diff --git a/records/sd1/sweep/6346bd3a62192d334487/output.webp b/records/sd1/sweep/6346bd3a62192d334487/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d130cc41df866a93f8f7c2a8de9c78167587a203
--- /dev/null
+++ b/records/sd1/sweep/6346bd3a62192d334487/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68eac84a2cdeddcd1ec029fb0606670ba314715ae692b821d465a36a42d1c453
+size 43880
diff --git a/records/sd1/sweep/634b3cf643a2fb2347ab/output.webp b/records/sd1/sweep/634b3cf643a2fb2347ab/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..875557476be0caba91524d3400ddc268b485add9
--- /dev/null
+++ b/records/sd1/sweep/634b3cf643a2fb2347ab/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:740182e11626088e462cc1c408f461fbb5072f14b09b715ca6762abae41cc808
+size 52742
diff --git a/records/sd1/sweep/634e2a4a2875cc8baf11/output.webp b/records/sd1/sweep/634e2a4a2875cc8baf11/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..97cf109603cb9f999ea909a3bc1e2a61c0cb0aed
--- /dev/null
+++ b/records/sd1/sweep/634e2a4a2875cc8baf11/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4300bb34d5c92d0429ba1fbbdc13e44455efbc94a4baf82d8d7799c265f482f4
+size 33846
diff --git a/records/sd1/sweep/6358623409eb1d38146f/output.webp b/records/sd1/sweep/6358623409eb1d38146f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e82e5da465162d3df994344c980e34d49dbab633
--- /dev/null
+++ b/records/sd1/sweep/6358623409eb1d38146f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57d6288751c11ba2575b3b9ee1cad50137b1084e1c0016cad584f3571c0d2540
+size 91560
diff --git a/records/sd1/sweep/636155c8b16199f4ddb0/output.webp b/records/sd1/sweep/636155c8b16199f4ddb0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..578295412ac27dbca475d289049a05fe95ecdea1
--- /dev/null
+++ b/records/sd1/sweep/636155c8b16199f4ddb0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05200db80c843a5a4989c30e63a79124dc6a13d1e02fbc8bac9a614025d46b17
+size 13666
diff --git a/records/sd1/sweep/63632e43fb6f2745c8a6/output.webp b/records/sd1/sweep/63632e43fb6f2745c8a6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f51b5bd308715dbfe1a5bba859a558868e7c3998
--- /dev/null
+++ b/records/sd1/sweep/63632e43fb6f2745c8a6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f14089a7d64ac8d9f20c50efd1076d60cf32bc834f514c90d2eff1987228cade
+size 21030
diff --git a/records/sd1/sweep/63697cbfc7715aff5d5c/output.webp b/records/sd1/sweep/63697cbfc7715aff5d5c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..207277d71b2e544e43723c6a466ebc35cf51b6a7
--- /dev/null
+++ b/records/sd1/sweep/63697cbfc7715aff5d5c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:729d77a20f0304130f0bf412a65df0a36fad642c14fbcd1143179f0fc8b8f1c0
+size 30584
diff --git a/records/sd1/sweep/63779e93cb6211bfdf69/output.webp b/records/sd1/sweep/63779e93cb6211bfdf69/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7828fbb11972f1576f3208f16c29bad62b83f887
--- /dev/null
+++ b/records/sd1/sweep/63779e93cb6211bfdf69/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa484d423d9c04759eb3850151fda00750b6e576d30dfe45f7652684412a7071
+size 3490
diff --git a/records/sd1/sweep/637829d71cf8a8892cc7/output.webp b/records/sd1/sweep/637829d71cf8a8892cc7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..da10b8a25f9f132ef7ad224b5037f55eed07c1eb
--- /dev/null
+++ b/records/sd1/sweep/637829d71cf8a8892cc7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b02b0b8b15b9fb45db15dce989643aa0a5ba2ebbbfe40b8b1fb2a025ca0c5913
+size 50694
diff --git a/records/sd1/sweep/637ac605c923ec96aadb/output.webp b/records/sd1/sweep/637ac605c923ec96aadb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b56a6b5b1beda24c13676306db559c05844004b4
--- /dev/null
+++ b/records/sd1/sweep/637ac605c923ec96aadb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b3cf9abc4221f2e5cd9f1279b22a90c4671af22bfe5cab970c9d35c14e6f795
+size 45126
diff --git a/records/sd1/sweep/637aed684434b6a05e84/output.webp b/records/sd1/sweep/637aed684434b6a05e84/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..aeec70c1da75be7e171879c8e03e0eda567ae039
--- /dev/null
+++ b/records/sd1/sweep/637aed684434b6a05e84/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01d37b019af781db52fbb5402fdc9d93963f51fde9c3004ebbad6b4eda79e99e
+size 55896
diff --git a/records/sd1/sweep/637be01c3930bf653203/output.webp b/records/sd1/sweep/637be01c3930bf653203/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..274151a19c6c7aed2bd4c0c393900b4c1a4e4c59
--- /dev/null
+++ b/records/sd1/sweep/637be01c3930bf653203/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd1f05361f671ad0fdba6d6344d5170d1014cbad6aba76f237711493c230a6fa
+size 22608
diff --git a/records/sd1/sweep/6384984c6a366d316eb0/output.webp b/records/sd1/sweep/6384984c6a366d316eb0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cd6ea4a7c8dd6bad14a7cc43660da272ec19aaa8
--- /dev/null
+++ b/records/sd1/sweep/6384984c6a366d316eb0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66334935050e4f1e05db0963a8b6183f89f1c7f6c86bc5652c9a26bf2b1e0117
+size 33254
diff --git a/records/sd1/sweep/6389c7869abfbe91b57b/output.webp b/records/sd1/sweep/6389c7869abfbe91b57b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a07e16ff73796bc96508a0e093a40889fc45ab02
--- /dev/null
+++ b/records/sd1/sweep/6389c7869abfbe91b57b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55c5e5cf058cce4b4a35cc60d9f147476074aaa8b15580cf5cb82604e991c682
+size 40214
diff --git a/records/sd1/sweep/638a8881f5c8376f41a4/output.webp b/records/sd1/sweep/638a8881f5c8376f41a4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e9dcf4c09cd2675f5a06ab8cd914e036fa87251e
--- /dev/null
+++ b/records/sd1/sweep/638a8881f5c8376f41a4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22af7fe741981f44bacf5e48c54d97deea427361630d188bed87bf71041d09cb
+size 23966
diff --git a/records/sd1/sweep/63924c646966d5a079fe/output.webp b/records/sd1/sweep/63924c646966d5a079fe/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..88731adb132259c7c785054b92392a8ae7c08562
--- /dev/null
+++ b/records/sd1/sweep/63924c646966d5a079fe/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32de541ae6137cd186cfc3fa043ab64a162520589e14acfa2878c7e8ce2cd63a
+size 22244
diff --git a/records/sd1/sweep/6395d12c17e577125915/output.webp b/records/sd1/sweep/6395d12c17e577125915/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6e5d3da19dd9125b2fcd3d122e88b5c53c4d65ed
--- /dev/null
+++ b/records/sd1/sweep/6395d12c17e577125915/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4e6556e9cb73654b46e4e0340a6c575679798b8a1f047fde2962107bd4a3eec
+size 42254
diff --git a/records/sd1/sweep/639bcc24a4646df5e595/output.webp b/records/sd1/sweep/639bcc24a4646df5e595/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..445aea142afcb3341bdfdb4954ed30354a6e55bf
--- /dev/null
+++ b/records/sd1/sweep/639bcc24a4646df5e595/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3008a4593bd93ed9dd4135bca3fedf52ff7d7490bad8e5c25a43b12413161155
+size 51856
diff --git a/records/sd1/sweep/639c64d15708c953cb37/output.webp b/records/sd1/sweep/639c64d15708c953cb37/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ac79ee501e23b214d6282d59ec462544446caf6a
--- /dev/null
+++ b/records/sd1/sweep/639c64d15708c953cb37/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27c065a1b3be3391c8add8889bfc2e76d4ff418ce9c1564f9c32d730a17dac79
+size 31204
diff --git a/records/sd1/sweep/63a153525fbccfbc1c90/output.webp b/records/sd1/sweep/63a153525fbccfbc1c90/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ee69f2f1b7f58d9ec72107a596b43e3946f5fa3a
--- /dev/null
+++ b/records/sd1/sweep/63a153525fbccfbc1c90/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35f7c641ed31785f252320a289a95f33e8651e8f68c825ea57d0ce8024ef32b7
+size 27800
diff --git a/records/sd1/sweep/63ab8062e1c5af1d27cf/output.webp b/records/sd1/sweep/63ab8062e1c5af1d27cf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..de290e2b07ec51bb061024b508af1ebe047732e6
--- /dev/null
+++ b/records/sd1/sweep/63ab8062e1c5af1d27cf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83a69e6fafacb3d683cc166cb993d8d541fb2b276ed64d88a768e8777e8195c7
+size 36928
diff --git a/records/sd1/sweep/63b023f4f2c9835f569c/output.webp b/records/sd1/sweep/63b023f4f2c9835f569c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..930e3e6867a4b695cf67f097ab51026412541ab3
--- /dev/null
+++ b/records/sd1/sweep/63b023f4f2c9835f569c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f17cadf68de0aacb28b25abbe8b33d95e5a6b82a79a8b13efbd85f1fa45c512
+size 36424
diff --git a/records/sd1/sweep/63be36476b43fc7a28d4/output.webp b/records/sd1/sweep/63be36476b43fc7a28d4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d0d7742e0e9262d643e2f13632e41d4f2c3b2db0
--- /dev/null
+++ b/records/sd1/sweep/63be36476b43fc7a28d4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d9d1d2979b8730d53bf91e0885b69b5422811a76696760fd29316de51618450
+size 47542
diff --git a/records/sd1/sweep/63c9ed43a74e5306e836/output.webp b/records/sd1/sweep/63c9ed43a74e5306e836/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fdc58a5ac9e1ef15bba0f10438f04f3b0f5167db
--- /dev/null
+++ b/records/sd1/sweep/63c9ed43a74e5306e836/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec784713bbd3fdc72f48f1c59f319d592ce3de8c25deed72145ae5876bc07d06
+size 35370
diff --git a/records/sd1/sweep/63caff795d3863058690/output.webp b/records/sd1/sweep/63caff795d3863058690/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6407a417767dd3e8437d0fcab6244e5df2c71167
--- /dev/null
+++ b/records/sd1/sweep/63caff795d3863058690/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:478efe989cdea92732f8ddbbe876560286e3f79c0d01e3544abde534b8c98efd
+size 26438
diff --git a/records/sd1/sweep/63cd063363ca6ef1282a/output.webp b/records/sd1/sweep/63cd063363ca6ef1282a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fe96c12b2d3d104572746a5aa5fb8260a98d8277
--- /dev/null
+++ b/records/sd1/sweep/63cd063363ca6ef1282a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61065816fe53df83c2c7085a07c598762f6f4dbe387c5cb7696676513eb8fc6b
+size 105292
diff --git a/records/sd1/sweep/63deec2bdef1fec08d12/output.webp b/records/sd1/sweep/63deec2bdef1fec08d12/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..680a10a5498d5ca0596077d1288bca9e67a9c485
--- /dev/null
+++ b/records/sd1/sweep/63deec2bdef1fec08d12/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:be7082f403ccb409e29dd6b9b01ecf30fcf19fc3fdd65dace496ffdbd90fad94
+size 36504
diff --git a/records/sd1/sweep/63df109e6b93ae816a37/output.webp b/records/sd1/sweep/63df109e6b93ae816a37/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..646fbeee821890e667ed4308e1735441a00b5f56
--- /dev/null
+++ b/records/sd1/sweep/63df109e6b93ae816a37/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3c4d1b905959adb15b6e90e3d7fe972bbf1f371d9e3e296e9446e66478c8c54
+size 61332
diff --git a/records/sd1/sweep/63e53bae80dc4dd2f478/output.webp b/records/sd1/sweep/63e53bae80dc4dd2f478/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9113c88e5b3ea8e53d70e35c721794dd419fd15d
--- /dev/null
+++ b/records/sd1/sweep/63e53bae80dc4dd2f478/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:24dd1738884c594f78ef6409193f5355517e18dca8cdb932e6f15b58a5a72ee5
+size 49422
diff --git a/records/sd1/sweep/63ec3517aba32ec0da87/output.webp b/records/sd1/sweep/63ec3517aba32ec0da87/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6a16e9049f8154aedb899b95bdff57f3862c0773
--- /dev/null
+++ b/records/sd1/sweep/63ec3517aba32ec0da87/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af78348d3b7af9077a69770d279bc222199bcf502b948b8a7ec26ab0d4fcb15c
+size 16552
diff --git a/records/sd1/sweep/63f8ff4fde4ba74e0793/output.webp b/records/sd1/sweep/63f8ff4fde4ba74e0793/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dcd3f710c8a667d31b84e359a8c3db9fd60e7732
--- /dev/null
+++ b/records/sd1/sweep/63f8ff4fde4ba74e0793/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fc22efbebf3bfa8b6b446c211921ebfa9519203c9bb5789066374559d1b624b
+size 14930
diff --git a/records/sd1/sweep/640e5d3fc29b7aa2b4f2/output.webp b/records/sd1/sweep/640e5d3fc29b7aa2b4f2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..69c9b852cc127382ea3f1dabe399deb8a141f69d
--- /dev/null
+++ b/records/sd1/sweep/640e5d3fc29b7aa2b4f2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18a7cf42b45379c1966d2c4e53099a442d4af7515c91aff197d6a713fb7558dc
+size 25148
diff --git a/records/sd1/sweep/6410a138beea8373b78b/output.webp b/records/sd1/sweep/6410a138beea8373b78b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f3cc0bda73e8addf00079b6e3057283e5ab55288
--- /dev/null
+++ b/records/sd1/sweep/6410a138beea8373b78b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b3829aeacf263895e15bd90cbe15feaa98da6eab7fc00971761bd86fec442e4
+size 37058
diff --git a/records/sd1/sweep/6416a3d92726cbe72a94/output.webp b/records/sd1/sweep/6416a3d92726cbe72a94/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..51b6191ceac7b70008f25907078d051071d44f90
--- /dev/null
+++ b/records/sd1/sweep/6416a3d92726cbe72a94/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31077d9cda0a2ff0fd06b32d0a62a1969de38ad7ed133fb8e9a2c4d2e10fbf2a
+size 19536
diff --git a/records/sd1/sweep/64191a8791523be2e99c/output.webp b/records/sd1/sweep/64191a8791523be2e99c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..797b04d859b8675d4d2e64198016f2802a0f5404
--- /dev/null
+++ b/records/sd1/sweep/64191a8791523be2e99c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94e68198182ea3e3e91aaa08aa929468a54204aa6b930f0dc77a8e9bccf740e7
+size 33098
diff --git a/records/sd1/sweep/641d4076a943268874e1/output.webp b/records/sd1/sweep/641d4076a943268874e1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c91e8231464e17c7f104fe7bc62952228f9fa7b9
--- /dev/null
+++ b/records/sd1/sweep/641d4076a943268874e1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eac1b33376f35bcf75df809aa5ccb340943b938dea8c6f614be052ebe9014ca5
+size 19932
diff --git a/records/sd1/sweep/64246d752e48cc99664a/output.webp b/records/sd1/sweep/64246d752e48cc99664a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0d0d026f7596a4ddf1a0f9af4edca2593557bfdb
--- /dev/null
+++ b/records/sd1/sweep/64246d752e48cc99664a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cda93df2ac35ed0cd89accd5097d05d53f34c633c2caef20fdde21c0514ae8e7
+size 17130
diff --git a/records/sd1/sweep/642d164a5a607a2b3619/output.webp b/records/sd1/sweep/642d164a5a607a2b3619/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f3055b112a2492e0f8fd0e18669d7009d172e436
--- /dev/null
+++ b/records/sd1/sweep/642d164a5a607a2b3619/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98cfa9811f70cc8482c5c23dc5588fcbe1d7f93087a96595bfdcd0edc4bdc60b
+size 35896
diff --git a/records/sd1/sweep/6441927f2a9bc81941e9/output.webp b/records/sd1/sweep/6441927f2a9bc81941e9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2788b0f895ead8cf811deed26f8fb5cf0587fa3f
--- /dev/null
+++ b/records/sd1/sweep/6441927f2a9bc81941e9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d109f4382cc980bf1616b7c3606f9c06f74f21f10f1f57f0ccff3542089348ac
+size 35894
diff --git a/records/sd1/sweep/64463e28caacf13adddd/output.webp b/records/sd1/sweep/64463e28caacf13adddd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fb02327e0b7f9c46da97a7a4087d4d021bdd2845
--- /dev/null
+++ b/records/sd1/sweep/64463e28caacf13adddd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4cedf664ecb1a1e3b38f52da01508e52f04383d01552825e535066a924951e1
+size 47776
diff --git a/records/sd1/sweep/6449ae5cab7129fb7ca8/output.webp b/records/sd1/sweep/6449ae5cab7129fb7ca8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1edd1fee7ff3e054da15f72539fafae12a12fe37
--- /dev/null
+++ b/records/sd1/sweep/6449ae5cab7129fb7ca8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cc6cc5480f0d1c40597e8824c146f6801ef7b303074b25b184a26b2f8dd51c3
+size 35826
diff --git a/records/sd1/sweep/644a5034d4b01c1d3f01/output.webp b/records/sd1/sweep/644a5034d4b01c1d3f01/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e4d5d57bdfd275ff7ea0aeb161326e43f2bfcdd5
--- /dev/null
+++ b/records/sd1/sweep/644a5034d4b01c1d3f01/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f41d833126ff7f8de78c534d8f57ff521fe4d58616bf3c00fe213c210d532d7
+size 53398
diff --git a/records/sd1/sweep/645f8ee1b71eeb0692c2/output.webp b/records/sd1/sweep/645f8ee1b71eeb0692c2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..577fba640dfd922050208af919ecc06279c9a230
--- /dev/null
+++ b/records/sd1/sweep/645f8ee1b71eeb0692c2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd141fcd674e37eed69c8889649875465947cc154bb9f1db078fa207dab62e3b
+size 19744
diff --git a/records/sd1/sweep/646d7f173096955db325/output.webp b/records/sd1/sweep/646d7f173096955db325/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..84454d5a10ce1ca621f204bd0b7d2c8cbf2ca8aa
--- /dev/null
+++ b/records/sd1/sweep/646d7f173096955db325/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87eff2b0c5255f6eba4dbb1a358c37554b790aa6037a21f45fb25c55c246949b
+size 35516
diff --git a/records/sd1/sweep/646f1e7c80f79a82a196/output.webp b/records/sd1/sweep/646f1e7c80f79a82a196/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..984dad45ab7ba8b3627dadb2e189c3a9cc16a312
--- /dev/null
+++ b/records/sd1/sweep/646f1e7c80f79a82a196/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:104da9ec8f73028ad8fe313eca6779185622fb2d4b33690e1d9f2a9a7e93dfc1
+size 21096
diff --git a/records/sd1/sweep/646f5d6fcb0b6688fd20/output.webp b/records/sd1/sweep/646f5d6fcb0b6688fd20/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4d3c349b6370222e0d950a6f1d32c45e0c130980
--- /dev/null
+++ b/records/sd1/sweep/646f5d6fcb0b6688fd20/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab0cf1c676d816dd936b80b21bb71b9d6443f88085fe3c81927cf5afb0550248
+size 37894
diff --git a/records/sd1/sweep/64724253cf82ba6ec2d1/output.webp b/records/sd1/sweep/64724253cf82ba6ec2d1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a300b7a734546b8c9d2fbc3d6932ad88eca2ba81
--- /dev/null
+++ b/records/sd1/sweep/64724253cf82ba6ec2d1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e715d75e04dd521c22af2533c5334ad99b4fd9e9364bd937ebd7244e4b41d03
+size 36046
diff --git a/records/sd1/sweep/6474fa4d6d55f42a65f9/output.webp b/records/sd1/sweep/6474fa4d6d55f42a65f9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..57f926772efc66a1d424422a498b27d60dd20d1a
--- /dev/null
+++ b/records/sd1/sweep/6474fa4d6d55f42a65f9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a94a8d1af4bea52f677e3254c0d8301c347177712f352b7cabb2d43b29f3ef69
+size 53836
diff --git a/records/sd1/sweep/647807ba2d35a8175fd6/output.webp b/records/sd1/sweep/647807ba2d35a8175fd6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..369f15e6f2806f070e8a6cf6e53361957baf020a
--- /dev/null
+++ b/records/sd1/sweep/647807ba2d35a8175fd6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6e67fe8b265b030675caf225d0a3eab8c88c858cb2ed7d23380d538cd7b3b98
+size 19930
diff --git a/records/sd1/sweep/6484c8908a784924e187/output.webp b/records/sd1/sweep/6484c8908a784924e187/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6f98634eeedfabc5864ecdfa54224964cf69ddba
--- /dev/null
+++ b/records/sd1/sweep/6484c8908a784924e187/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7162c8b0d264a60cd5526ddbe4c7cfffc9801a88e4def7d690d77acac527db0
+size 22586
diff --git a/records/sd1/sweep/648521ef44d52025856b/output.webp b/records/sd1/sweep/648521ef44d52025856b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..90fb2c7f551cb645291d02fd73ce9643dea41919
--- /dev/null
+++ b/records/sd1/sweep/648521ef44d52025856b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a23bc783939f22360477102ab4f56f3a8ca41ae3929eebbea5919fdd435637f
+size 15810
diff --git a/records/sd1/sweep/6489e9279b73b8e77533/output.webp b/records/sd1/sweep/6489e9279b73b8e77533/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c57bf073a1bdbd28bf1f4f1f484d9ae24feb2d53
--- /dev/null
+++ b/records/sd1/sweep/6489e9279b73b8e77533/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73bba67ed6538a6814f2b4984c9f349e2bc198949d17225245ca3fe0cc6632b4
+size 40662
diff --git a/records/sd1/sweep/649c6d979a23a3d644e6/output.webp b/records/sd1/sweep/649c6d979a23a3d644e6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..16cb77856f01027bbc2b619b57151ce9ba092ff2
--- /dev/null
+++ b/records/sd1/sweep/649c6d979a23a3d644e6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d35893f4548735046f2ef67a3c3487d9a61b1cffeca109577606e32cbbc9300
+size 16942
diff --git a/records/sd1/sweep/64a15808cb47fcef7771/output.webp b/records/sd1/sweep/64a15808cb47fcef7771/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..388b71cb552c852d80fb961f570656acdbc652d3
--- /dev/null
+++ b/records/sd1/sweep/64a15808cb47fcef7771/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0017644c8bab7d699992f8bfc2359e613c552550f848761b92061b6f3119fd9d
+size 31296
diff --git a/records/sd1/sweep/64a4bb173d1fdaf154ae/output.webp b/records/sd1/sweep/64a4bb173d1fdaf154ae/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..67351e023203cd42c38a1642c6ede32d2d19aada
--- /dev/null
+++ b/records/sd1/sweep/64a4bb173d1fdaf154ae/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d9dea0f9dc7fb2b5c06b79ea661e7512c71e39c06b6c4d556a5f690e44a857e
+size 49332
diff --git a/records/sd1/sweep/64b9892cd69eb8b7a8fa/output.webp b/records/sd1/sweep/64b9892cd69eb8b7a8fa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c8ae85c360149f16226a900941df3690ce5c62f1
--- /dev/null
+++ b/records/sd1/sweep/64b9892cd69eb8b7a8fa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05b2a06940e3adf895d37dc572e5b6321606555dcddffbb64eb8a3f14278c0d9
+size 31690
diff --git a/records/sd1/sweep/64b9bf0d5229a15655ca/output.webp b/records/sd1/sweep/64b9bf0d5229a15655ca/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f2803c9c5b785ecbd011701ca1d3fb43be4dd06d
--- /dev/null
+++ b/records/sd1/sweep/64b9bf0d5229a15655ca/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:055eab35d7798e6401acfec6b0319bb053f942d56bea27d7f2c20f8cf5f74073
+size 32208
diff --git a/records/sd1/sweep/64b9d709f44f72ad3993/output.webp b/records/sd1/sweep/64b9d709f44f72ad3993/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..06c7f3930f3eb0c69bf55f79c9a303237c86ed03
--- /dev/null
+++ b/records/sd1/sweep/64b9d709f44f72ad3993/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10c8ebdf701ac4a70f999a8137da914f0d788ed6296b32b4c4bd96f3eff2d086
+size 4102
diff --git a/records/sd1/sweep/64bdb06bff16ca086f1d/output.webp b/records/sd1/sweep/64bdb06bff16ca086f1d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f1e1db7da9bc9f59b2e06ba8d0d734e1a2a1ee45
--- /dev/null
+++ b/records/sd1/sweep/64bdb06bff16ca086f1d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e2bd78464510818eda34f3d6fc2377ec6f6e828a27da2b67b156696f0648934
+size 29890
diff --git a/records/sd1/sweep/64bf2643e10413b597aa/output.webp b/records/sd1/sweep/64bf2643e10413b597aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ca12e81478a66c816835059bae5dbe3553527bf7
--- /dev/null
+++ b/records/sd1/sweep/64bf2643e10413b597aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc1fbfe25ffe8d751b87159ead367e684dd0023bc3c3e5362b66f56d5e0079a8
+size 28330
diff --git a/records/sd1/sweep/64cb07915497d7e07e2a/output.webp b/records/sd1/sweep/64cb07915497d7e07e2a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c4f1ed631758e714d12edb7f14b537116ac85441
--- /dev/null
+++ b/records/sd1/sweep/64cb07915497d7e07e2a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3017ea4bae860ae01362a1f43fca11d56995954c236b459ba9d19c4b8ad58a35
+size 16616
diff --git a/records/sd1/sweep/64d041132019d9db1a19/output.webp b/records/sd1/sweep/64d041132019d9db1a19/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9fd39c76cb892fff675c4aeaac275b2448dc7998
--- /dev/null
+++ b/records/sd1/sweep/64d041132019d9db1a19/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d86173b3508e35dc3a8dd9f07018b523e311a45c2e23d7f78aee8e64831c248
+size 32024
diff --git a/records/sd1/sweep/64d24bb8638fb270a26f/output.webp b/records/sd1/sweep/64d24bb8638fb270a26f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..22b6244ec6851f166d8278902565d2c0ae8fa17c
--- /dev/null
+++ b/records/sd1/sweep/64d24bb8638fb270a26f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:891f8c27901ea74d2b74015c569c739b0da1f3e432fd365181e4fc9449aeb2e4
+size 54670
diff --git a/records/sd1/sweep/64d40feb9d90920a40c9/output.webp b/records/sd1/sweep/64d40feb9d90920a40c9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..068c26b1831b6c32b3ef56b80f9ae7969ffa67f0
--- /dev/null
+++ b/records/sd1/sweep/64d40feb9d90920a40c9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b53a487117c7a859bf42dca5ae86c4cd30c00ccae593df9da812db05f601d6e
+size 35564
diff --git a/records/sd1/sweep/64e6d5fe0ffed857232b/output.webp b/records/sd1/sweep/64e6d5fe0ffed857232b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..842cac795106d82f05bc56dab7c3696c2bb5ef73
--- /dev/null
+++ b/records/sd1/sweep/64e6d5fe0ffed857232b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed65ae5c09274b36485321866ebb43da2a9ab27e9d9a6cf39bbd1b9f41cc3036
+size 18678
diff --git a/records/sd1/sweep/64e86cdbb2c3477f5879/output.webp b/records/sd1/sweep/64e86cdbb2c3477f5879/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fd1e136fbd7cc61171f89c23bc814cf912e8a7b5
--- /dev/null
+++ b/records/sd1/sweep/64e86cdbb2c3477f5879/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e46caeb60802c9a9d9bff4e532a3e1b07794d369b65954caf4347c6cd2b8c5b
+size 41662
diff --git a/records/sd1/sweep/64ee8898c0e9af0825a2/output.webp b/records/sd1/sweep/64ee8898c0e9af0825a2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..73046ad046153c424dad44d9ef2157e1c11e64db
--- /dev/null
+++ b/records/sd1/sweep/64ee8898c0e9af0825a2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b4c15512e539ba616f2932ee47616f9a5a7b894d05875f3f857f047b56974c9
+size 16700
diff --git a/records/sd1/sweep/64eee0e1302a997b4319/output.webp b/records/sd1/sweep/64eee0e1302a997b4319/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7b1e7d149dc9163574a9d558ec0a1c20d977e389
--- /dev/null
+++ b/records/sd1/sweep/64eee0e1302a997b4319/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0073df169378b05f740c6be4de90a2293b9e9a331876ff64ea62d8387c00ac9
+size 24342
diff --git a/records/sd1/sweep/6500119f479ee0964041/output.webp b/records/sd1/sweep/6500119f479ee0964041/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ae313199595647d6b01b0f75dbb6b83a8878357b
--- /dev/null
+++ b/records/sd1/sweep/6500119f479ee0964041/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aef7b353946b125412afd31a42462e9083f5493011e7e2300578b2f73dccbae1
+size 54570
diff --git a/records/sd1/sweep/65012406cf2e46dd7c5c/output.webp b/records/sd1/sweep/65012406cf2e46dd7c5c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0a0613d98cdbf153adfe0a3fef554eee64d2e357
--- /dev/null
+++ b/records/sd1/sweep/65012406cf2e46dd7c5c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10c61bda0c4157ea209d195c821cdaa2e049273f1499357fade375c775747626
+size 33436
diff --git a/records/sd1/sweep/650cd901ce597d8597a3/output.webp b/records/sd1/sweep/650cd901ce597d8597a3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fb23e68f6ed9b6e3884d0144c6e1c8eec78ce004
--- /dev/null
+++ b/records/sd1/sweep/650cd901ce597d8597a3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:866b041f9f3e373d1aac5e80a1803942fa6dac314349537b6f568f141c8bb71c
+size 19932
diff --git a/records/sd1/sweep/6523ef8e15b8ac27ba38/output.webp b/records/sd1/sweep/6523ef8e15b8ac27ba38/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1e35e6886d0430bfb465d819c6414fcf58b4954d
--- /dev/null
+++ b/records/sd1/sweep/6523ef8e15b8ac27ba38/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d05210c115168ebeea135893fb89c45427540145cdce05c637a6df14aa0a6c7f
+size 34594
diff --git a/records/sd1/sweep/6535f5c2c349fb9cd28a/output.webp b/records/sd1/sweep/6535f5c2c349fb9cd28a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..56581834e35eb7fd42e1a518efe57704f4acde45
--- /dev/null
+++ b/records/sd1/sweep/6535f5c2c349fb9cd28a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cc50ab7e1accdf02a3123dbcb08ae9694a9113ef17ef9341e25f7f82a9bbc7f
+size 55716
diff --git a/records/sd1/sweep/65395c3e625885d385b1/output.webp b/records/sd1/sweep/65395c3e625885d385b1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..601196ac91afe4cd8d834ae09116e0a96115c584
--- /dev/null
+++ b/records/sd1/sweep/65395c3e625885d385b1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:187c40cdfb6f6ae0ddce8014936a3f14436136c19475840a80c60aded4bc231b
+size 19004
diff --git a/records/sd1/sweep/65418a3aa7a60bd94db9/output.webp b/records/sd1/sweep/65418a3aa7a60bd94db9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2a220fdd5f52c436a7239732696ad61ea3d0350f
--- /dev/null
+++ b/records/sd1/sweep/65418a3aa7a60bd94db9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:002b00fd4a0d6d261c4a5c54b42d8838f848b388224bf768aa597e86d006efcb
+size 63450
diff --git a/records/sd1/sweep/65445e495d2ede6b69a1/output.webp b/records/sd1/sweep/65445e495d2ede6b69a1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..08f27d8f72992fdeee16ea39458d3aeae943d7bd
--- /dev/null
+++ b/records/sd1/sweep/65445e495d2ede6b69a1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04ae38787453014abb1447f2ac090a7cbaf1216e2c2ad2d5850ee0b82d36c6e8
+size 33104
diff --git a/records/sd1/sweep/6546a4de334cbd7d68ac/output.webp b/records/sd1/sweep/6546a4de334cbd7d68ac/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c697ee562b882d62118951f2ecbfe0fc7c750251
--- /dev/null
+++ b/records/sd1/sweep/6546a4de334cbd7d68ac/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:30d41b09767d1bf29bfee236954c760a04df0bda4d8b8b4684a56471913b66ea
+size 47240
diff --git a/records/sd1/sweep/6547494445555d24c7ca/output.webp b/records/sd1/sweep/6547494445555d24c7ca/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..72427fcf15734811e60ceb11734e0a36169ac04f
--- /dev/null
+++ b/records/sd1/sweep/6547494445555d24c7ca/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ba468d9dc2c4687aeadbacb92bdf54bdff00656708d838f0cce30c04a0032e3
+size 27638
diff --git a/records/sd1/sweep/6554a08f5d132064a23a/output.webp b/records/sd1/sweep/6554a08f5d132064a23a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c861db2491cac180183e3a1ff33557de12ccdb9c
--- /dev/null
+++ b/records/sd1/sweep/6554a08f5d132064a23a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4e0ffe971b224831b8946c5c9b8c54b10e025f114d503c40ad045259af91bcc
+size 42818
diff --git a/records/sd1/sweep/6559856bfe01fd50a771/output.webp b/records/sd1/sweep/6559856bfe01fd50a771/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d2581501c06105541c76a1ff6e184faec8d7841a
--- /dev/null
+++ b/records/sd1/sweep/6559856bfe01fd50a771/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83f1013ab5af4513269ff0d71c8c9cebbf6abdd41a69b3ea8f84cf1d573b0127
+size 48016
diff --git a/records/sd1/sweep/6569441f8076779938c3/output.webp b/records/sd1/sweep/6569441f8076779938c3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f413920dc2bef2df637755f4f8c293014796a239
--- /dev/null
+++ b/records/sd1/sweep/6569441f8076779938c3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5de5e77647ce9e3f415d1ad3fac60afa47829de213c3a294a2e005bb4c6a6181
+size 36500
diff --git a/records/sd1/sweep/658429851950216d6f62/output.webp b/records/sd1/sweep/658429851950216d6f62/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a130a9ff9b79990751f153949b2c79348d9cdb61
--- /dev/null
+++ b/records/sd1/sweep/658429851950216d6f62/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c42d48e3a5f7373cd86d75b20e53ec33be685c72fd2102f57a68d5b5ca03ba24
+size 26482
diff --git a/records/sd1/sweep/6585ee4c5db1bd731987/output.webp b/records/sd1/sweep/6585ee4c5db1bd731987/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e188caec1a6c9773732881da973417eddcfd4439
--- /dev/null
+++ b/records/sd1/sweep/6585ee4c5db1bd731987/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2483d5d3d2a7526bf62054c4b5c88b4add0e45f7795471e0f174ce37661ca45
+size 29480
diff --git a/records/sd1/sweep/65930957b5f31bb85dd5/output.webp b/records/sd1/sweep/65930957b5f31bb85dd5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3bb8fe839f72931ff93eaed9c20116ddbc2c7f83
--- /dev/null
+++ b/records/sd1/sweep/65930957b5f31bb85dd5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb01bde4852ff520dc6debc5d05768a0e8954cc2fa6ff4fb5f9105df1135f580
+size 33888
diff --git a/records/sd1/sweep/659b3e3a0f4cd95591bf/output.webp b/records/sd1/sweep/659b3e3a0f4cd95591bf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c58aeeb09d4d8fa6c49d4836e6648a81729ff37a
--- /dev/null
+++ b/records/sd1/sweep/659b3e3a0f4cd95591bf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82a4626df5efca205fe0f96299e5b0fea097b5c6034ad691b69503950a05b9af
+size 24970
diff --git a/records/sd1/sweep/659b691b34ae3391d052/output.webp b/records/sd1/sweep/659b691b34ae3391d052/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..30d46661974bfa218694357886234faf23fb58da
--- /dev/null
+++ b/records/sd1/sweep/659b691b34ae3391d052/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6b3aba51f476996c24595f8c11daa9709c947a92a81c7482178d9b7372b7fe1
+size 56846
diff --git a/records/sd1/sweep/65a57707b39213849ed8/output.webp b/records/sd1/sweep/65a57707b39213849ed8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..432c0d49f69b75c2671658cf8706a691fa363dcc
--- /dev/null
+++ b/records/sd1/sweep/65a57707b39213849ed8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2272ab9516eb58d3038cef9178a365e3d266ce02227620451239b96a2295b0a4
+size 22832
diff --git a/records/sd1/sweep/65ab9d9e964a766fcefc/output.webp b/records/sd1/sweep/65ab9d9e964a766fcefc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..312678573a1603c93b3baeef52559b85fcd18e96
--- /dev/null
+++ b/records/sd1/sweep/65ab9d9e964a766fcefc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cf6402e31ded252d3dd34a13352fbdd7f947229e8f0c7eeb0c61c841ad1df96
+size 29902
diff --git a/records/sd1/sweep/65ba1b146d393acd6a20/output.webp b/records/sd1/sweep/65ba1b146d393acd6a20/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a4c7dbe1dd684790daf130c3b7954eb58819e004
--- /dev/null
+++ b/records/sd1/sweep/65ba1b146d393acd6a20/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d2ad9fe21153fb7c9acdfc4bf2158df27472d281c273b7b9a105a1312f4d965
+size 26250
diff --git a/records/sd1/sweep/65bbd5aee8d354393c1e/output.webp b/records/sd1/sweep/65bbd5aee8d354393c1e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7cda6f911301b884eb6ff5da70f94780b4dd0854
--- /dev/null
+++ b/records/sd1/sweep/65bbd5aee8d354393c1e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b7fe9834d1860df8d5edbc497dd6c24b3d123113a9209079d6d13f70175a853
+size 31794
diff --git a/records/sd1/sweep/65c19ce552cd4d9c33ce/output.webp b/records/sd1/sweep/65c19ce552cd4d9c33ce/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3c66ad491e64bfa83cdf22fb611812d232bc8b1c
--- /dev/null
+++ b/records/sd1/sweep/65c19ce552cd4d9c33ce/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:166f324b4ff4f46a6f7e702e341929978c7eec75e40104c58d3e1e0d0f127009
+size 49236
diff --git a/records/sd1/sweep/65c26b789a1384dbf38a/output.webp b/records/sd1/sweep/65c26b789a1384dbf38a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0d992620009e5e8b75525681b10dffd5532b4d21
--- /dev/null
+++ b/records/sd1/sweep/65c26b789a1384dbf38a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58d8bb908cce92769677beff189034fcd48bec5970d9661cd13187da6dd4b65e
+size 18672
diff --git a/records/sd1/sweep/65ca0082ccd427ca00c9/output.webp b/records/sd1/sweep/65ca0082ccd427ca00c9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b77d14a999d5cda64ffe5b962b50e19d9a76ab8b
--- /dev/null
+++ b/records/sd1/sweep/65ca0082ccd427ca00c9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:567c570ad0b12297aaa21ff9e02f77c9ffa7e3bc19ed1e1c596c19f2e8561e41
+size 20614
diff --git a/records/sd1/sweep/65cf14228693d1c5237e/output.webp b/records/sd1/sweep/65cf14228693d1c5237e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8a4e105c8677f5c6326235df4aa91fd7bffdcfca
--- /dev/null
+++ b/records/sd1/sweep/65cf14228693d1c5237e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:baf4218fbb2f3ecadbc47787ed83b5b8b274d0ca9f45007ca8ced47e445a2705
+size 50680
diff --git a/records/sd1/sweep/65cf4a2bdd041b397295/output.webp b/records/sd1/sweep/65cf4a2bdd041b397295/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3dbfc9bab07ffdc60f3eeba391f99c0c14f986f5
--- /dev/null
+++ b/records/sd1/sweep/65cf4a2bdd041b397295/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c550c07e2d1b69cf72066c8dbbbfc72f891b696fe9b211c139b8bcb4aa35273
+size 36406
diff --git a/records/sd1/sweep/65d05b4400760800d5bc/output.webp b/records/sd1/sweep/65d05b4400760800d5bc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..28bf0397b656edfc327d878012a05e96955a83ef
--- /dev/null
+++ b/records/sd1/sweep/65d05b4400760800d5bc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b5598e2ea687bb9dea195741ac7c457160119cc05c9d9ef703145fadeb999cf
+size 36498
diff --git a/records/sd1/sweep/65d9158a26aaafad77b8/output.webp b/records/sd1/sweep/65d9158a26aaafad77b8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..623ae295f28af32c091f9536d456f984b42290ae
--- /dev/null
+++ b/records/sd1/sweep/65d9158a26aaafad77b8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeed650cadbbaec946bf1653e935917f2dcf627fc99ddf944da7a5cb275232a6
+size 25316
diff --git a/records/sd1/sweep/65decbb0fe2d4cbeede1/output.webp b/records/sd1/sweep/65decbb0fe2d4cbeede1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f1670a368a319aa616fe48ff5da275a7d68d3a25
--- /dev/null
+++ b/records/sd1/sweep/65decbb0fe2d4cbeede1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c7cff98b78e3fe572192f879075d8503ecde7472ae978b627979a2ae93b9208
+size 23702
diff --git a/records/sd1/sweep/65e8533bce4a7366c3db/output.webp b/records/sd1/sweep/65e8533bce4a7366c3db/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..df17288f70f3d4f840b3bbe89953ee20b536092d
--- /dev/null
+++ b/records/sd1/sweep/65e8533bce4a7366c3db/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dad56b30ecc474b6f68827f9a679977567b55ba7915a2723d2355a57ffab3316
+size 48924
diff --git a/records/sd1/sweep/65f088499114b20db8d1/output.webp b/records/sd1/sweep/65f088499114b20db8d1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4403c24d2561d4540d26cf3d2d99e083ae84c207
--- /dev/null
+++ b/records/sd1/sweep/65f088499114b20db8d1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:983f23ea4c0fd379e50087b6104e6b61e4ef3daafb75aa8bc47be808a47b9cb5
+size 49184
diff --git a/records/sd1/sweep/65fb11e9b2cdec73fdcf/output.webp b/records/sd1/sweep/65fb11e9b2cdec73fdcf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..03dea9d73053e9067be2127e38fca50b4d5d89a2
--- /dev/null
+++ b/records/sd1/sweep/65fb11e9b2cdec73fdcf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26843876777164cddeb74a1ff187c62a069292cac22de7697698660c0768e9e3
+size 34376
diff --git a/records/sd1/sweep/6602366721fc977556a7/output.webp b/records/sd1/sweep/6602366721fc977556a7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d4e044e7ef3909e209969879e6afeb00ec0470a2
--- /dev/null
+++ b/records/sd1/sweep/6602366721fc977556a7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87a21c4f77005b2925611b075b05139735ed008c03d3aff7d8d1cf10e2f20a18
+size 19544
diff --git a/records/sd1/sweep/66071e3bf21d5eed8518/output.webp b/records/sd1/sweep/66071e3bf21d5eed8518/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8aca58f9af2dc3ee91d30b191b4dfb5ba715c7f5
--- /dev/null
+++ b/records/sd1/sweep/66071e3bf21d5eed8518/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf46559062434a828d6dcafd0e26379e239175a6478aaab25e5618cc520b5800
+size 27088
diff --git a/records/sd1/sweep/66077a828012946af892/output.webp b/records/sd1/sweep/66077a828012946af892/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1bad229bad36d22f87e72afa5780a6111e6a13cf
--- /dev/null
+++ b/records/sd1/sweep/66077a828012946af892/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d01cb5d61112c207703759f069b4e5ffb47a4756201af2e298ce016482a13c91
+size 49250
diff --git a/records/sd1/sweep/660b81d3fbab286b7551/output.webp b/records/sd1/sweep/660b81d3fbab286b7551/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e99f12fd5eb19c1a6a032ae67fc46fd148965411
--- /dev/null
+++ b/records/sd1/sweep/660b81d3fbab286b7551/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27d34bb5931868e67047f32398e800fbe2de579bfdd2ac41f566c6753ba63489
+size 20648
diff --git a/records/sd1/sweep/66137f62607558e48e9e/output.webp b/records/sd1/sweep/66137f62607558e48e9e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..092bb6a574b037243f19470b6e1f8bd4c14d57e3
--- /dev/null
+++ b/records/sd1/sweep/66137f62607558e48e9e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55b6adcdf4824d382c95ec771a617fc67f9a71c93c80a7b469af7ba6f4a15add
+size 27874
diff --git a/records/sd1/sweep/661df6e5fe603c3091a1/output.webp b/records/sd1/sweep/661df6e5fe603c3091a1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3f4c8584511cf941015a3248f0ab12ea4c39ef6b
--- /dev/null
+++ b/records/sd1/sweep/661df6e5fe603c3091a1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a73787464e11611dd0320de00ca06fe2ace051102a54f07575079fb478595d6
+size 38886
diff --git a/records/sd1/sweep/6623e738541d02781635/output.webp b/records/sd1/sweep/6623e738541d02781635/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e25084512dddedf431fda86bef37f0b60dae03e8
--- /dev/null
+++ b/records/sd1/sweep/6623e738541d02781635/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5fb92a077233fee62f12dc1a21342d7b6a53ee3c35e9d90a047e9cb3cf5831d0
+size 27714
diff --git a/records/sd1/sweep/662801ff95572ceea25a/output.webp b/records/sd1/sweep/662801ff95572ceea25a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2125eb80cff2ee9deace551557419dc65157bd90
--- /dev/null
+++ b/records/sd1/sweep/662801ff95572ceea25a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a245d9c60766de17e1d018b7dc251c63ee6652dd8fa27b9d2067c4360959771
+size 51308
diff --git a/records/sd1/sweep/662e07a41a650a99bc26/output.webp b/records/sd1/sweep/662e07a41a650a99bc26/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2a749863ad32b2fd3da8d2d95a1f3a0ea9760f60
--- /dev/null
+++ b/records/sd1/sweep/662e07a41a650a99bc26/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee5242e0b01eba9567f3b2505ef1475e733e1dc604dd81e2c4e5c13d69e7ec29
+size 10564
diff --git a/records/sd1/sweep/662f96e09f0c1ac581ee/output.webp b/records/sd1/sweep/662f96e09f0c1ac581ee/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7dc837ee9b2474b8689187412a96623eadab2fe4
--- /dev/null
+++ b/records/sd1/sweep/662f96e09f0c1ac581ee/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c66b11c21f71afd74ba93074ecc53f733b4307255909c111a27718cc43f7376c
+size 19664
diff --git a/records/sd1/sweep/66406588af7ed19d94c2/output.webp b/records/sd1/sweep/66406588af7ed19d94c2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1ee85e5d1afa048d5a1e4b00d6cdc7ca4df171fc
--- /dev/null
+++ b/records/sd1/sweep/66406588af7ed19d94c2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:732863e94f41c7e928ddace850c973ffdae24a1ec51584f5fb138f8323c9b8ae
+size 54802
diff --git a/records/sd1/sweep/664486617da6e06ba85a/output.webp b/records/sd1/sweep/664486617da6e06ba85a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e43b4f84efaf7dfb1033d8f57fb64d7db3812a17
--- /dev/null
+++ b/records/sd1/sweep/664486617da6e06ba85a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04ca08c5e7ca4340a6b8459e1794fd494631130dec5f57e40b88f929e605f078
+size 22538
diff --git a/records/sd1/sweep/664f717ecaf492bc8ccb/output.webp b/records/sd1/sweep/664f717ecaf492bc8ccb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fc2f31c5540569f4f994f57238d895daa4dfaf90
--- /dev/null
+++ b/records/sd1/sweep/664f717ecaf492bc8ccb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9321561761c9dbd138c52d9fd7f94a96ad00369936665b935c428f0bb0d13256
+size 18138
diff --git a/records/sd1/sweep/665349077ab66c92c58b/output.webp b/records/sd1/sweep/665349077ab66c92c58b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e26ee606b0ccdc07d8377cbcec8b0b9e93ed67b6
--- /dev/null
+++ b/records/sd1/sweep/665349077ab66c92c58b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e136a2cf46afb390e5e3910957f6ed373c9cacd93f3df14beda5d4a9cc76e06d
+size 53920
diff --git a/records/sd1/sweep/665f0fc6042a3a6e8698/output.webp b/records/sd1/sweep/665f0fc6042a3a6e8698/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..43d91bc2b0cad99c3717e34e2d362717bd6125cc
--- /dev/null
+++ b/records/sd1/sweep/665f0fc6042a3a6e8698/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb95eccc5fc6cd7b55d8d2cd742ed66c2c976d67b49f360d3a3028dfefc9f048
+size 52204
diff --git a/records/sd1/sweep/66655058250035e341f8/output.webp b/records/sd1/sweep/66655058250035e341f8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..00ef98e90783f7d24fe7267d0283f3032a3046bd
--- /dev/null
+++ b/records/sd1/sweep/66655058250035e341f8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ace6ed0c9186a10f9bfbd961abafb513a7976c804eb960d60d8c3cd1e5a7257
+size 67272
diff --git a/records/sd1/sweep/666abc2f1f09c20bb085/output.webp b/records/sd1/sweep/666abc2f1f09c20bb085/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d4d8b647cfec42c1c7dbbc8ffcb28a721da7ee41
--- /dev/null
+++ b/records/sd1/sweep/666abc2f1f09c20bb085/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7fe0befa2a33563ca55a7fd2d939d888a73523c52289ad280c74c31fb019b6c
+size 40542
diff --git a/records/sd1/sweep/669a6a2f78a0f20662a7/output.webp b/records/sd1/sweep/669a6a2f78a0f20662a7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c36dd6809b7ec714f5a6162e9ac2f371f7f4bcf4
--- /dev/null
+++ b/records/sd1/sweep/669a6a2f78a0f20662a7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b164b369271d092171a2e9980dae4c00a68e8ba90ce10b4b7169f8b0ee5d2f6
+size 28238
diff --git a/records/sd1/sweep/66a18a29b49224c03fa0/output.webp b/records/sd1/sweep/66a18a29b49224c03fa0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..685a9c2a07a692d00729540d048697020fea1948
--- /dev/null
+++ b/records/sd1/sweep/66a18a29b49224c03fa0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7432e0f9526f2da38926c4aec113343098b7e3e0d39a6157c6873ab327e1ca53
+size 7108
diff --git a/records/sd1/sweep/66a79210a6cd70947f1c/output.webp b/records/sd1/sweep/66a79210a6cd70947f1c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1fa91480d973ef17121bd9f7726a11d01f821287
--- /dev/null
+++ b/records/sd1/sweep/66a79210a6cd70947f1c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4741be8c617825ef21c5d8a3b224ff0f3f3ad7e22f3b37a855074fc718314071
+size 19142
diff --git a/records/sd1/sweep/66a9422f8ccd9f7ae1c4/output.webp b/records/sd1/sweep/66a9422f8ccd9f7ae1c4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..925473077685890a97fab964e00a11c25004b87f
--- /dev/null
+++ b/records/sd1/sweep/66a9422f8ccd9f7ae1c4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b21ee3df7e02a342f9eb84bf99093de78dd5d0ead83b6e5366c56a391c310c6
+size 20012
diff --git a/records/sd1/sweep/66acb61110e5b3182657/output.webp b/records/sd1/sweep/66acb61110e5b3182657/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a8d585052acb9e12aab772398666af313305153b
--- /dev/null
+++ b/records/sd1/sweep/66acb61110e5b3182657/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a01d96b96e6c65fe5db06da5d71da727fcfd5117f2f9fb2aa8f871d85285b81
+size 18986
diff --git a/records/sd1/sweep/66c802ab8bb66d43118c/output.webp b/records/sd1/sweep/66c802ab8bb66d43118c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b23cf3b07451985ecf3551928b3478e7ef9c7a50
--- /dev/null
+++ b/records/sd1/sweep/66c802ab8bb66d43118c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3379433868de7b2c9345fe07c9e8c985bf486797b34a84baefa21c25a8cc3eb2
+size 19570
diff --git a/records/sd1/sweep/66ca170df4a028e3403c/output.webp b/records/sd1/sweep/66ca170df4a028e3403c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3e4499dd436a1224e3a2ff5d14af799306362d3b
--- /dev/null
+++ b/records/sd1/sweep/66ca170df4a028e3403c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f8cc37d42a3823b90f0b879987aad477834a20f2906b962e65f4f7bc9cd56e3
+size 17436
diff --git a/records/sd1/sweep/66ca36c7f4c4d17aa482/output.webp b/records/sd1/sweep/66ca36c7f4c4d17aa482/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5f0314e43acc35f49c2a1735cc14d91b6a49e6bd
--- /dev/null
+++ b/records/sd1/sweep/66ca36c7f4c4d17aa482/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3aafe5c882ecc8e57233079f9c6e3659b43ed0305e8e14ba007fb435555b8e3d
+size 33330
diff --git a/records/sd1/sweep/66cdef88d0c460a59467/output.webp b/records/sd1/sweep/66cdef88d0c460a59467/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b3958d0f12729dd64bdecc12e69e4bbf02bd7a39
--- /dev/null
+++ b/records/sd1/sweep/66cdef88d0c460a59467/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:179de119df0dafe7b54c48de9c1b467a5c4dc75a58a89cdb9962aaa5540534d5
+size 41590
diff --git a/records/sd1/sweep/66d56ac5be694c7f762d/output.webp b/records/sd1/sweep/66d56ac5be694c7f762d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f307f33d9941b24efe32e5313f4fb21b73a6849a
--- /dev/null
+++ b/records/sd1/sweep/66d56ac5be694c7f762d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:987e15a893a6db655c6e540811be58bf372e652efd5510e3bf99d7c70230bf89
+size 29244
diff --git a/records/sd1/sweep/66d764730f5ff3467e1e/output.webp b/records/sd1/sweep/66d764730f5ff3467e1e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..878705c4c86f5250a39673a22001a772eda6265c
--- /dev/null
+++ b/records/sd1/sweep/66d764730f5ff3467e1e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85acb28fef85457270c9ac71f97b3ea40c953cafbe368f3c42df12b2f815e275
+size 71624
diff --git a/records/sd1/sweep/66f85ce85871a2dfbdb8/output.webp b/records/sd1/sweep/66f85ce85871a2dfbdb8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cd481e41a15f86fba558819a2b3ca95b4fba0f04
--- /dev/null
+++ b/records/sd1/sweep/66f85ce85871a2dfbdb8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78684843d1b73e2724d6555fc214f987acdc072f229e23ee5535756b1f2349c8
+size 31456
diff --git a/records/sd1/sweep/6701b7e0114ba9b3fd66/output.webp b/records/sd1/sweep/6701b7e0114ba9b3fd66/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3462c7a334b1b319032dbe64a7c6924959df2b07
--- /dev/null
+++ b/records/sd1/sweep/6701b7e0114ba9b3fd66/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:faa08173eb2686696b7146059d9128f40d8af5610be594d07951e2e36520d239
+size 36498
diff --git a/records/sd1/sweep/67022ff1e6603d886834/output.webp b/records/sd1/sweep/67022ff1e6603d886834/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8401715211a942c04de895082bdbd10cc0a350e0
--- /dev/null
+++ b/records/sd1/sweep/67022ff1e6603d886834/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dea1622487e4759201c7094776eff84181f368b22a4dfce64dd36b1c278ca7e
+size 30838
diff --git a/records/sd1/sweep/67097ccbd59f124f8e15/output.webp b/records/sd1/sweep/67097ccbd59f124f8e15/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..637ff72946c548c0072c3099925c8fd6469cf21a
--- /dev/null
+++ b/records/sd1/sweep/67097ccbd59f124f8e15/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d605726dcd16fb5b3e05a6d5d7705ce1414bfcbdf165093b4b56d6495b1566b6
+size 35710
diff --git a/records/sd1/sweep/670a8b2c9d38d5964f7d/output.webp b/records/sd1/sweep/670a8b2c9d38d5964f7d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7a283f27f66a9d691595545d533568a26243b328
--- /dev/null
+++ b/records/sd1/sweep/670a8b2c9d38d5964f7d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b0316c1071ea538efd2f6cd612e188f1e4881332270188adf48b6f28b7147fd
+size 54184
diff --git a/records/sd1/sweep/6720ca1847a9a57e40e7/output.webp b/records/sd1/sweep/6720ca1847a9a57e40e7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9753a2b8963c52a81858aa483e38787c7735a9ab
--- /dev/null
+++ b/records/sd1/sweep/6720ca1847a9a57e40e7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b6aa8ae6fa95d0936a73ae5aafe903a4ce89e993bbbd34cfd0d462c8b26bcbc
+size 22080
diff --git a/records/sd1/sweep/67219e110207cfe10600/output.webp b/records/sd1/sweep/67219e110207cfe10600/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..06943c46097297eca26980654ef2b2e8be0c7b32
--- /dev/null
+++ b/records/sd1/sweep/67219e110207cfe10600/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcdc2db8c93c167f0735865aea314203118133804e09088b3cb55d791d482f7c
+size 35958
diff --git a/records/sd1/sweep/6725bb0d99a73c00a958/output.webp b/records/sd1/sweep/6725bb0d99a73c00a958/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8fb7103905d9e6b2afe69567b32e91b67ce15610
--- /dev/null
+++ b/records/sd1/sweep/6725bb0d99a73c00a958/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:594e881b4aff9060ca7b418d4b2e6fd2ec944542bc5835f20f79060bb5e32941
+size 32922
diff --git a/records/sd1/sweep/672bb4f03d082985a875/output.webp b/records/sd1/sweep/672bb4f03d082985a875/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..232ad38ea02b9c2a0ad7742cd0234b605f3d2a84
--- /dev/null
+++ b/records/sd1/sweep/672bb4f03d082985a875/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7d2c9e8bf3c624012268b900a08c5f4c7c68cdc2779afb8cfccbaf71ecbd235
+size 20648
diff --git a/records/sd1/sweep/673bef9d461af124e0ed/output.webp b/records/sd1/sweep/673bef9d461af124e0ed/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..59bdf1c71c501010496c566e7a48cd7c3bf0ff08
--- /dev/null
+++ b/records/sd1/sweep/673bef9d461af124e0ed/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49c45cf1488b3f55a9a9b385878b7604f1c720605123fc54c2db808b9106bbad
+size 19934
diff --git a/records/sd1/sweep/673eb9f9dfab8fa33028/output.webp b/records/sd1/sweep/673eb9f9dfab8fa33028/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d19fb0f46c9c376635bfb5f83a9ce37e35589ce3
--- /dev/null
+++ b/records/sd1/sweep/673eb9f9dfab8fa33028/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3156c73be8c240b1fb9245dc70434dd8bb08388e01448b1915de7290613587cf
+size 39612
diff --git a/records/sd1/sweep/6742c551e04cb5caa01e/output.webp b/records/sd1/sweep/6742c551e04cb5caa01e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d25e9ce3b24d75604c0399494f337cc4b88c2b33
--- /dev/null
+++ b/records/sd1/sweep/6742c551e04cb5caa01e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa02ac7f2ef9638de994590b8b7ae24f0d326ffbffe2404fc1218608817ed32e
+size 39646
diff --git a/records/sd1/sweep/6745f7f56129193d2559/output.webp b/records/sd1/sweep/6745f7f56129193d2559/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..99d582410d625dfb7f10607515f6fabac9fa1502
--- /dev/null
+++ b/records/sd1/sweep/6745f7f56129193d2559/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb209017b8caf267c8126acd8030842e7c131a7c6526dcb847743913328e83d0
+size 51482
diff --git a/records/sd1/sweep/674671862ae0396789ad/output.webp b/records/sd1/sweep/674671862ae0396789ad/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..37241787e2d4691de2959bd2c64cc9bfcf9f51ea
--- /dev/null
+++ b/records/sd1/sweep/674671862ae0396789ad/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1533c9263ad955eaaf6e8705278ecd06a3908c8eeee58bbe1e5d4ba44eea6da
+size 49854
diff --git a/records/sd1/sweep/676a8e4f546d66b54e46/output.webp b/records/sd1/sweep/676a8e4f546d66b54e46/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1278895cacc4b785ade20678738d617054017521
--- /dev/null
+++ b/records/sd1/sweep/676a8e4f546d66b54e46/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee89a3bf15909b19dc77dbc3f8f6c5146eb0cb484b885ca74c3a0e482863cf11
+size 39174
diff --git a/records/sd1/sweep/676d0cfca2d6165d5f7e/output.webp b/records/sd1/sweep/676d0cfca2d6165d5f7e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e2e1201fb07fefc05fe46fccb2a7b2112cdd9e09
--- /dev/null
+++ b/records/sd1/sweep/676d0cfca2d6165d5f7e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65872a34d404258263f92c1308c2ebf9d609b2f59034fa051a8d6ca8bf769660
+size 34348
diff --git a/records/sd1/sweep/67768f0ac0b71ae447ec/output.webp b/records/sd1/sweep/67768f0ac0b71ae447ec/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f3bd8345956d9c8b455940d4d895c146cf46d5c9
--- /dev/null
+++ b/records/sd1/sweep/67768f0ac0b71ae447ec/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0c9a9bda9ee55ea66802a472c43a0a03575e024402e7317480b562f70331818
+size 34110
diff --git a/records/sd1/sweep/6777133e07745a7dba8a/output.webp b/records/sd1/sweep/6777133e07745a7dba8a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..56cffff4a9f3563fd136896557cef2acbe540b9d
--- /dev/null
+++ b/records/sd1/sweep/6777133e07745a7dba8a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:028dea1e2d7b5ecee34d16ee0c9ab7624499d31c2de8fa3c755c2491cefe2993
+size 26696
diff --git a/records/sd1/sweep/677803d3015330e19cdc/output.webp b/records/sd1/sweep/677803d3015330e19cdc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6f40ef7168a0ac28771a3436763539a864d76a88
--- /dev/null
+++ b/records/sd1/sweep/677803d3015330e19cdc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47c8165720d0df7bf2c3153b3ccd90f521ab93d3b0011d029b957b9985c3e544
+size 35398
diff --git a/records/sd1/sweep/677c9d68f32554b81767/output.webp b/records/sd1/sweep/677c9d68f32554b81767/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cc99703cf1c5d1f8a4fc2729a92c61c5dbe5c7f9
--- /dev/null
+++ b/records/sd1/sweep/677c9d68f32554b81767/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f28254e1f146eba017020aaae672caf33cf65d00ca098492b6641db89143040
+size 18278
diff --git a/records/sd1/sweep/677d2fcfcfad1c59a06b/output.webp b/records/sd1/sweep/677d2fcfcfad1c59a06b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f6cfdcc4b7b2c189af80600622c3f428ec8211b2
--- /dev/null
+++ b/records/sd1/sweep/677d2fcfcfad1c59a06b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc2695124cf3012b27ed679c50c30ae8c12b673512328351fa73047a1f549642
+size 48332
diff --git a/records/sd1/sweep/6788b6a35e87476f4f85/output.webp b/records/sd1/sweep/6788b6a35e87476f4f85/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b7c1027510529caa29470806a8e169ada504d9fc
--- /dev/null
+++ b/records/sd1/sweep/6788b6a35e87476f4f85/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0f646dffc4367aa65a6d37ac378104e18fc5d7496b0fd41fc338541efa7b3a7
+size 26730
diff --git a/records/sd1/sweep/678feeaaf0983a057233/output.webp b/records/sd1/sweep/678feeaaf0983a057233/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2d22d79a1d3c40841721b775ece9c1bc489eb424
--- /dev/null
+++ b/records/sd1/sweep/678feeaaf0983a057233/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1c40dba10cab111f5892fc9bf86f1a57a5a678421f95764b4a8f658a7e31155
+size 51398
diff --git a/records/sd1/sweep/679e6be5a54cf0329430/output.webp b/records/sd1/sweep/679e6be5a54cf0329430/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e217c31d183afe2457f9d79a59be44129287122c
--- /dev/null
+++ b/records/sd1/sweep/679e6be5a54cf0329430/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6af42397c02aecc16ebe7c28790f6a873a9d0fa21f21ec843a3200ed1dd63ef6
+size 24144
diff --git a/records/sd1/sweep/67b48f3e2aa8bcafaa2b/output.webp b/records/sd1/sweep/67b48f3e2aa8bcafaa2b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..73a1268c5a9e322e36182ca57af3ff64aca95263
--- /dev/null
+++ b/records/sd1/sweep/67b48f3e2aa8bcafaa2b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc67be25e6232fe91f94694396cd7b659558317432fc8c6056a95d3d3bd73d5b
+size 58296
diff --git a/records/sd1/sweep/67b9cac9f45ff0d1fda7/output.webp b/records/sd1/sweep/67b9cac9f45ff0d1fda7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a6669883e3c2ac86e53f33f4643ab97083378c5b
--- /dev/null
+++ b/records/sd1/sweep/67b9cac9f45ff0d1fda7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7a7a88ab445c4d3f5be636d073f15e49199a869d8de4c3e1835df9bd363f452
+size 32630
diff --git a/records/sd1/sweep/67ba85cf8f9639c87b85/output.webp b/records/sd1/sweep/67ba85cf8f9639c87b85/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7e590c6949e76575d8076eba191162588d9376f1
--- /dev/null
+++ b/records/sd1/sweep/67ba85cf8f9639c87b85/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b33646168bed05b637c015c1b69f3fbe9df7cde31d871a061d30b647a74a0a21
+size 30156
diff --git a/records/sd1/sweep/67be89d81315a7c2777d/output.webp b/records/sd1/sweep/67be89d81315a7c2777d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..461df989ba2060bb4053ea794f2600a75b9d55a1
--- /dev/null
+++ b/records/sd1/sweep/67be89d81315a7c2777d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d34e39a892d243fb7c5057a70e3c97d5de9b3c308d39d2d0eccdc78ed4c5a42
+size 35874
diff --git a/records/sd1/sweep/67bed7844aae73e602c8/output.webp b/records/sd1/sweep/67bed7844aae73e602c8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9e00c2341c5347877840c92faf9f72523374b068
--- /dev/null
+++ b/records/sd1/sweep/67bed7844aae73e602c8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1bad362e062a1203a8a06242ab4a6e97fd93fd2e8ff5b3adbc728530b50c9cc
+size 19746
diff --git a/records/sd1/sweep/67c2c3d316b11b1f5509/output.webp b/records/sd1/sweep/67c2c3d316b11b1f5509/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0b875f90559cd1387accbdf717365dc99d1c6ee5
--- /dev/null
+++ b/records/sd1/sweep/67c2c3d316b11b1f5509/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:511b593ffa33233fb9d2497a67be4b82081d5aa2644edbf7dd9436e40fb444aa
+size 26880
diff --git a/records/sd1/sweep/67d517c8abbdfff2a38c/output.webp b/records/sd1/sweep/67d517c8abbdfff2a38c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..431da05abf7a5017d84326e8575d359118525eaa
--- /dev/null
+++ b/records/sd1/sweep/67d517c8abbdfff2a38c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f99a8a33e87da73bcd0207da653f2c5496f1ec96b8823f0169249c738d266723
+size 39112
diff --git a/records/sd1/sweep/67d8c068bc137fe903f6/output.webp b/records/sd1/sweep/67d8c068bc137fe903f6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..907f498742e02cab4a5b89a6a7c62b899d5d7f29
--- /dev/null
+++ b/records/sd1/sweep/67d8c068bc137fe903f6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca268f2638d51af31dc705d00ac8714f7676fbc9152313f036ac964c978a82ff
+size 43022
diff --git a/records/sd1/sweep/67eb6642ca5b15d9406d/output.webp b/records/sd1/sweep/67eb6642ca5b15d9406d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c612023acee359120983c6a0a759d679944b6e6a
--- /dev/null
+++ b/records/sd1/sweep/67eb6642ca5b15d9406d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:406387fc43dccd1d83d4b561a33d99f8059a0c8e0d514b7290392dce857f3bb7
+size 58046
diff --git a/records/sd1/sweep/67eb7a12c044dcdc2ed4/output.webp b/records/sd1/sweep/67eb7a12c044dcdc2ed4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e8e4bfe7445f80ebdc6e750926a009a325782e93
--- /dev/null
+++ b/records/sd1/sweep/67eb7a12c044dcdc2ed4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7f9840f7c93f11a8e7c83d0142a7e11f0baf44650588a18a274c63f327c4578
+size 18102
diff --git a/records/sd1/sweep/67ec8453e926df88b441/output.webp b/records/sd1/sweep/67ec8453e926df88b441/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ee159e95037ec80b285c93cdd45564493e82bf4e
--- /dev/null
+++ b/records/sd1/sweep/67ec8453e926df88b441/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1aecad254024c45de3af7ccdbc82507f818bba0c375b8bd76eaa6ec218c4084a
+size 90878
diff --git a/records/sd1/sweep/67f43ea274aaf8dbf6d0/output.webp b/records/sd1/sweep/67f43ea274aaf8dbf6d0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..83dd74ea61d727f847dd6ea568d612488de63b5b
--- /dev/null
+++ b/records/sd1/sweep/67f43ea274aaf8dbf6d0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ad98c48589f3ae0c804aee13ff43743ab28db6cb6807bf34bef0eeb1f4f4606
+size 51580
diff --git a/records/sd1/sweep/67f662e3457b8f017d20/output.webp b/records/sd1/sweep/67f662e3457b8f017d20/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d0fdd12fb75905ea1eea138865ace3c75860bbef
--- /dev/null
+++ b/records/sd1/sweep/67f662e3457b8f017d20/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b4761a6fd0d65a934950266a1a496228a25ce8b40fd013b6dc8d79413923e0c
+size 28870
diff --git a/records/sd1/sweep/67f8db0b7e75789fc578/output.webp b/records/sd1/sweep/67f8db0b7e75789fc578/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..76dad6ef2552530e62b90b33e494bd88fd095006
--- /dev/null
+++ b/records/sd1/sweep/67f8db0b7e75789fc578/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e515b1f2844664ac8f83f7d820833931b5955fd8876fd7417aaaa501e785b9d
+size 48758
diff --git a/records/sd1/sweep/67f91d725aa6b0c84997/output.webp b/records/sd1/sweep/67f91d725aa6b0c84997/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..57136f5b8272917f03a665a7a68dab3a22542cdf
--- /dev/null
+++ b/records/sd1/sweep/67f91d725aa6b0c84997/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e043c91d06bb883dc33efe03329f8c238413548fa82ac056880517955edff4e
+size 45210
diff --git a/records/sd1/sweep/67fa022e4df80328f725/output.webp b/records/sd1/sweep/67fa022e4df80328f725/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ac1e24a816a5ff8b406278228292a850c289496a
--- /dev/null
+++ b/records/sd1/sweep/67fa022e4df80328f725/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1994eaf5e849f99f2d37cd51d85dbf4fbcdd583d65828be45d1b68782be3821d
+size 39396
diff --git a/records/sd1/sweep/680342e87fd258b37cad/output.webp b/records/sd1/sweep/680342e87fd258b37cad/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1967227e3e3d9a01c6d6034579717c2dced55120
--- /dev/null
+++ b/records/sd1/sweep/680342e87fd258b37cad/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95ffa341ce22e4743c142d082592e7a04a86f358ddb3cd4dc3ba3363f1305c2f
+size 27600
diff --git a/records/sd1/sweep/68054b04fe72ba8a461e/output.webp b/records/sd1/sweep/68054b04fe72ba8a461e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7992c8e04d1da9c5452354081f2a3dc5f7af3e0e
--- /dev/null
+++ b/records/sd1/sweep/68054b04fe72ba8a461e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9ef2a43c6220922a938230c2d1d6bdd480a4fbcd3c6409948041b337af1eb653
+size 47034
diff --git a/records/sd1/sweep/680788b0043335fc2977/output.webp b/records/sd1/sweep/680788b0043335fc2977/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..56b34843d2a7fbb03a3ff11d3defb0690b6cf724
--- /dev/null
+++ b/records/sd1/sweep/680788b0043335fc2977/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c2e7ae0e0a441b5fbe37d3bb818db81ec223f11679190d0aa56668b457e4459
+size 26678
diff --git a/records/sd1/sweep/68079e8aa2f0cce5d97c/output.webp b/records/sd1/sweep/68079e8aa2f0cce5d97c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3a49421f5504486e478fb4820fcbd5f6237c74a8
--- /dev/null
+++ b/records/sd1/sweep/68079e8aa2f0cce5d97c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ea06ca8641758ea973e35623b6b1a21e98752711b19f74b983ccf7cf46ce474
+size 19934
diff --git a/records/sd1/sweep/681028ce2540aa56e237/output.webp b/records/sd1/sweep/681028ce2540aa56e237/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a067f10f3d092e623dec6237fcd70bce7bccddf8
--- /dev/null
+++ b/records/sd1/sweep/681028ce2540aa56e237/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00d5ffcd82a4b3c2d22f8abea3231fb346863f26522ff4ad00ecad3a5c898344
+size 37218
diff --git a/records/sd1/sweep/6814225cba6f56ea84c9/output.webp b/records/sd1/sweep/6814225cba6f56ea84c9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f0c8c7c4dc790ee44696a56db1b4d3dc2066f217
--- /dev/null
+++ b/records/sd1/sweep/6814225cba6f56ea84c9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc6591570e795bdbb587fc73944ccc61929ae7a3d1f51cdbe6921393bfe1a90c
+size 34446
diff --git a/records/sd1/sweep/681b21ff97daaa1d9115/output.webp b/records/sd1/sweep/681b21ff97daaa1d9115/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9b9379be85c72e5686cff220102c1e40848e9739
--- /dev/null
+++ b/records/sd1/sweep/681b21ff97daaa1d9115/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc070cfb13eb81fae6aee1b4ede16455873488425252eb9e732053dc6bc38bda
+size 50830
diff --git a/records/sd1/sweep/681f5db08c2f7f00df05/output.webp b/records/sd1/sweep/681f5db08c2f7f00df05/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c791e472dfaef9b550d596b471f16d033827b170
--- /dev/null
+++ b/records/sd1/sweep/681f5db08c2f7f00df05/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0cb0d3ae1846595a6af85b226bddca468ae02e5b24e68672b1618d52a878b74
+size 29000
diff --git a/records/sd1/sweep/6821dbd74238d45351db/output.webp b/records/sd1/sweep/6821dbd74238d45351db/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cbc8ad6c1c8038f0ac40c199095e30f46c65c2d2
--- /dev/null
+++ b/records/sd1/sweep/6821dbd74238d45351db/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:395e5436a95cbf65b8385b9abf7512bfe90a0bcdf1496e185cf125457c22858c
+size 14476
diff --git a/records/sd1/sweep/682ab64609837d2eebb6/output.webp b/records/sd1/sweep/682ab64609837d2eebb6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..877f5a54edf2f19f76814f72c8b6df03ad9fb84f
--- /dev/null
+++ b/records/sd1/sweep/682ab64609837d2eebb6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f3ddd091ca13e91f57f514140282d4e2a93741198fdfb6c6d94083a13ade86b
+size 31542
diff --git a/records/sd1/sweep/682ea1965f0f34367d28/output.webp b/records/sd1/sweep/682ea1965f0f34367d28/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6acd5148b5e2af9362b3eb8dc30d887bd64d5979
--- /dev/null
+++ b/records/sd1/sweep/682ea1965f0f34367d28/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff59a27136973cce8eebc861e7cc1ff8a65e97296c1bbe54e3cfe8b2ebd8d515
+size 36126
diff --git a/records/sd1/sweep/68303eacd080b92a6a74/output.webp b/records/sd1/sweep/68303eacd080b92a6a74/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9ad1da6e07bcd16a0d1c5709cbcdb70c2f99bb7d
--- /dev/null
+++ b/records/sd1/sweep/68303eacd080b92a6a74/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d760b38b0657bbd1619fb37990523f60953ff36ae6d90a010a1b299e1ba52b7
+size 25404
diff --git a/records/sd1/sweep/68317133fde66d476370/output.webp b/records/sd1/sweep/68317133fde66d476370/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c1b3d435df0b11f6d34a22f56bdd929ec9138082
--- /dev/null
+++ b/records/sd1/sweep/68317133fde66d476370/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6389a9f944bc63a60e7c3326ab629abe6e8a329da543b2c8f79f7f9c3230e2b9
+size 16830
diff --git a/records/sd1/sweep/683acd4b7b4ca32e3670/output.webp b/records/sd1/sweep/683acd4b7b4ca32e3670/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..29478639518d91bc36161f2ed31f815b6d1c0ca9
--- /dev/null
+++ b/records/sd1/sweep/683acd4b7b4ca32e3670/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:186ad4f1ea34b4e3eca2f60c80fc41fd4de7ec5979eb682c02e336c080868c01
+size 43676
diff --git a/records/sd1/sweep/683bfbc0f6ed37df238c/output.webp b/records/sd1/sweep/683bfbc0f6ed37df238c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3778da6312d15f285747e71ca0fe794b3bd7225c
--- /dev/null
+++ b/records/sd1/sweep/683bfbc0f6ed37df238c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1056138217b7ce18a262bfec4d7e7af5a5c68c064a2c5489971f61436e50a910
+size 26366
diff --git a/records/sd1/sweep/683c444b631e1a5fd583/output.webp b/records/sd1/sweep/683c444b631e1a5fd583/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6936255e8ad784857f9fffbeba8c5c890821df87
--- /dev/null
+++ b/records/sd1/sweep/683c444b631e1a5fd583/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4cdcab5ad38a0740ef8ea428c121c5d744eb8256e524d7f1cdef789c53982123
+size 37494
diff --git a/records/sd1/sweep/684463945576e47759ed/output.webp b/records/sd1/sweep/684463945576e47759ed/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b7c6a7834be105ad3e54109afa667a5c719f2cb3
--- /dev/null
+++ b/records/sd1/sweep/684463945576e47759ed/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbe5d5d6bf94833691804899e600cce851f54f8b0e9e1e7bd8eb68f72757fe93
+size 18808
diff --git a/records/sd1/sweep/685b0fbc1468ab0de892/output.webp b/records/sd1/sweep/685b0fbc1468ab0de892/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..92ed69457127201163061a7ab5874fa53b94ec53
--- /dev/null
+++ b/records/sd1/sweep/685b0fbc1468ab0de892/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73373e8c21f58376798baa5883ad6f81c97119f51beb63ea97bb7f7cea03ec53
+size 33644
diff --git a/records/sd1/sweep/686c922f85c3ae72a868/output.webp b/records/sd1/sweep/686c922f85c3ae72a868/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9ddff574955496d0837e99e9ce7fc626be28b151
--- /dev/null
+++ b/records/sd1/sweep/686c922f85c3ae72a868/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1e00ec4b492a9757d47f03028d609531d5e078874632f102802115509c532457
+size 60014
diff --git a/records/sd1/sweep/68776d8fe62b21b9b4e4/output.webp b/records/sd1/sweep/68776d8fe62b21b9b4e4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bfc816816b88f8d52fe51f4830dbdcd574f9a4cc
--- /dev/null
+++ b/records/sd1/sweep/68776d8fe62b21b9b4e4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d138730ae41c56118615ffccc28876c3f254b14daf8dab3a2c45c9ecc9fd43a7
+size 56426
diff --git a/records/sd1/sweep/687ce307297105a16c4d/output.webp b/records/sd1/sweep/687ce307297105a16c4d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5968b6b8b2b16fcb0885d5fc7b958309fca80fd4
--- /dev/null
+++ b/records/sd1/sweep/687ce307297105a16c4d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69b0e15e70a26bb339204c5151286bfcf0d138e2366866a490e138f0fe8f4f13
+size 86356
diff --git a/records/sd1/sweep/6881b7ba3f8bd6757672/output.webp b/records/sd1/sweep/6881b7ba3f8bd6757672/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0348f36c77391df3c53aeb6c721d741bc630de9c
--- /dev/null
+++ b/records/sd1/sweep/6881b7ba3f8bd6757672/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cdc66d69ab265666d1809d3d42481424f69a7b8b0114939893babcac6219a0b
+size 37848
diff --git a/records/sd1/sweep/689a281a3ffda9235255/output.webp b/records/sd1/sweep/689a281a3ffda9235255/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c102480b517f34cc701900855739a8b47510dd48
--- /dev/null
+++ b/records/sd1/sweep/689a281a3ffda9235255/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f4dbb0b610e3c09913683a1cb78e09fb3eef5ce27b906755adb05156437bf94
+size 29266
diff --git a/records/sd1/sweep/689a97ff13b1516c806a/output.webp b/records/sd1/sweep/689a97ff13b1516c806a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e3ee59ab821e66dcb4d5c32755fc228937dcb118
--- /dev/null
+++ b/records/sd1/sweep/689a97ff13b1516c806a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b8a6eaad7dc7b0b36dda3bd0c9da9b2b195e124fb6ec6c946e3b061a0f7cbf89
+size 29164
diff --git a/records/sd1/sweep/689fd44c7ae64167aa25/output.webp b/records/sd1/sweep/689fd44c7ae64167aa25/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ff17c6920336325e451d1ebffe18c2a8319c78fa
--- /dev/null
+++ b/records/sd1/sweep/689fd44c7ae64167aa25/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee8e5704d2714d7210890051eafe11b8fe5c90f270764ad275c6cd865c6ee8ca
+size 21120
diff --git a/records/sd1/sweep/68bae18b78c3fd43d698/output.webp b/records/sd1/sweep/68bae18b78c3fd43d698/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3843c68d3da5f932f195410fd258e1a4ea932194
--- /dev/null
+++ b/records/sd1/sweep/68bae18b78c3fd43d698/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b83a7114fb0a54da9b19006970b837c41e52715475d365edb3ea0bdcb0e8838
+size 30272
diff --git a/records/sd1/sweep/68c1049e47bde241ee1b/output.webp b/records/sd1/sweep/68c1049e47bde241ee1b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3a3d2135026955fad0d80033464fe6d703f06fad
--- /dev/null
+++ b/records/sd1/sweep/68c1049e47bde241ee1b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c01c93de72f2c5f95e5437840f99c0037ea1e826860acacb7504816e9d6f0da4
+size 42974
diff --git a/records/sd1/sweep/68c1f077b69c996a2471/output.webp b/records/sd1/sweep/68c1f077b69c996a2471/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cf3990e142ab7d08c7a482cc89b83a71a52cbba3
--- /dev/null
+++ b/records/sd1/sweep/68c1f077b69c996a2471/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6d379bf6ea2edce32acf3973121d9171878a8c70bd42c15985f2d4af445092b
+size 7346
diff --git a/records/sd1/sweep/68ccabcf8c56ae7c0741/output.webp b/records/sd1/sweep/68ccabcf8c56ae7c0741/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fcabfc24c047956da2e61ac591b8cdf76281765c
--- /dev/null
+++ b/records/sd1/sweep/68ccabcf8c56ae7c0741/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfbe94d2b9ec870151d815391120db77b3e663ba9e6c61c097ecc1ca97096816
+size 49916
diff --git a/records/sd1/sweep/68ccf816419fe554ff65/output.webp b/records/sd1/sweep/68ccf816419fe554ff65/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f3f6611fbb16d7ce4caa8ade0186a36030f7b961
--- /dev/null
+++ b/records/sd1/sweep/68ccf816419fe554ff65/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbf061e888e176a1427d3a20956e9c9b8ee3c7adc07673e60030e5d4b5fe1297
+size 27464
diff --git a/records/sd1/sweep/68d51f52ecc75b80a386/output.webp b/records/sd1/sweep/68d51f52ecc75b80a386/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a11d1ba3f9faa6f52f8ae202af85d30d273a37d2
--- /dev/null
+++ b/records/sd1/sweep/68d51f52ecc75b80a386/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65ca248ec190ba4dfbaa6a8084bdd7be251ac4482456386689d0a0963e602d31
+size 27704
diff --git a/records/sd1/sweep/68d98eab39e6d400aa52/output.webp b/records/sd1/sweep/68d98eab39e6d400aa52/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..06b0980136c7ecf245fd83d490ba8102fca9be8c
--- /dev/null
+++ b/records/sd1/sweep/68d98eab39e6d400aa52/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66251280fdbb1bfdca13fbe8f75dc9562c7b92edade5842d1d693170dd6eafe2
+size 31402
diff --git a/records/sd1/sweep/68e6a02e22b3da834ca8/output.webp b/records/sd1/sweep/68e6a02e22b3da834ca8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7eb7f656a91f654a6679f49dd3bb92f30ef23422
--- /dev/null
+++ b/records/sd1/sweep/68e6a02e22b3da834ca8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04e917582d17025178febb6e64b20431073fb740bf0d7badb66a84cbf8b88689
+size 18426
diff --git a/records/sd1/sweep/68ec578f155461b0e15f/output.webp b/records/sd1/sweep/68ec578f155461b0e15f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..99159907cbaf98b4302a5c42028083b8c261b6d8
--- /dev/null
+++ b/records/sd1/sweep/68ec578f155461b0e15f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5627bdbaa0c4a473e3ef91edb8da525386b4cb1bd835f4cf52412505596f4ac
+size 18996
diff --git a/records/sd1/sweep/68ecec5ad961cdf7222e/output.webp b/records/sd1/sweep/68ecec5ad961cdf7222e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cd3dc497b6582f8bc61b6ccb2d8b350a24fbb359
--- /dev/null
+++ b/records/sd1/sweep/68ecec5ad961cdf7222e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef8ae06dee905741fe93f3b2cd5e0bbdca6cd59a039463ddc4cbf0aad174265c
+size 36496
diff --git a/records/sd1/sweep/68ee167801ebec2c4424/output.webp b/records/sd1/sweep/68ee167801ebec2c4424/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..acc4643fe57a8e17c9c7956a1dd4e6f02b0c8ccd
--- /dev/null
+++ b/records/sd1/sweep/68ee167801ebec2c4424/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f0ed64d6f2d8a4816b4d25c0202fe0d8ecd3bb43480ffccd4240b957251b4ba
+size 21022
diff --git a/records/sd1/sweep/68f63c8a48a1213ee538/output.webp b/records/sd1/sweep/68f63c8a48a1213ee538/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..18f881f3ed4f4b4027cb33251ca39dfe2c4d9494
--- /dev/null
+++ b/records/sd1/sweep/68f63c8a48a1213ee538/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:931338214c7f73bf87fa06d0596b8a8fe492b0e21f14624b65fc3f4f7ecc715b
+size 42984
diff --git a/records/sd1/sweep/690c88a83ff3ca14253e/output.webp b/records/sd1/sweep/690c88a83ff3ca14253e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..aa4e0c0867d7dc8761a5a1b100bb7faa9c4a4228
--- /dev/null
+++ b/records/sd1/sweep/690c88a83ff3ca14253e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66cab9688f9da3207abb19429d084b881eac0e157db0630bfbb3437b5deddda7
+size 17488
diff --git a/records/sd1/sweep/690e71eed8013523bc47/output.webp b/records/sd1/sweep/690e71eed8013523bc47/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..96b5001b86594e1a46d1d797790b3b7e624f695d
--- /dev/null
+++ b/records/sd1/sweep/690e71eed8013523bc47/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:065bca471227efbef35ef6027858e66cbd33fd2378e12ddc4d1450bac7ad1467
+size 20398
diff --git a/records/sd1/sweep/6924769dbb892066786b/output.webp b/records/sd1/sweep/6924769dbb892066786b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b831016209405d2ddc3c19e734b40136dc0cf3da
--- /dev/null
+++ b/records/sd1/sweep/6924769dbb892066786b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02e1730d18edaa77319af2862641fe59c160587beed8061ba636d56868179c18
+size 27808
diff --git a/records/sd1/sweep/6929dc703e290a2aafdf/output.webp b/records/sd1/sweep/6929dc703e290a2aafdf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3e05bd54acf32aa17402d20bd8bbefff81c44ddb
--- /dev/null
+++ b/records/sd1/sweep/6929dc703e290a2aafdf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8b8d7e65c5772af25db040393bec0f76dc043068a6affe28e3597ead3a02045b
+size 18152
diff --git a/records/sd1/sweep/693258172c7f753311a7/output.webp b/records/sd1/sweep/693258172c7f753311a7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6ebaf2454d14a689a50ca96d1c6ffb2643dcfca0
--- /dev/null
+++ b/records/sd1/sweep/693258172c7f753311a7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa1ea54a8c0eb20aa95b746a2b5e915fb904f45c62ca8be1ec0ba3e936884617
+size 17624
diff --git a/records/sd1/sweep/6942b5a39032989ef569/output.webp b/records/sd1/sweep/6942b5a39032989ef569/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ef95ee402ec40c4f8bd1b8804a46efeac9be4b2c
--- /dev/null
+++ b/records/sd1/sweep/6942b5a39032989ef569/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d73b6363011c64c0cb37257d182a24aec54276f6c7a00cd2f946aad9c16df69d
+size 21820
diff --git a/records/sd1/sweep/6943609a32d1eced8b6d/output.webp b/records/sd1/sweep/6943609a32d1eced8b6d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e958244a7fbfc23c32376a5a50bd671e6de13e28
--- /dev/null
+++ b/records/sd1/sweep/6943609a32d1eced8b6d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc4118449f8a236c343c8c6fdea3236db3c80e6a88c5891f3668f15f85503ffe
+size 47614
diff --git a/records/sd1/sweep/694765ccc080f2256d3b/output.webp b/records/sd1/sweep/694765ccc080f2256d3b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b86a452394c0009ab683117c26410fa1ae342fc8
--- /dev/null
+++ b/records/sd1/sweep/694765ccc080f2256d3b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbc28bfb6cc1b0032e3327111b992e44003c548f0f77e68f44b5510f6808b617
+size 38836
diff --git a/records/sd1/sweep/6948ef9edfe2df80c0be/output.webp b/records/sd1/sweep/6948ef9edfe2df80c0be/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..924cb6cecf2cd2cb09f7452735d2948c1dec936d
--- /dev/null
+++ b/records/sd1/sweep/6948ef9edfe2df80c0be/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d76a343bf6f137a67f92ebd6c81a8e119cfd581204c3de47f39814ba67b46584
+size 27112
diff --git a/records/sd1/sweep/695432c27f23e39c4e48/output.webp b/records/sd1/sweep/695432c27f23e39c4e48/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60bd52794aea1b40c18db727f2c418287ae4f8e1
--- /dev/null
+++ b/records/sd1/sweep/695432c27f23e39c4e48/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5142c7f518adef4f99f16cdfaf18a90c8abbddc6ae2d7043b99d9b4ca9fbfce5
+size 51246
diff --git a/records/sd1/sweep/695ca85769982411b0ec/output.webp b/records/sd1/sweep/695ca85769982411b0ec/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5a1ab05bbc7bbf6d7eebe7d3796c2ae702029272
--- /dev/null
+++ b/records/sd1/sweep/695ca85769982411b0ec/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:780c4a65ed16abd12aa3e901497e51c810ab90911fa0bce2504288437f95c205
+size 47066
diff --git a/records/sd1/sweep/69790bf16048d5278be0/output.webp b/records/sd1/sweep/69790bf16048d5278be0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..346fd7c3f770f606df19480c00f7fa4f67abdf5e
--- /dev/null
+++ b/records/sd1/sweep/69790bf16048d5278be0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1a9b204dd4d880f560f7e65a28c7c78e9f1e76971dc2dcc118305dc87adc093
+size 36504
diff --git a/records/sd1/sweep/697d0808de422ac0786d/output.webp b/records/sd1/sweep/697d0808de422ac0786d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a14c505f32c5fe9334c35a944079211baf5177a6
--- /dev/null
+++ b/records/sd1/sweep/697d0808de422ac0786d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81e6b0c2300da0abedcde75ba4fd90064d06606a1e6451f30ceca1ff97a6be8f
+size 33022
diff --git a/records/sd1/sweep/6980efe9ebce0e6b605e/output.webp b/records/sd1/sweep/6980efe9ebce0e6b605e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6320c8610c3f40e0a37f7f71f704f4df62f4b3e6
--- /dev/null
+++ b/records/sd1/sweep/6980efe9ebce0e6b605e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:580d0acdb301b1c75b935650bdccafce5ef1137f4a7333fb7124f7d52dc6407a
+size 36504
diff --git a/records/sd1/sweep/6987dd2c9c2e51d32c5c/output.webp b/records/sd1/sweep/6987dd2c9c2e51d32c5c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6f39988284143bf7c17536a82840bdf39d88f9a2
--- /dev/null
+++ b/records/sd1/sweep/6987dd2c9c2e51d32c5c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25cd7b4760027d23fe7096a91b6621cf3fb835cd84974286cf9b457811c36846
+size 34106
diff --git a/records/sd1/sweep/6989c84bded4e990364c/output.webp b/records/sd1/sweep/6989c84bded4e990364c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c508bf8e7487b325cfdf807414618e0a00e75553
--- /dev/null
+++ b/records/sd1/sweep/6989c84bded4e990364c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:179a89f923b49202c14fa252b4ceff40afe7db92d8bb8cb0589d35b58d0c8594
+size 36506
diff --git a/records/sd1/sweep/698f62d7b89ae3202444/output.webp b/records/sd1/sweep/698f62d7b89ae3202444/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fd979058f9dcd9d2c11610d0a215eea226597e9d
--- /dev/null
+++ b/records/sd1/sweep/698f62d7b89ae3202444/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccefe2af1b13583f705a38f1a8e9b3a64d9fdd49f90425409d84c6314d6c96e6
+size 50558
diff --git a/records/sd1/sweep/699dbe1945ddd9277be8/output.webp b/records/sd1/sweep/699dbe1945ddd9277be8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..82a5ccc521647edcedc0b2c0abc982497884b1a9
--- /dev/null
+++ b/records/sd1/sweep/699dbe1945ddd9277be8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:299d37970a17766a94d4cb9fda28b0c58dec818143e4341ec00fb505bab74257
+size 60356
diff --git a/records/sd1/sweep/69a6ae38f14a8da53fcc/output.webp b/records/sd1/sweep/69a6ae38f14a8da53fcc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..480bb2ae96d1407a2149520a80bd44e00c5ed6ae
--- /dev/null
+++ b/records/sd1/sweep/69a6ae38f14a8da53fcc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d17a6a7df322b02e29873400a9fc971a21c0676b79e04cab985875a84e6abf8
+size 55418
diff --git a/records/sd1/sweep/69a9b5ab1f494b273299/output.webp b/records/sd1/sweep/69a9b5ab1f494b273299/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d7556790d380e0810ec13233c79909d753ece30c
--- /dev/null
+++ b/records/sd1/sweep/69a9b5ab1f494b273299/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a86354e0ec0ed311bdcd4b50b5217a6012e6ee910be31ae69824aff5b2cecc27
+size 16160
diff --git a/records/sd1/sweep/69ae6eb39eba634fbb3f/output.webp b/records/sd1/sweep/69ae6eb39eba634fbb3f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7d41ce5d62ce535d2f860dc7c70e60fe398ee913
--- /dev/null
+++ b/records/sd1/sweep/69ae6eb39eba634fbb3f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:383ca379d7ccfcb15e1e712cb61f762262fc4bf72f4b00f40bd394d462f63e13
+size 19940
diff --git a/records/sd1/sweep/69b46353d9e5c6b26e0a/output.webp b/records/sd1/sweep/69b46353d9e5c6b26e0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3b0e5507f16bdcc3a91ab72142edac2357dfdd9a
--- /dev/null
+++ b/records/sd1/sweep/69b46353d9e5c6b26e0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd57d1fefd3acc01f122b40b40bf9173a9a841dc962088e8ed1490b5b7e74c80
+size 17506
diff --git a/records/sd1/sweep/69b95ed588f89af282fd/output.webp b/records/sd1/sweep/69b95ed588f89af282fd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..80f67d4a27bf393d9d2bb3b49295eb521ecb6446
--- /dev/null
+++ b/records/sd1/sweep/69b95ed588f89af282fd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52720b375307ac3c2293ccec13901032764ab68d9054da1d35285d44a57d38bb
+size 50964
diff --git a/records/sd1/sweep/69bed267a747ad864877/output.webp b/records/sd1/sweep/69bed267a747ad864877/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fe990a0aef6261ce23f8241c4b0dbf326cae0eb7
--- /dev/null
+++ b/records/sd1/sweep/69bed267a747ad864877/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55fc7b736db88a9284e1ead50273866ef4b6dbda881006608cdabc8a8aee9ccd
+size 32542
diff --git a/records/sd1/sweep/69c166c44f0894337ecb/output.webp b/records/sd1/sweep/69c166c44f0894337ecb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4271798b90ae579040584914f993173b65da91a1
--- /dev/null
+++ b/records/sd1/sweep/69c166c44f0894337ecb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0c2851cebf911d2eda99331b1316489bcf6317a78cdb12b59b8e238bf275a80
+size 55250
diff --git a/records/sd1/sweep/69c541601f5d650c97c1/output.webp b/records/sd1/sweep/69c541601f5d650c97c1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..abef6b69f068960409321b2136a474922d83fcdb
--- /dev/null
+++ b/records/sd1/sweep/69c541601f5d650c97c1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23b1e335b6e105519d829eae39f4e48e4fecbf21d1328ebaa3ef459ce309df5a
+size 27468
diff --git a/records/sd1/sweep/69c595edfe4aae11e518/output.webp b/records/sd1/sweep/69c595edfe4aae11e518/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e27e4c924c4bc1ad6c752fb4f7011d04118e78de
--- /dev/null
+++ b/records/sd1/sweep/69c595edfe4aae11e518/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:78831be80960d05de5d44c8afb529d00744cb2204be0e6ce607ba0ec1197ddb0
+size 49460
diff --git a/records/sd1/sweep/69c9c9f96fcea9105cf2/output.webp b/records/sd1/sweep/69c9c9f96fcea9105cf2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d0549b9932ca6d6c7f1a11299dae0f30a9ecd26a
--- /dev/null
+++ b/records/sd1/sweep/69c9c9f96fcea9105cf2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:106722185e39484c299995693145508987795ece51146ae8aae984d0479d581f
+size 36494
diff --git a/records/sd1/sweep/69d637049641cc3f3449/output.webp b/records/sd1/sweep/69d637049641cc3f3449/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2bb8f2ac0d3be4bde96ddeb2cacaff1599bcca50
--- /dev/null
+++ b/records/sd1/sweep/69d637049641cc3f3449/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16abed155b6552e222121824d662590618118b5427e7230acd226fcd8accfd1e
+size 29308
diff --git a/records/sd1/sweep/69d76f36ff891a0f99a1/output.webp b/records/sd1/sweep/69d76f36ff891a0f99a1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bc37f9e5acce8df8080f7caa5183c0440bb72aba
--- /dev/null
+++ b/records/sd1/sweep/69d76f36ff891a0f99a1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44e46dcf8b8f4f57832d61b6bb89b043e099a90148102e7b80f5dd97e3f638df
+size 57126
diff --git a/records/sd1/sweep/72f10af829385f698900/record.json b/records/sd1/sweep/72f10af829385f698900/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..44b714833372dfb3baf0429989ff1ed4fe277090
--- /dev/null
+++ b/records/sd1/sweep/72f10af829385f698900/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72f10af829385f698900",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.mid",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ce94a57616ffc9cfc76be6f51958016cd0f6d05bdb4ba11f903feac67c5f758f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ce94a57616ffc9cfc76be6f51958016cd0f6d05bdb4ba11f903feac67c5f758f",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "6452aebd9e5032e2a83f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f3d8ac2f466462da67/interpretations.jsonl b/records/sd1/sweep/72f3d8ac2f466462da67/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6a05f2672be45ece86890dbcd4f9421c0868df76
--- /dev/null
+++ b/records/sd1/sweep/72f3d8ac2f466462da67/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72f3d8ac2f466462da67"}, "kind": "caption", "value": "Portrait of bob-haired woman with pale smoothed skin and a dark smudge near the mouth", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:17:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8acb631dae34a04f"}
+{"target": {"record": "72f3d8ac2f466462da67"}, "kind": "change", "value": "Composition stays close but the face becomes flatter and paler with a smeared dark artifact on the mouth and cheek, and the frame edge is lighter and plainer.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:17:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "14438440c34d442e"}
+{"target": {"record": "72f3d8ac2f466462da67"}, "kind": "keywords", "value": ["smeared mouth", "pale skin", "face artifact", "flat lighting", "bob haircut"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:17:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e8bff7156eaa4650"}
+{"target": {"record": "72f3d8ac2f466462da67"}, "kind": "effect_tags", "value": ["distorted", "bright", "subject_kept", "glitch"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:17:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "49a299cd610ced4f"}
diff --git a/records/sd1/sweep/72f3d8ac2f466462da67/measurements.json b/records/sd1/sweep/72f3d8ac2f466462da67/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1a7483ff20dd589cf1200908d079c86d17af62d
--- /dev/null
+++ b/records/sd1/sweep/72f3d8ac2f466462da67/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72f3d8ac2f466462da67",
+ "values": {
+  "cosine_distance": {
+   "value": 0.129226,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.172833,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.116405,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.101609,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.726,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00298,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.374,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0183,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3993,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f3d8ac2f466462da67/record.json b/records/sd1/sweep/72f3d8ac2f466462da67/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f96f9628c2c8dc09c7c8bd4b1fbaf5e8d627338
--- /dev/null
+++ b/records/sd1/sweep/72f3d8ac2f466462da67/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72f3d8ac2f466462da67",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.out_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/2cf9758becc7da416d597a151b553e5857683fa6e3169cb9aaecaa7a839bbac7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2cf9758becc7da416d597a151b553e5857683fa6e3169cb9aaecaa7a839bbac7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "30056dc810a0b572a666"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f3d8ac2f466462da67/workflow.json b/records/sd1/sweep/72f3d8ac2f466462da67/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..823d0c5fae7460a02055805529455e5f68fe87b2
--- /dev/null
+++ b/records/sd1/sweep/72f3d8ac2f466462da67/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72f3d8ac2f466462da67)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.out_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f887128d77cc17d072/measurements.json b/records/sd1/sweep/72f887128d77cc17d072/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c44647e14b4c33d24d435a109c346f354b85cf62
--- /dev/null
+++ b/records/sd1/sweep/72f887128d77cc17d072/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72f887128d77cc17d072",
+ "values": {
+  "std": {
+   "value": 34.478,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01248,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.79,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.009308,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.059153,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.03313,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00925,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.01,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2391,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f887128d77cc17d072/record.json b/records/sd1/sweep/72f887128d77cc17d072/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5d1aadbb560789eba3ef61040c4b636e6697d8c
--- /dev/null
+++ b/records/sd1/sweep/72f887128d77cc17d072/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "72f887128d77cc17d072",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:30:41Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "de797b6838bf486f4621"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f887128d77cc17d072/workflow.json b/records/sd1/sweep/72f887128d77cc17d072/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..87fe477bf04e96fece0b6727c3c5d49049cca5ba
--- /dev/null
+++ b/records/sd1/sweep/72f887128d77cc17d072/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72f887128d77cc17d072)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72fbce8849f449a75909/measurements.json b/records/sd1/sweep/72fbce8849f449a75909/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7810e5b09573058720efaaab369020cb04696c96
--- /dev/null
+++ b/records/sd1/sweep/72fbce8849f449a75909/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72fbce8849f449a75909",
+ "values": {
+  "cosine_distance": {
+   "value": 0.018721,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.025576,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.017821,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.005919,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.93,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00559,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.394,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9903,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1319,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72fbce8849f449a75909/record.json b/records/sd1/sweep/72fbce8849f449a75909/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a13271342b274b498073b69898d39b324f604e12
--- /dev/null
+++ b/records/sd1/sweep/72fbce8849f449a75909/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "72fbce8849f449a75909",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:36Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/fb9b34ff73947008e623e5782e360b7da6dc2b873c35ee77daf3b5d08fab6158.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "fb9b34ff73947008e623e5782e360b7da6dc2b873c35ee77daf3b5d08fab6158",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8e7db73ee7805a9fe8e4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72fbce8849f449a75909/workflow.json b/records/sd1/sweep/72fbce8849f449a75909/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d3f3f5351ffced86f2779b26118aa75579044e6
--- /dev/null
+++ b/records/sd1/sweep/72fbce8849f449a75909/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72fbce8849f449a75909)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/730c5c69715b0987256f/interpretations.jsonl b/records/sd1/sweep/730c5c69715b0987256f/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ed755fbb746f8a17c6518b95a4c0c6966fed2c23
--- /dev/null
+++ b/records/sd1/sweep/730c5c69715b0987256f/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "730c5c69715b0987256f"}, "kind": "caption", "value": "Small flat graphic portrait of a man on a card, centred on beige canvas", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "97d46dd86bddf9e5"}
+{"target": {"record": "730c5c69715b0987256f"}, "kind": "change", "value": "The photo becomes a small, flat cutout-style graphic portrait with red and black shapes, framed like a card on a plain beige surface.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b130f8db4affe0b8"}
+{"target": {"record": "730c5c69715b0987256f"}, "kind": "keywords", "value": ["graphic cutout", "flat illustration", "picture within picture", "beige canvas", "red and black"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bb13c57100b02e1c"}
+{"target": {"record": "730c5c69715b0987256f"}, "kind": "effect_tags", "value": ["style_change", "framing", "poster", "geometric", "new_composition"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c070a53875bc8652"}
diff --git a/records/sd1/sweep/730c5c69715b0987256f/measurements.json b/records/sd1/sweep/730c5c69715b0987256f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cd021d0c87633b2e8c408241da00449f9cc0e4b
--- /dev/null
+++ b/records/sd1/sweep/730c5c69715b0987256f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "730c5c69715b0987256f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.715827,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.627828,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.663119,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.359894,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 40.936,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00627,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 51.51,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8663,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3609,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/730c5c69715b0987256f/record.json b/records/sd1/sweep/730c5c69715b0987256f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0fb9fa271d90e404f75b0f443a2c08a4bed341a7
--- /dev/null
+++ b/records/sd1/sweep/730c5c69715b0987256f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "730c5c69715b0987256f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:42Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.6.0.op",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.mid",
+   "kind": "downsample",
+   "module_type": "Conv2d",
+   "container": "Downsample",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/a2960053c7126875669eb6678a956fff46ea5d5dd33421960ff1013d1e6e54f6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a2960053c7126875669eb6678a956fff46ea5d5dd33421960ff1013d1e6e54f6",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.Downsample.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "7a02063df734642744d2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/730c5c69715b0987256f/workflow.json b/records/sd1/sweep/730c5c69715b0987256f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a79a5847994daf391930545563e89589985f62ac
--- /dev/null
+++ b/records/sd1/sweep/730c5c69715b0987256f/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 730c5c69715b0987256f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.6.0.op\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/730f92eec2f6f3125e0a/interpretations.jsonl b/records/sd1/sweep/730f92eec2f6f3125e0a/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e38fe86650cbd838149decc92c559984e1e38cf7
--- /dev/null
+++ b/records/sd1/sweep/730f92eec2f6f3125e0a/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "730f92eec2f6f3125e0a"}, "kind": "caption", "value": "Pale woman with straighter hair and a rough painted grey backdrop on an aged paper border", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4e19c60a934f2593"}
+{"target": {"record": "730f92eec2f6f3125e0a"}, "kind": "change", "value": "The face becomes paler and softer with lighter eyes, the hair hangs straighter and lighter, and the background turns a slightly lighter grey with rougher edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c49b36e9e67a47c7"}
+{"target": {"record": "730f92eec2f6f3125e0a"}, "kind": "keywords", "value": ["pale portrait", "straight hair", "rough brushwork", "aged paper", "face variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "62754b37b94f30e2"}
+{"target": {"record": "730f92eec2f6f3125e0a"}, "kind": "effect_tags", "value": ["subject_kept", "washed_out", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "917549dde92fe13d"}
diff --git a/records/sd1/sweep/730f92eec2f6f3125e0a/measurements.json b/records/sd1/sweep/730f92eec2f6f3125e0a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..474654802b72e16321fe40e386b6b8b268624aa2
--- /dev/null
+++ b/records/sd1/sweep/730f92eec2f6f3125e0a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "730f92eec2f6f3125e0a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.084012,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.094325,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.036223,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.026495,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.191,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00423,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.706,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9472,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.196,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/730f92eec2f6f3125e0a/record.json b/records/sd1/sweep/730f92eec2f6f3125e0a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e7960b444e16f3287bf3fba01ad6a7cd7a2f6a4
--- /dev/null
+++ b/records/sd1/sweep/730f92eec2f6f3125e0a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "730f92eec2f6f3125e0a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:38Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "mid",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/592e857cd7550b58d64a827b798587876c5e3f05f0683c89b857a58340606f05.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "592e857cd7550b58d64a827b798587876c5e3f05f0683c89b857a58340606f05",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6c4fc65d937b42dd5a7d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/730f92eec2f6f3125e0a/workflow.json b/records/sd1/sweep/730f92eec2f6f3125e0a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..78130d23c2b1ca49784a83626205089e4a3eb9ad
--- /dev/null
+++ b/records/sd1/sweep/730f92eec2f6f3125e0a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 730f92eec2f6f3125e0a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.skip_connection\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7316eb0b2b21066ebd43/interpretations.jsonl b/records/sd1/sweep/7316eb0b2b21066ebd43/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..41d681ea4b2d1c481d943ce8f44a784fa88c319d
--- /dev/null
+++ b/records/sd1/sweep/7316eb0b2b21066ebd43/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7316eb0b2b21066ebd43"}, "kind": "caption", "value": "Flat cartoon illustration of a woman with an auburn bob and red lips on teal", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fc9db52f5a7e081e"}
+{"target": {"record": "7316eb0b2b21066ebd43"}, "kind": "change", "value": "The photo becomes a bold vector-style cartoon with thick black outlines, flat fills, a red mouth and a teal background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5c8b1209ef6dbc7e"}
+{"target": {"record": "7316eb0b2b21066ebd43"}, "kind": "keywords", "value": ["cartoon", "vector illustration", "flat colour", "outline art", "pop art"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f9278c7820e277f4"}
+{"target": {"record": "7316eb0b2b21066ebd43"}, "kind": "effect_tags", "value": ["style_change", "poster", "saturated", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "59303a6e71b45e74"}
diff --git a/records/sd1/sweep/7316eb0b2b21066ebd43/measurements.json b/records/sd1/sweep/7316eb0b2b21066ebd43/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..27dbda9ccaa6c482b3f2b0fd8a35b3d0dae8a796
--- /dev/null
+++ b/records/sd1/sweep/7316eb0b2b21066ebd43/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7316eb0b2b21066ebd43",
+ "values": {
+  "cosine_distance": {
+   "value": 0.524726,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.563747,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.695964,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.316029,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 88.257,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01166,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 63.582,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9768,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0564,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7316eb0b2b21066ebd43/record.json b/records/sd1/sweep/7316eb0b2b21066ebd43/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bcf740f4c74a782d4c56ae176bc44bf358a459f
--- /dev/null
+++ b/records/sd1/sweep/7316eb0b2b21066ebd43/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7316eb0b2b21066ebd43",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.norm",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/7ea8b8982bded3ef39b56424b2d047c98edcf565165ff06a0170073ad807bdf3.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7ea8b8982bded3ef39b56424b2d047c98edcf565165ff06a0170073ad807bdf3",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "022ed04934e773b41c27"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7316eb0b2b21066ebd43/workflow.json b/records/sd1/sweep/7316eb0b2b21066ebd43/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6560749fb6322081ace5521b407601886cec19c9
--- /dev/null
+++ b/records/sd1/sweep/7316eb0b2b21066ebd43/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7316eb0b2b21066ebd43)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.norm\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73219689c6028d6fc083/interpretations.jsonl b/records/sd1/sweep/73219689c6028d6fc083/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6e40e557f8ccab0a2a4fe927030201c0e34b6fda
--- /dev/null
+++ b/records/sd1/sweep/73219689c6028d6fc083/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "73219689c6028d6fc083"}, "kind": "caption", "value": "Full-frame green crackle texture like dried paint or cracked leather, with no subject.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "546cc9a8f8e0ec64"}
+{"target": {"record": "73219689c6028d6fc083"}, "kind": "change", "value": "The portrait is replaced entirely by a uniform teal-green surface covered in fine dark crack lines.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "731d7a7fd7d906d4"}
+{"target": {"record": "73219689c6028d6fc083"}, "kind": "keywords", "value": ["crackle texture", "cracked paint", "green surface", "craquelure", "no subject"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "966842449ef80df2"}
+{"target": {"record": "73219689c6028d6fc083"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "texture", "pattern", "colour_shift", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "43cf2a3d52276db4"}
diff --git a/records/sd1/sweep/73219689c6028d6fc083/measurements.json b/records/sd1/sweep/73219689c6028d6fc083/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f2bf3ecd8f504dbc7d65b1ab4241e3f0dc7740d
--- /dev/null
+++ b/records/sd1/sweep/73219689c6028d6fc083/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "73219689c6028d6fc083",
+ "values": {
+  "cosine_distance": {
+   "value": 0.841017,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.987514,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.915217,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.527945,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 14.357,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.43404,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "noise"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "noise"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 83.295,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.6991,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1605,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73219689c6028d6fc083/record.json b/records/sd1/sweep/73219689c6028d6fc083/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b78be53123dd855b16cbc8be8a783c09fb5872a3
--- /dev/null
+++ b/records/sd1/sweep/73219689c6028d6fc083/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73219689c6028d6fc083",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/15e083d258100fed21fb036bd9140790c32d43921ba2fbe454968969a9b7f2c8.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "15e083d258100fed21fb036bd9140790c32d43921ba2fbe454968969a9b7f2c8",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "7601338b442966b3b54b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73219689c6028d6fc083/workflow.json b/records/sd1/sweep/73219689c6028d6fc083/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9903c20050ecdb0a1ba9a9af2420854494d4e945
--- /dev/null
+++ b/records/sd1/sweep/73219689c6028d6fc083/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73219689c6028d6fc083)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.in_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73226bc2d0661341572c/measurements.json b/records/sd1/sweep/73226bc2d0661341572c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..63c7740d16f5ee29f3bf88cc61e9d731cb7bc4e5
--- /dev/null
+++ b/records/sd1/sweep/73226bc2d0661341572c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73226bc2d0661341572c",
+ "values": {
+  "std": {
+   "value": 53.109,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01533,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.573,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.118545,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.217172,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.206229,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027183,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9907,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1754,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73226bc2d0661341572c/record.json b/records/sd1/sweep/73226bc2d0661341572c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..47b0f0dde9eaccf7c77af226e5e528c0d3cb0167
--- /dev/null
+++ b/records/sd1/sweep/73226bc2d0661341572c/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "73226bc2d0661341572c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:00:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "medium",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_noise",
+   0.2,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73226bc2d0661341572c/workflow.json b/records/sd1/sweep/73226bc2d0661341572c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b59bf5f1c98ee2a778bfb1f157d25c364e342ae
--- /dev/null
+++ b/records/sd1/sweep/73226bc2d0661341572c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73226bc2d0661341572c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732b393b4b7ccb9f3acf/measurements.json b/records/sd1/sweep/732b393b4b7ccb9f3acf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..211c3380c3be4c0cafd53cff779fa5959082cff2
--- /dev/null
+++ b/records/sd1/sweep/732b393b4b7ccb9f3acf/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "732b393b4b7ccb9f3acf",
+ "values": {
+  "std": {
+   "value": 31.075,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02941,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.789,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.469716,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.562533,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.505553,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.148121,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9913,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0138,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732b393b4b7ccb9f3acf/record.json b/records/sd1/sweep/732b393b4b7ccb9f3acf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef5838af4367340f9b6a23b551fe4633212998b8
--- /dev/null
+++ b/records/sd1/sweep/732b393b4b7ccb9f3acf/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "732b393b4b7ccb9f3acf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:05:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "rotate",
+   180.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5dbb0af64b34319bc504"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732b393b4b7ccb9f3acf/workflow.json b/records/sd1/sweep/732b393b4b7ccb9f3acf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..010021633c6add9f91237ab04bb1729e9d77711e
--- /dev/null
+++ b/records/sd1/sweep/732b393b4b7ccb9f3acf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 732b393b4b7ccb9f3acf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732d8899e03ebf59f70f/measurements.json b/records/sd1/sweep/732d8899e03ebf59f70f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..635c336d13c15019e6152d37a1994f443312da8c
--- /dev/null
+++ b/records/sd1/sweep/732d8899e03ebf59f70f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "732d8899e03ebf59f70f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.170965,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.244929,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.066927,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.067981,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.437,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00588,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.134,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9677,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0986,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732d8899e03ebf59f70f/record.json b/records/sd1/sweep/732d8899e03ebf59f70f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fac67f6c19c68ebf7ac9339b78c19aeac72b57e9
--- /dev/null
+++ b/records/sd1/sweep/732d8899e03ebf59f70f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "732d8899e03ebf59f70f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/4dd5d016e4b80ce968f4461a3d29f5422be8bee3db9274dcdb7896b41694a2cd.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4dd5d016e4b80ce968f4461a3d29f5422be8bee3db9274dcdb7896b41694a2cd",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "466270109d97b2c82610"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732d8899e03ebf59f70f/workflow.json b/records/sd1/sweep/732d8899e03ebf59f70f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..611ee3988bba13d8f4332d6dc4c1fc0aaa418119
--- /dev/null
+++ b/records/sd1/sweep/732d8899e03ebf59f70f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 732d8899e03ebf59f70f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732f94fc011903bcaaf6/measurements.json b/records/sd1/sweep/732f94fc011903bcaaf6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d417cd63eee8efae95df1c9c645e80fe7541d435
--- /dev/null
+++ b/records/sd1/sweep/732f94fc011903bcaaf6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "732f94fc011903bcaaf6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.168895,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.226625,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.13096,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.14993,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.964,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00585,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.075,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0104,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1103,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732f94fc011903bcaaf6/record.json b/records/sd1/sweep/732f94fc011903bcaaf6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d61956c6503fe5d198ff76ae47a5ec12115bc59
--- /dev/null
+++ b/records/sd1/sweep/732f94fc011903bcaaf6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "732f94fc011903bcaaf6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/55e75a3dccbbfaa4a8c77bb4e48c247b494487ff88f96e171e018ba7045a70a5.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "55e75a3dccbbfaa4a8c77bb4e48c247b494487ff88f96e171e018ba7045a70a5",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "65c5b9106d0ea8f9755b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/732f94fc011903bcaaf6/workflow.json b/records/sd1/sweep/732f94fc011903bcaaf6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..52e1f6d5954a16f8f897bef0e96093482fc25eb3
--- /dev/null
+++ b/records/sd1/sweep/732f94fc011903bcaaf6/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 732f94fc011903bcaaf6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.out_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7336f9026ff06236cd6b/measurements.json b/records/sd1/sweep/7336f9026ff06236cd6b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8181605320fc10a3a21e293152887da3d0d92c94
--- /dev/null
+++ b/records/sd1/sweep/7336f9026ff06236cd6b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7336f9026ff06236cd6b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.444187,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.609696,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.667184,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.416184,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.155,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00416,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 46.937,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9831,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2152,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7336f9026ff06236cd6b/record.json b/records/sd1/sweep/7336f9026ff06236cd6b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45d4c2d1db228c66195fc8ecdccab0bb930ce45b
--- /dev/null
+++ b/records/sd1/sweep/7336f9026ff06236cd6b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7336f9026ff06236cd6b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:00Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b126cfc4d092545f533f27a887d5d767d4830ce739a618bed03caa6ed6b9d50a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b126cfc4d092545f533f27a887d5d767d4830ce739a618bed03caa6ed6b9d50a",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "a80065bfce2b24e65693"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7336f9026ff06236cd6b/workflow.json b/records/sd1/sweep/7336f9026ff06236cd6b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9cb66d8ebb88bc4a7c8388cc449c91e71efdeaf1
--- /dev/null
+++ b/records/sd1/sweep/7336f9026ff06236cd6b/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7336f9026ff06236cd6b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.skip_connection\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73376160781fd728bc55/measurements.json b/records/sd1/sweep/73376160781fd728bc55/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..78e2702a1eaf9999b5880d60a9b1e680a876da44
--- /dev/null
+++ b/records/sd1/sweep/73376160781fd728bc55/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73376160781fd728bc55",
+ "values": {
+  "cosine_distance": {
+   "value": 0.311068,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.351508,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.254618,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.107638,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.189,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00579,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.143,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0401,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.107,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73376160781fd728bc55/record.json b/records/sd1/sweep/73376160781fd728bc55/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..adb226d844dde768e3ccebbb32d4b71c1863240d
--- /dev/null
+++ b/records/sd1/sweep/73376160781fd728bc55/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73376160781fd728bc55",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/d77511dc6a654a3008e8570b518608a3782f46e7e108e0b724653239efb8bdb1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d77511dc6a654a3008e8570b518608a3782f46e7e108e0b724653239efb8bdb1",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "61be08834af11925ab86"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73376160781fd728bc55/workflow.json b/records/sd1/sweep/73376160781fd728bc55/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..579eced315376b9c81469f817f726e3fb5a904e6
--- /dev/null
+++ b/records/sd1/sweep/73376160781fd728bc55/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73376160781fd728bc55)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/733a3f07eeef31d34c69/measurements.json b/records/sd1/sweep/733a3f07eeef31d34c69/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e97bef1ba30a985becd95342e14dbc110b285393
--- /dev/null
+++ b/records/sd1/sweep/733a3f07eeef31d34c69/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "733a3f07eeef31d34c69",
+ "values": {
+  "cosine_distance": {
+   "value": 0.189027,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.198655,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.087961,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.049219,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.159,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00594,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.822,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9753,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.23,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/733a3f07eeef31d34c69/record.json b/records/sd1/sweep/733a3f07eeef31d34c69/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e914e6f805873389f7b49fef17fc9fc8706e47f6
--- /dev/null
+++ b/records/sd1/sweep/733a3f07eeef31d34c69/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "733a3f07eeef31d34c69",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.proj_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5c1e345bc4b5b69726a7d6eb548b5b7c8bd08d5b60c6642092141ac97c80ad7d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5c1e345bc4b5b69726a7d6eb548b5b7c8bd08d5b60c6642092141ac97c80ad7d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "acc1ab17eba294d1e719"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/733a3f07eeef31d34c69/workflow.json b/records/sd1/sweep/733a3f07eeef31d34c69/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9aecb4a36e8f39380c1af233b436907ebced7270
--- /dev/null
+++ b/records/sd1/sweep/733a3f07eeef31d34c69/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 733a3f07eeef31d34c69)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.proj_out\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/733f6770a9639e7ab512/interpretations.jsonl b/records/sd1/sweep/733f6770a9639e7ab512/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7ee129f024f22f912bc0c57c90b395ff805bc102
--- /dev/null
+++ b/records/sd1/sweep/733f6770a9639e7ab512/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "733f6770a9639e7ab512"}, "kind": "caption", "value": "High-contrast black-and-white portrait of a pale person with tousled dark hair, circular vignette", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f536cfea176ccfdd"}
+{"target": {"record": "733f6770a9639e7ab512"}, "kind": "change", "value": "The image becomes a stark, pale monochrome portrait of a slimmer, androgynous face inside a light circular vignette on textured paper.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cd477723e0f602c0"}
+{"target": {"record": "733f6770a9639e7ab512"}, "kind": "keywords", "value": ["monochrome portrait", "circular vignette", "pale skin", "textured paper", "high contrast"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "823d28cca8a8e6e1"}
+{"target": {"record": "733f6770a9639e7ab512"}, "kind": "effect_tags", "value": ["desaturated", "high_contrast", "framing", "subject_kept", "style_change", "bright"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8d9a5e60bab7b758"}
diff --git a/records/sd1/sweep/733f6770a9639e7ab512/measurements.json b/records/sd1/sweep/733f6770a9639e7ab512/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c42cb4fc1f1cf8ea20e0cb9a6c880e684cd72c88
--- /dev/null
+++ b/records/sd1/sweep/733f6770a9639e7ab512/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "733f6770a9639e7ab512",
+ "values": {
+  "cosine_distance": {
+   "value": 0.255425,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.341599,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.192481,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.161231,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.71,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00641,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.238,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9088,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6667,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/733f6770a9639e7ab512/record.json b/records/sd1/sweep/733f6770a9639e7ab512/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4be38ea63dc1fc1650e78fc2199766b3e9c6ead2
--- /dev/null
+++ b/records/sd1/sweep/733f6770a9639e7ab512/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "733f6770a9639e7ab512",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/335a68469f98dba1c418785675fc67dd749489525c9ff071ff65dc30858f2608.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "335a68469f98dba1c418785675fc67dd749489525c9ff071ff65dc30858f2608",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "ee769d16a580f2ac0e98"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/733f6770a9639e7ab512/workflow.json b/records/sd1/sweep/733f6770a9639e7ab512/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..61779a3a44e0b002bdc9cf83932f993521d32f37
--- /dev/null
+++ b/records/sd1/sweep/733f6770a9639e7ab512/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 733f6770a9639e7ab512)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/734afedf39a825340b59/measurements.json b/records/sd1/sweep/734afedf39a825340b59/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..29f4586be1f543dc5010d9a06dcf48f0f0f17cc0
--- /dev/null
+++ b/records/sd1/sweep/734afedf39a825340b59/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "734afedf39a825340b59",
+ "values": {
+  "cosine_distance": {
+   "value": 0.101432,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.163694,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.050945,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.042285,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.245,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00582,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 16.042,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.989,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0577,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/734afedf39a825340b59/record.json b/records/sd1/sweep/734afedf39a825340b59/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..76ca0e7db04d30decb0ffb9884304de9587c757e
--- /dev/null
+++ b/records/sd1/sweep/734afedf39a825340b59/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "734afedf39a825340b59",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e6daa7ef97062c050a2ddcd42e4090f8e81dc216c2ec727d48868e2801b9764b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e6daa7ef97062c050a2ddcd42e4090f8e81dc216c2ec727d48868e2801b9764b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7dcbf418d80b8a4f768d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/734afedf39a825340b59/workflow.json b/records/sd1/sweep/734afedf39a825340b59/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5742e6632a7ab7ba8c6ac795b2e359f647557b2f
--- /dev/null
+++ b/records/sd1/sweep/734afedf39a825340b59/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 734afedf39a825340b59)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/734b81edd74addba3f1d/measurements.json b/records/sd1/sweep/734b81edd74addba3f1d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..11755692648d3fb63dc76efa6dbca6d9250a77c7
--- /dev/null
+++ b/records/sd1/sweep/734b81edd74addba3f1d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "734b81edd74addba3f1d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0483,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.078077,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.029621,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.009061,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.242,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00602,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.887,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9931,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1044,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/734b81edd74addba3f1d/record.json b/records/sd1/sweep/734b81edd74addba3f1d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..de5315bf53ab7ac9c78dc1939e5885cfb4d7f63f
--- /dev/null
+++ b/records/sd1/sweep/734b81edd74addba3f1d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "734b81edd74addba3f1d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:36Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7469e0622a80f1cd8147c0d627a28ba2bd88267b956eb673d09994aeed5575d9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7469e0622a80f1cd8147c0d627a28ba2bd88267b956eb673d09994aeed5575d9",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "643e3ea79e9333537d5c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/734b81edd74addba3f1d/workflow.json b/records/sd1/sweep/734b81edd74addba3f1d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..86a598b78ff9e04809583ef5c3aa383f4dba426c
--- /dev/null
+++ b/records/sd1/sweep/734b81edd74addba3f1d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 734b81edd74addba3f1d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73559c1b91a5fc2af581/measurements.json b/records/sd1/sweep/73559c1b91a5fc2af581/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..920834d55a8d80bc30ed2e4bf7551d7c7ea24c5c
--- /dev/null
+++ b/records/sd1/sweep/73559c1b91a5fc2af581/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73559c1b91a5fc2af581",
+ "values": {
+  "cosine_distance": {
+   "value": 0.145942,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.182752,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.138892,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.072961,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.12,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00355,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.642,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9427,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.058,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73559c1b91a5fc2af581/record.json b/records/sd1/sweep/73559c1b91a5fc2af581/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c08c7eac9b53ce03d654bbbfdb7c1f023731a1a
--- /dev/null
+++ b/records/sd1/sweep/73559c1b91a5fc2af581/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73559c1b91a5fc2af581",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9f0140754c549b811fae4e4dfd91bbb8111cd1f42a5d8ddb050e1f358c80121b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9f0140754c549b811fae4e4dfd91bbb8111cd1f42a5d8ddb050e1f358c80121b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e9404e0f08c96f16ab0b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73559c1b91a5fc2af581/workflow.json b/records/sd1/sweep/73559c1b91a5fc2af581/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f8b3f689fb50de37656cdde744c8d57c3fd7da3
--- /dev/null
+++ b/records/sd1/sweep/73559c1b91a5fc2af581/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73559c1b91a5fc2af581)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/736285ae44fdda792df5/interpretations.jsonl b/records/sd1/sweep/736285ae44fdda792df5/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..aad2187aeb66e1070e8de58cbe5306414b072789
--- /dev/null
+++ b/records/sd1/sweep/736285ae44fdda792df5/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "736285ae44fdda792df5"}, "kind": "caption", "value": "Loose painterly abstraction of peach, black and blue-grey cloudy brushwork with a faint face", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "aa972fa7b71b942b"}
+{"target": {"record": "736285ae44fdda792df5"}, "kind": "change", "value": "The crisp portrait breaks down into expressive smudgy brushwork where only a ghostly peach face and dark hair mass are hinted at among cloudy blue-grey strokes.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fc921ffa9c10e7f9"}
+{"target": {"record": "736285ae44fdda792df5"}, "kind": "keywords", "value": ["expressive brushwork", "ghost face", "abstract portrait", "cloudy paint", "impressionist"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5593e2bbe436838a"}
+{"target": {"record": "736285ae44fdda792df5"}, "kind": "effect_tags", "value": ["painterly", "abstract", "texture", "distorted", "subject_lost"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "29ba36ddcabfbb9b"}
diff --git a/records/sd1/sweep/736285ae44fdda792df5/measurements.json b/records/sd1/sweep/736285ae44fdda792df5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..88d4463309307ecb4bcb9f3b41eef0c89747b90d
--- /dev/null
+++ b/records/sd1/sweep/736285ae44fdda792df5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "736285ae44fdda792df5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.426852,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.62198,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.766883,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.312329,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.375,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00291,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 46.857,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9385,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2166,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/736285ae44fdda792df5/record.json b/records/sd1/sweep/736285ae44fdda792df5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d70b015e4f04d04649848952ab9dd8db3b3614b3
--- /dev/null
+++ b/records/sd1/sweep/736285ae44fdda792df5/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "736285ae44fdda792df5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.3.0.op",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.hi",
+   "kind": "downsample",
+   "module_type": "Conv2d",
+   "container": "Downsample",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5118522a4c6732b026f8c21b1f0912c04cbf688b44c0cc739e604041613e7bce.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5118522a4c6732b026f8c21b1f0912c04cbf688b44c0cc739e604041613e7bce",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.Downsample.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0faa7adbc87a72cdeaa6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/736285ae44fdda792df5/workflow.json b/records/sd1/sweep/736285ae44fdda792df5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..84762c20d388a2c6c0e241edee751bc2cb4f069b
--- /dev/null
+++ b/records/sd1/sweep/736285ae44fdda792df5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 736285ae44fdda792df5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.3.0.op\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/739183a62a8374194c5f/measurements.json b/records/sd1/sweep/739183a62a8374194c5f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef495b3d0653adf68e533f11bd6fa8e231245775
--- /dev/null
+++ b/records/sd1/sweep/739183a62a8374194c5f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "739183a62a8374194c5f",
+ "values": {
+  "std": {
+   "value": 25.169,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0241,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.43,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.447105,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.655716,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.589511,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.148659,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7912,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7077,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/739183a62a8374194c5f/record.json b/records/sd1/sweep/739183a62a8374194c5f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0a9ca45fca38e0c4f6d56f3b77c695d834c911f
--- /dev/null
+++ b/records/sd1/sweep/739183a62a8374194c5f/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "739183a62a8374194c5f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:28:27Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   180.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "3d9b126da9af77f961ad"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/739183a62a8374194c5f/workflow.json b/records/sd1/sweep/739183a62a8374194c5f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..44beed5e6159a427e9bc7e99cc2fdde3ba2fbc1a
--- /dev/null
+++ b/records/sd1/sweep/739183a62a8374194c5f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 739183a62a8374194c5f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73a766727a1845225b2c/measurements.json b/records/sd1/sweep/73a766727a1845225b2c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2029ba749d4387edf7a70a77fdecbdcfb2704571
--- /dev/null
+++ b/records/sd1/sweep/73a766727a1845225b2c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73a766727a1845225b2c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.471006,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.457119,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.358214,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.227368,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.231,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00425,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.011,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9518,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3606,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73a766727a1845225b2c/record.json b/records/sd1/sweep/73a766727a1845225b2c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45bd12013c3ec05087245df7e8cf4556987e3ac8
--- /dev/null
+++ b/records/sd1/sweep/73a766727a1845225b2c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73a766727a1845225b2c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ec646afcee472c64180242039ba7ee53b51443435267b49ce3a7740fe5421a80.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ec646afcee472c64180242039ba7ee53b51443435267b49ce3a7740fe5421a80",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "8aa332de6bffe1497f14"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73a766727a1845225b2c/workflow.json b/records/sd1/sweep/73a766727a1845225b2c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..87498a1b4f7649327b15b5b2eaf60a13180cdca7
--- /dev/null
+++ b/records/sd1/sweep/73a766727a1845225b2c/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73a766727a1845225b2c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73ae6eed500115b0f86a/measurements.json b/records/sd1/sweep/73ae6eed500115b0f86a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c50799bd511f92de4174b761e527b99df235f2ed
--- /dev/null
+++ b/records/sd1/sweep/73ae6eed500115b0f86a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73ae6eed500115b0f86a",
+ "values": {
+  "std": {
+   "value": 50.674,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00675,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.779,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.349389,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.507159,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.42644,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.178597,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9466,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3382,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73ae6eed500115b0f86a/record.json b/records/sd1/sweep/73ae6eed500115b0f86a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f3cff6b6f507541e46db05d472ff089bb76d5c4
--- /dev/null
+++ b/records/sd1/sweep/73ae6eed500115b0f86a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "73ae6eed500115b0f86a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:40:42Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "grow",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "scale",
+   1.4,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "582635fd10a6325f001d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73ae6eed500115b0f86a/workflow.json b/records/sd1/sweep/73ae6eed500115b0f86a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd3b0777f00ad4b5ede85d39071440277789cdb7
--- /dev/null
+++ b/records/sd1/sweep/73ae6eed500115b0f86a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73ae6eed500115b0f86a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73b675c11fcf0dd9d856/measurements.json b/records/sd1/sweep/73b675c11fcf0dd9d856/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d5f91000b715414d2786f115492d496e232613f
--- /dev/null
+++ b/records/sd1/sweep/73b675c11fcf0dd9d856/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73b675c11fcf0dd9d856",
+ "values": {
+  "cosine_distance": {
+   "value": 0.304404,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.411161,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.234726,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.128899,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 42.301,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00434,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.26,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9109,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1635,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73b675c11fcf0dd9d856/record.json b/records/sd1/sweep/73b675c11fcf0dd9d856/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f17c38bb4e911c89586da055a44c0b8b50691b6
--- /dev/null
+++ b/records/sd1/sweep/73b675c11fcf0dd9d856/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73b675c11fcf0dd9d856",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/6a4afb3077e849d47e4ddb028483a2a985d626550bbe37f6975ca53f2fedfc38.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6a4afb3077e849d47e4ddb028483a2a985d626550bbe37f6975ca53f2fedfc38",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4f515942d51ffa707076"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73b675c11fcf0dd9d856/workflow.json b/records/sd1/sweep/73b675c11fcf0dd9d856/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a8d10d5c8a767144435e47b29150d30380d8673
--- /dev/null
+++ b/records/sd1/sweep/73b675c11fcf0dd9d856/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73b675c11fcf0dd9d856)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.proj_out\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73bb2db8e85e07039d09/measurements.json b/records/sd1/sweep/73bb2db8e85e07039d09/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5f437bfcc81adbb11e9c302e4eb4c0496a50836
--- /dev/null
+++ b/records/sd1/sweep/73bb2db8e85e07039d09/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73bb2db8e85e07039d09",
+ "values": {
+  "cosine_distance": {
+   "value": 0.231068,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.39949,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.189217,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.109359,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.873,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0036,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.005,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.025,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1527,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73bb2db8e85e07039d09/record.json b/records/sd1/sweep/73bb2db8e85e07039d09/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..abe162a9a5f1431d7a8a2312844f9baa3bd79d13
--- /dev/null
+++ b/records/sd1/sweep/73bb2db8e85e07039d09/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73bb2db8e85e07039d09",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:25Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ff892aea6ae5a9739329f2374de88efd46937f2568230b2e4f8a485f2ca21eba.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ff892aea6ae5a9739329f2374de88efd46937f2568230b2e4f8a485f2ca21eba",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c73518cf5f82533f8e55"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73bb2db8e85e07039d09/workflow.json b/records/sd1/sweep/73bb2db8e85e07039d09/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..658b63dcffea5ab3712e6f9473e31e7a27f31682
--- /dev/null
+++ b/records/sd1/sweep/73bb2db8e85e07039d09/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73bb2db8e85e07039d09)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.proj_in\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73bcfe7e51da7c7d9096/measurements.json b/records/sd1/sweep/73bcfe7e51da7c7d9096/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d0a87fee48cf65ae795003633468731e275c97b
--- /dev/null
+++ b/records/sd1/sweep/73bcfe7e51da7c7d9096/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73bcfe7e51da7c7d9096",
+ "values": {
+  "cosine_distance": {
+   "value": 0.115158,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.137056,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.145305,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.103847,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.182,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00539,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 16.145,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9793,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1402,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73bcfe7e51da7c7d9096/record.json b/records/sd1/sweep/73bcfe7e51da7c7d9096/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e61f5c50c91ee0893c5ec487b2b73abfdfc58df
--- /dev/null
+++ b/records/sd1/sweep/73bcfe7e51da7c7d9096/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73bcfe7e51da7c7d9096",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.proj_in",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e0b13d76efa8908260bed0dc2ffc0ba780e32d8b7bfa124d78b7398ac9c6d43d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e0b13d76efa8908260bed0dc2ffc0ba780e32d8b7bfa124d78b7398ac9c6d43d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c14f9dcd6a11061a402e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73bcfe7e51da7c7d9096/workflow.json b/records/sd1/sweep/73bcfe7e51da7c7d9096/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbea633f61e8838f9c9106c53de37c3b66ecf24c
--- /dev/null
+++ b/records/sd1/sweep/73bcfe7e51da7c7d9096/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73bcfe7e51da7c7d9096)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.proj_in\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73be2c1d5c1e2b169da6/measurements.json b/records/sd1/sweep/73be2c1d5c1e2b169da6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4024d68181c674f1b5ceb9d2776ace04d903491
--- /dev/null
+++ b/records/sd1/sweep/73be2c1d5c1e2b169da6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73be2c1d5c1e2b169da6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.206705,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.402374,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.316084,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.290181,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.742,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00328,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.629,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0169,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4167,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73be2c1d5c1e2b169da6/record.json b/records/sd1/sweep/73be2c1d5c1e2b169da6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5a1f241a1faf6be6a0e2e017beebb99d4c06386
--- /dev/null
+++ b/records/sd1/sweep/73be2c1d5c1e2b169da6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73be2c1d5c1e2b169da6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/a7eadecdc1fe72e648d8f7b342880c78144da7c0034c41759639948f98ccf976.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a7eadecdc1fe72e648d8f7b342880c78144da7c0034c41759639948f98ccf976",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "b6ced5e41c611232508e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73be2c1d5c1e2b169da6/workflow.json b/records/sd1/sweep/73be2c1d5c1e2b169da6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a705e218f0fa00c265c79486bc211f37f1491b55
--- /dev/null
+++ b/records/sd1/sweep/73be2c1d5c1e2b169da6/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73be2c1d5c1e2b169da6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73c85a7d3b4381917ba8/measurements.json b/records/sd1/sweep/73c85a7d3b4381917ba8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..246c5cb2f43efef1ebd8b8082392624e7f800830
--- /dev/null
+++ b/records/sd1/sweep/73c85a7d3b4381917ba8/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73c85a7d3b4381917ba8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.230433,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.209522,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.165153,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.169361,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 73.171,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00527,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.744,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8958,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3105,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73c85a7d3b4381917ba8/record.json b/records/sd1/sweep/73c85a7d3b4381917ba8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e92cf38964f0c4c3b6c7d23bed31abe33bf3ff7
--- /dev/null
+++ b/records/sd1/sweep/73c85a7d3b4381917ba8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73c85a7d3b4381917ba8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:44Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f4a578170506c3443a107d6d1c74bf337e10c6ea50253244f6eab7c43ebc0f41.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f4a578170506c3443a107d6d1c74bf337e10c6ea50253244f6eab7c43ebc0f41",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "b6319c3608b4f72e3dea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73c85a7d3b4381917ba8/workflow.json b/records/sd1/sweep/73c85a7d3b4381917ba8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ace1315a1db588f3a00d2b951a3353950e38aec
--- /dev/null
+++ b/records/sd1/sweep/73c85a7d3b4381917ba8/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73c85a7d3b4381917ba8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.proj_in\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73e10a6dd3a667d8ba1d/interpretations.jsonl b/records/sd1/sweep/73e10a6dd3a667d8ba1d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..fe18c64a9f217861563e7edd68ee3bbe96905934
--- /dev/null
+++ b/records/sd1/sweep/73e10a6dd3a667d8ba1d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "73e10a6dd3a667d8ba1d"}, "kind": "caption", "value": "Painted woman with wavy hair, sharper features and blue eyes on a grey panel", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "04589c15c9e5a4a7"}
+{"target": {"record": "73e10a6dd3a667d8ba1d"}, "kind": "change", "value": "Almost no visible change; the features become slightly sharper and the eyes bluer.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7f26b24083078a6c"}
+{"target": {"record": "73e10a6dd3a667d8ba1d"}, "kind": "keywords", "value": ["portrait", "minor variation", "blue eyes", "painted portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "76d61523410b1ce3"}
+{"target": {"record": "73e10a6dd3a667d8ba1d"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "087230deb5ef41d1"}
diff --git a/records/sd1/sweep/73e10a6dd3a667d8ba1d/measurements.json b/records/sd1/sweep/73e10a6dd3a667d8ba1d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d3ff13d3668fe1f20455e4f762f25f736fa10630
--- /dev/null
+++ b/records/sd1/sweep/73e10a6dd3a667d8ba1d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73e10a6dd3a667d8ba1d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.075562,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.103612,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.034743,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.016092,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.916,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00699,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.532,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0039,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1456,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73e10a6dd3a667d8ba1d/record.json b/records/sd1/sweep/73e10a6dd3a667d8ba1d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2c944f72336c14917e349af5348bf02b65d0c7d
--- /dev/null
+++ b/records/sd1/sweep/73e10a6dd3a667d8ba1d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73e10a6dd3a667d8ba1d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:59Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5952f120b2e5dde1f250a4a6f45d720beb5dc9fea767bc47d6b13fa19cd9f891.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5952f120b2e5dde1f250a4a6f45d720beb5dc9fea767bc47d6b13fa19cd9f891",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "9ee1060b8634c556231c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73e10a6dd3a667d8ba1d/workflow.json b/records/sd1/sweep/73e10a6dd3a667d8ba1d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7769c4cdf9626232ea7200a5a841b340068804b
--- /dev/null
+++ b/records/sd1/sweep/73e10a6dd3a667d8ba1d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73e10a6dd3a667d8ba1d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.out_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73e338bcf73603b1089a/measurements.json b/records/sd1/sweep/73e338bcf73603b1089a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ac3f1b8ab12e9e7f79f9c6217bb16cbabc7755c
--- /dev/null
+++ b/records/sd1/sweep/73e338bcf73603b1089a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73e338bcf73603b1089a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.048987,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.066661,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.047657,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.011575,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.307,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0054,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.087,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0089,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1289,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73e338bcf73603b1089a/record.json b/records/sd1/sweep/73e338bcf73603b1089a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..daba0c1b367988f4d84387da2b485e0bfc6cd8fb
--- /dev/null
+++ b/records/sd1/sweep/73e338bcf73603b1089a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73e338bcf73603b1089a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/fff637a976c252d78f337b4d94f17293f03188c7d2c94c3c498612936cc27868.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "fff637a976c252d78f337b4d94f17293f03188c7d2c94c3c498612936cc27868",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f14c8ddcdba429a40346"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73e338bcf73603b1089a/workflow.json b/records/sd1/sweep/73e338bcf73603b1089a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6589439eaf96ab96129715e539780f067e61306d
--- /dev/null
+++ b/records/sd1/sweep/73e338bcf73603b1089a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73e338bcf73603b1089a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73ec2da2367997be24b8/measurements.json b/records/sd1/sweep/73ec2da2367997be24b8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bb272f5d6fa9d05f6fe9f74a75e49635a546382
--- /dev/null
+++ b/records/sd1/sweep/73ec2da2367997be24b8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73ec2da2367997be24b8",
+ "values": {
+  "std": {
+   "value": 44.341,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01336,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.441,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.325984,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.4281,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.118064,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.058081,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0408,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0125,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73ec2da2367997be24b8/record.json b/records/sd1/sweep/73ec2da2367997be24b8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..04eda5dd430a42966d27cc2409026be89b499875
--- /dev/null
+++ b/records/sd1/sweep/73ec2da2367997be24b8/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "73ec2da2367997be24b8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:49:00Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "damp",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "multiply",
+   0.5,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "71292d0a3a0aa71889fb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73ec2da2367997be24b8/workflow.json b/records/sd1/sweep/73ec2da2367997be24b8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f74fab602d5f201160e5253c12347639b8dee0e
--- /dev/null
+++ b/records/sd1/sweep/73ec2da2367997be24b8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73ec2da2367997be24b8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73f220e9adfcaa1cf470/interpretations.jsonl b/records/sd1/sweep/73f220e9adfcaa1cf470/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..cfd5bf4ea0c406c6f0f9b66615200fd72cb5faec
--- /dev/null
+++ b/records/sd1/sweep/73f220e9adfcaa1cf470/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "73f220e9adfcaa1cf470"}, "kind": "caption", "value": "Smeared, ghosted portrait of a curly-haired woman with doubled eyes and mouth", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "24bc167c7278d176"}
+{"target": {"record": "73f220e9adfcaa1cf470"}, "kind": "change", "value": "The image dissolves into an out-of-focus smear with ghosted, doubled facial features and curly brown hair.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "400e02826b4509b9"}
+{"target": {"record": "73f220e9adfcaa1cf470"}, "kind": "keywords", "value": ["smeared face", "ghosting", "blurry portrait", "doubled features"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ae23bf11d9fc96e7"}
+{"target": {"record": "73f220e9adfcaa1cf470"}, "kind": "effect_tags", "value": ["blur", "distorted", "washed_out", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "47a946dbd648679b"}
diff --git a/records/sd1/sweep/73f220e9adfcaa1cf470/measurements.json b/records/sd1/sweep/73f220e9adfcaa1cf470/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f246389ac8a948ad06a4beb0769471403f42e22
--- /dev/null
+++ b/records/sd1/sweep/73f220e9adfcaa1cf470/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73f220e9adfcaa1cf470",
+ "values": {
+  "cosine_distance": {
+   "value": 0.147022,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.383911,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.259713,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.135412,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.302,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00148,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.651,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0343,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0299,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73f220e9adfcaa1cf470/record.json b/records/sd1/sweep/73f220e9adfcaa1cf470/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..48add93c36b713e08340f58753c156e18724273a
--- /dev/null
+++ b/records/sd1/sweep/73f220e9adfcaa1cf470/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73f220e9adfcaa1cf470",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/b79ce1a6a1661a46d655ed1d4f8889983b391ed53516f9f04d45d185bdbc7f67.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b79ce1a6a1661a46d655ed1d4f8889983b391ed53516f9f04d45d185bdbc7f67",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "457e9637354453c7c519"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73f220e9adfcaa1cf470/workflow.json b/records/sd1/sweep/73f220e9adfcaa1cf470/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bf738bc40f2b3f2f02b62dda830237775268e3a
--- /dev/null
+++ b/records/sd1/sweep/73f220e9adfcaa1cf470/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73f220e9adfcaa1cf470)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73f665316e96d5258c89/interpretations.jsonl b/records/sd1/sweep/73f665316e96d5258c89/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e06ef33f99374dfa86da34e9ff176588e3876c9b
--- /dev/null
+++ b/records/sd1/sweep/73f665316e96d5258c89/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "73f665316e96d5258c89"}, "kind": "caption", "value": "Uncanny doll-like face with purple eyeshadow inside a hexagonal teal and brown geometric frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ddf30fdbf57eae6c"}
+{"target": {"record": "73f665316e96d5258c89"}, "kind": "change", "value": "The portrait becomes a crude, smooth, mask-like CG face with an open mouth, set inside flat geometric teal shapes, losing the photographic look entirely.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d6932f2f0db44ee6"}
+{"target": {"record": "73f665316e96d5258c89"}, "kind": "keywords", "value": ["uncanny face", "mask", "geometric frame", "cg render", "teal shapes"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0d2d6973465b9c8b"}
+{"target": {"record": "73f665316e96d5258c89"}, "kind": "effect_tags", "value": ["style_change", "geometric", "surreal", "distorted", "abstract", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e2566b806766127c"}
diff --git a/records/sd1/sweep/73f665316e96d5258c89/measurements.json b/records/sd1/sweep/73f665316e96d5258c89/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e35b60732245371e81d367dd0f969686ab6ba404
--- /dev/null
+++ b/records/sd1/sweep/73f665316e96d5258c89/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "73f665316e96d5258c89",
+ "values": {
+  "cosine_distance": {
+   "value": 0.511634,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.553472,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.605613,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.421967,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 74.875,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00309,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 53.1,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8362,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7605,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73f665316e96d5258c89/record.json b/records/sd1/sweep/73f665316e96d5258c89/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9d903290244a39c91db9042649d8b4a8266f59c
--- /dev/null
+++ b/records/sd1/sweep/73f665316e96d5258c89/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "73f665316e96d5258c89",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "ff",
+   "module_type": "Linear",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/529cafd329e9b1ba6db8617b5ccf30d59d8624a11bd337c09b1c25d8e60382e6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "529cafd329e9b1ba6db8617b5ccf30d59d8624a11bd337c09b1c25d8e60382e6",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "efcf67ebceafdb56f0c4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73f665316e96d5258c89/workflow.json b/records/sd1/sweep/73f665316e96d5258c89/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..53868e1bc25d2fd7ea043a761438491ad50b24b4
--- /dev/null
+++ b/records/sd1/sweep/73f665316e96d5258c89/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73f665316e96d5258c89)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.ff.net.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73fa5ade84571fbaf1fe/measurements.json b/records/sd1/sweep/73fa5ade84571fbaf1fe/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..05a3994ce57f01488c973f04807092b155354ddc
--- /dev/null
+++ b/records/sd1/sweep/73fa5ade84571fbaf1fe/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "73fa5ade84571fbaf1fe",
+ "values": {
+  "cosine_distance": {
+   "value": 0.45975,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.498131,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.28508,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.100637,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 39.757,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0043,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.229,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9549,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0732,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73fa5ade84571fbaf1fe/record.json b/records/sd1/sweep/73fa5ade84571fbaf1fe/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff692a5f1a14a79bed1aba9bdef1bb2943930b3c
--- /dev/null
+++ b/records/sd1/sweep/73fa5ade84571fbaf1fe/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "73fa5ade84571fbaf1fe",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:25Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/85f5265fe0998ef06badc117958624f5a977d0e034452ca3c9fe31f8d5f27285.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "85f5265fe0998ef06badc117958624f5a977d0e034452ca3c9fe31f8d5f27285",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "67043a66f2f6924b486e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/73fa5ade84571fbaf1fe/workflow.json b/records/sd1/sweep/73fa5ade84571fbaf1fe/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a99b226605bac566aef5f51997864a6877e4982
--- /dev/null
+++ b/records/sd1/sweep/73fa5ade84571fbaf1fe/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 73fa5ade84571fbaf1fe)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7404cf5a30eedab1f384/measurements.json b/records/sd1/sweep/7404cf5a30eedab1f384/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3db8966c1d49d6510e728eef8989639096f979cc
--- /dev/null
+++ b/records/sd1/sweep/7404cf5a30eedab1f384/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7404cf5a30eedab1f384",
+ "values": {
+  "cosine_distance": {
+   "value": 0.313056,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.507907,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.2417,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.274895,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.297,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00916,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.627,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9699,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2694,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7404cf5a30eedab1f384/record.json b/records/sd1/sweep/7404cf5a30eedab1f384/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..16a1f2fe01149c3cc9a613da2d257b784f4e4c1f
--- /dev/null
+++ b/records/sd1/sweep/7404cf5a30eedab1f384/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7404cf5a30eedab1f384",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:42Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/284df9eec07ebead84a9987a40a02705e3e7c895043cbd6d0688ad1c8c71a660.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "284df9eec07ebead84a9987a40a02705e3e7c895043cbd6d0688ad1c8c71a660",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "b2c44d0b2b9f6e574a0f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7404cf5a30eedab1f384/workflow.json b/records/sd1/sweep/7404cf5a30eedab1f384/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..368e5c4b99ca8cb13242f19173bda8215bd83b3d
--- /dev/null
+++ b/records/sd1/sweep/7404cf5a30eedab1f384/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7404cf5a30eedab1f384)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74057ca6a5c0195915a7/measurements.json b/records/sd1/sweep/74057ca6a5c0195915a7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..43b11ea8abfff4f58116bf9d5d170ba6ee6568cf
--- /dev/null
+++ b/records/sd1/sweep/74057ca6a5c0195915a7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74057ca6a5c0195915a7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.109001,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.184823,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.321588,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.048075,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 76.019,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01186,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.053,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9874,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3278,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74057ca6a5c0195915a7/record.json b/records/sd1/sweep/74057ca6a5c0195915a7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef025d2504973f8d841751056af13b99167dcb77
--- /dev/null
+++ b/records/sd1/sweep/74057ca6a5c0195915a7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74057ca6a5c0195915a7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/23a49f9e97c97f91ad042d1bd9f96228dfd7093524de817b509127ebf0d5e20c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "23a49f9e97c97f91ad042d1bd9f96228dfd7093524de817b509127ebf0d5e20c",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "195405073af577522bea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74057ca6a5c0195915a7/workflow.json b/records/sd1/sweep/74057ca6a5c0195915a7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..407404b4966ceb3205cad42fc9f90bf31142bd9a
--- /dev/null
+++ b/records/sd1/sweep/74057ca6a5c0195915a7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74057ca6a5c0195915a7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7405efd905d50c99d67e/measurements.json b/records/sd1/sweep/7405efd905d50c99d67e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4758ae49235241b3761c810a50d70badbcec36d3
--- /dev/null
+++ b/records/sd1/sweep/7405efd905d50c99d67e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7405efd905d50c99d67e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.326951,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.431343,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.37453,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.179036,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.564,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00264,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.75,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0363,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1854,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7405efd905d50c99d67e/record.json b/records/sd1/sweep/7405efd905d50c99d67e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f88c057c4a392275b6a7cdbde1e2119205e8b258
--- /dev/null
+++ b/records/sd1/sweep/7405efd905d50c99d67e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7405efd905d50c99d67e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0.in_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/185e0f88dce900a9472ca5566d2d683efa9ab5849fd55051d3668dcfcb89f7c8.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "185e0f88dce900a9472ca5566d2d683efa9ab5849fd55051d3668dcfcb89f7c8",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "174468c2d83947a9e51c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7405efd905d50c99d67e/workflow.json b/records/sd1/sweep/7405efd905d50c99d67e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..26745cb904fe1648e1c0e375c84988e4b619e9a5
--- /dev/null
+++ b/records/sd1/sweep/7405efd905d50c99d67e/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7405efd905d50c99d67e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.in_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/740a6f29135d8bb09549/measurements.json b/records/sd1/sweep/740a6f29135d8bb09549/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ca407494123510ef22527b42aeb0b2f0645f367
--- /dev/null
+++ b/records/sd1/sweep/740a6f29135d8bb09549/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "740a6f29135d8bb09549",
+ "values": {
+  "cosine_distance": {
+   "value": 0.185682,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.227908,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.097551,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.042256,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.119,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01106,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.417,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0182,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0689,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/740a6f29135d8bb09549/record.json b/records/sd1/sweep/740a6f29135d8bb09549/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..827337400f6d47482da90d590b38bbb9ea2cd66f
--- /dev/null
+++ b/records/sd1/sweep/740a6f29135d8bb09549/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "740a6f29135d8bb09549",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:17Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d3c89e413dce62c3a79d2191531f3b94c04a2a9bb7e63dd0ff7417deb45f6cd8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d3c89e413dce62c3a79d2191531f3b94c04a2a9bb7e63dd0ff7417deb45f6cd8",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7a8a4c08c8dbd75e05ac"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/740a6f29135d8bb09549/workflow.json b/records/sd1/sweep/740a6f29135d8bb09549/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..98f04c698ecb82048114aa523faae27b3d7becf1
--- /dev/null
+++ b/records/sd1/sweep/740a6f29135d8bb09549/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 740a6f29135d8bb09549)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/741bd67fb33512f04634/measurements.json b/records/sd1/sweep/741bd67fb33512f04634/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..745a78615e2c032328ef46c6b9aa8d418f33129b
--- /dev/null
+++ b/records/sd1/sweep/741bd67fb33512f04634/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "741bd67fb33512f04634",
+ "values": {
+  "std": {
+   "value": 26.59,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00401,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.577,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.548287,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.808534,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.971076,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.461117,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6442,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4147,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/741bd67fb33512f04634/record.json b/records/sd1/sweep/741bd67fb33512f04634/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f227108cef692c6c4ecef62f99f3300468ee3b2
--- /dev/null
+++ b/records/sd1/sweep/741bd67fb33512f04634/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "741bd67fb33512f04634",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:58:41Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "57b42f19d2f0d223bf23"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/741bd67fb33512f04634/workflow.json b/records/sd1/sweep/741bd67fb33512f04634/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..51f3780172f7b5b4d12c1bbd4c86a48fff946043
--- /dev/null
+++ b/records/sd1/sweep/741bd67fb33512f04634/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 741bd67fb33512f04634)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/742256ab1f88cf4609b5/interpretations.jsonl b/records/sd1/sweep/742256ab1f88cf4609b5/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e08bf21eab6abe9a89cb7c154ec2a68749ededc2
--- /dev/null
+++ b/records/sd1/sweep/742256ab1f88cf4609b5/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "742256ab1f88cf4609b5"}, "kind": "caption", "value": "Old man with white beard and broad wool hat in front of a grey sea", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2e6eebcbeb862990"}
+{"target": {"record": "742256ab1f88cf4609b5"}, "kind": "change", "value": "Almost no visible change apart from minor differences in the face and hat texture.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "77f4ed3219be5ef8"}
+{"target": {"record": "742256ab1f88cf4609b5"}, "kind": "keywords", "value": ["old man portrait", "wool hat", "unchanged"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "09b7c8f572a54580"}
+{"target": {"record": "742256ab1f88cf4609b5"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ceeadedef5633785"}
diff --git a/records/sd1/sweep/742256ab1f88cf4609b5/measurements.json b/records/sd1/sweep/742256ab1f88cf4609b5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..12b9fc8391abafd3ce0009070518bc6d1e13053e
--- /dev/null
+++ b/records/sd1/sweep/742256ab1f88cf4609b5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "742256ab1f88cf4609b5",
+ "values": {
+  "std": {
+   "value": 58.822,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01009,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.33,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.005785,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.011688,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.003121,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002237,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0063,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2899,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/742256ab1f88cf4609b5/record.json b/records/sd1/sweep/742256ab1f88cf4609b5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..29a94f29f77965652acc01b7e60bbbf6c86d52e6
--- /dev/null
+++ b/records/sd1/sweep/742256ab1f88cf4609b5/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "742256ab1f88cf4609b5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:40:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "grow",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "scale",
+   1.4,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "886c5aab294a893a4988"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/742256ab1f88cf4609b5/workflow.json b/records/sd1/sweep/742256ab1f88cf4609b5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..704789e62de0fc812b6be3d9dfd4fc4ca1f040b1
--- /dev/null
+++ b/records/sd1/sweep/742256ab1f88cf4609b5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 742256ab1f88cf4609b5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7424872fb64798275128/measurements.json b/records/sd1/sweep/7424872fb64798275128/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bad8fe86a38eaf22f31bbcf1a72b41e477e710a0
--- /dev/null
+++ b/records/sd1/sweep/7424872fb64798275128/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7424872fb64798275128",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7424872fb64798275128/record.json b/records/sd1/sweep/7424872fb64798275128/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb10fcd7aacb9952f9fa7e1231f4f770f9544371
--- /dev/null
+++ b/records/sd1/sweep/7424872fb64798275128/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7424872fb64798275128",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/32466eb2eebb49a2a71a9a10caf253a7e1fc30e5a6aea6cdc8e70bfed7401999.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "32466eb2eebb49a2a71a9a10caf253a7e1fc30e5a6aea6cdc8e70bfed7401999",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c2d386cea60f9ec8acf6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7424872fb64798275128/workflow.json b/records/sd1/sweep/7424872fb64798275128/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..df1d4169cf499ed4412c349b9a14e47a45b9cca3
--- /dev/null
+++ b/records/sd1/sweep/7424872fb64798275128/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7424872fb64798275128)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.h_upd\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/742d79019e57b2ffc91e/measurements.json b/records/sd1/sweep/742d79019e57b2ffc91e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f32b6d210a75d1e44230d49ef47dcd0caaf51a60
--- /dev/null
+++ b/records/sd1/sweep/742d79019e57b2ffc91e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "742d79019e57b2ffc91e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.144513,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.321937,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.230383,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.117223,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.235,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00353,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.277,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0301,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0375,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/742d79019e57b2ffc91e/record.json b/records/sd1/sweep/742d79019e57b2ffc91e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ec01681203f6711c3a9f7d348a0cb3ad116f3e6
--- /dev/null
+++ b/records/sd1/sweep/742d79019e57b2ffc91e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "742d79019e57b2ffc91e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/692d05803b548d28864f02f6040c3e64463bc8236a589ec48bc2a0482a50abed.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "692d05803b548d28864f02f6040c3e64463bc8236a589ec48bc2a0482a50abed",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "b8131173560a849e350e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/742d79019e57b2ffc91e/workflow.json b/records/sd1/sweep/742d79019e57b2ffc91e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f14a9fdd0af4f2802ce535c186df7ebe926169c
--- /dev/null
+++ b/records/sd1/sweep/742d79019e57b2ffc91e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 742d79019e57b2ffc91e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74307a4d6b60bb5daff9/measurements.json b/records/sd1/sweep/74307a4d6b60bb5daff9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f85830445a186386564e2187c39c6de7fce851c
--- /dev/null
+++ b/records/sd1/sweep/74307a4d6b60bb5daff9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74307a4d6b60bb5daff9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.764751,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.808541,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.935011,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.42794,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 89.651,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.04159,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 94.696,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9191,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9031,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74307a4d6b60bb5daff9/record.json b/records/sd1/sweep/74307a4d6b60bb5daff9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c49245de114e9cc2a668dd70cac96ce391a0cf9d
--- /dev/null
+++ b/records/sd1/sweep/74307a4d6b60bb5daff9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74307a4d6b60bb5daff9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:25Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "time_embed.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "embed",
+   "kind": "time_embed",
+   "module_type": "Linear",
+   "container": "Sequential",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ba3337432224d176fc3f4f6f8322d2eec5d8cd4d47cd411ef8dfcf4f848d2187.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ba3337432224d176fc3f4f6f8322d2eec5d8cd4d47cd411ef8dfcf4f848d2187",
+  "type_path": "UNetModel.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ff02fa4fc988ab31792e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74307a4d6b60bb5daff9/workflow.json b/records/sd1/sweep/74307a4d6b60bb5daff9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..79e16081b33a9ae1932b56128c2a17eed2466b74
--- /dev/null
+++ b/records/sd1/sweep/74307a4d6b60bb5daff9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74307a4d6b60bb5daff9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"time_embed.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7439bde027b9df6fa082/measurements.json b/records/sd1/sweep/7439bde027b9df6fa082/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ec2d07339f4a59f9fa00bdc56a8bc335b8a4214
--- /dev/null
+++ b/records/sd1/sweep/7439bde027b9df6fa082/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7439bde027b9df6fa082",
+ "values": {
+  "cosine_distance": {
+   "value": 0.268651,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.310386,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.116919,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044746,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.507,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01653,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 35.736,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9346,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1169,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7439bde027b9df6fa082/record.json b/records/sd1/sweep/7439bde027b9df6fa082/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5153a38523d6088b30fcc9bd85f053cc91abba85
--- /dev/null
+++ b/records/sd1/sweep/7439bde027b9df6fa082/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7439bde027b9df6fa082",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.hi",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ec20d84c2c95236386264736781e52a5a778ad232a1565283ed7ccf678b37745.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ec20d84c2c95236386264736781e52a5a778ad232a1565283ed7ccf678b37745",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "95f2859788b31b885202"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7439bde027b9df6fa082/workflow.json b/records/sd1/sweep/7439bde027b9df6fa082/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..741b17b335a3a4e137e84cc26a976d7b03a992de
--- /dev/null
+++ b/records/sd1/sweep/7439bde027b9df6fa082/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7439bde027b9df6fa082)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.transformer_blocks.0.norm3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743a0f1b7e59b738e41a/interpretations.jsonl b/records/sd1/sweep/743a0f1b7e59b738e41a/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..85c1da54c58960fa6e3dba6d75cbb21b6c2b6919
--- /dev/null
+++ b/records/sd1/sweep/743a0f1b7e59b738e41a/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "743a0f1b7e59b738e41a"}, "kind": "caption", "value": "Painted portrait of a woman with wavy brown hair on a light grey brushy background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3ede74d143890915"}
+{"target": {"record": "743a0f1b7e59b738e41a"}, "kind": "change", "value": "A subtle restyle: the background turns a lighter grey, the hair becomes wavier and darker, the face is slightly altered, and the paper border is removed.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6f001671c0b2351f"}
+{"target": {"record": "743a0f1b7e59b738e41a"}, "kind": "keywords", "value": ["portrait", "grey background", "wavy hair", "subtle variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "85a9ca43f9c02f32"}
+{"target": {"record": "743a0f1b7e59b738e41a"}, "kind": "effect_tags", "value": ["subtle", "subject_kept", "bright"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "179f4397b971e6ad"}
diff --git a/records/sd1/sweep/743a0f1b7e59b738e41a/measurements.json b/records/sd1/sweep/743a0f1b7e59b738e41a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee2a8cff0a89c8b36cc45569df9faa6b1f615d4e
--- /dev/null
+++ b/records/sd1/sweep/743a0f1b7e59b738e41a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "743a0f1b7e59b738e41a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.221291,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.296295,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.135236,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.081657,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.123,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00565,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.186,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9524,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0578,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743a0f1b7e59b738e41a/record.json b/records/sd1/sweep/743a0f1b7e59b738e41a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..201813b2a9ce46802457876a5933c48a442e063c
--- /dev/null
+++ b/records/sd1/sweep/743a0f1b7e59b738e41a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "743a0f1b7e59b738e41a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.hi",
+   "kind": "self_attn",
+   "module_type": "Dropout",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1748f07c1c0854165366aa1247231ac2089e19e22dc99526685bdf4cf007b478.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1748f07c1c0854165366aa1247231ac2089e19e22dc99526685bdf4cf007b478",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "23efebddc89a91f4a57a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743a0f1b7e59b738e41a/workflow.json b/records/sd1/sweep/743a0f1b7e59b738e41a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f35ee83f217031a3af5ae57fad165c37bb47f3ad
--- /dev/null
+++ b/records/sd1/sweep/743a0f1b7e59b738e41a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 743a0f1b7e59b738e41a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.transformer_blocks.0.attn1.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743ab557aabd5946d746/interpretations.jsonl b/records/sd1/sweep/743ab557aabd5946d746/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..12aa598cff798661482d39fed9b57a966fcf3c27
--- /dev/null
+++ b/records/sd1/sweep/743ab557aabd5946d746/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "743ab557aabd5946d746"}, "kind": "caption", "value": "Soft sepia-toned portrait of a woman with brown hair on dark brown background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "29b078bf4c26eb43"}
+{"target": {"record": "743ab557aabd5946d746"}, "kind": "change", "value": "The image softens and warms into sepia browns, with lower contrast, slightly blurred hair and a muted brown background in place of the cool grey.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9909af4b08108e57"}
+{"target": {"record": "743ab557aabd5946d746"}, "kind": "keywords", "value": ["sepia", "soft focus", "warm brown", "vintage portrait", "muted"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "54c7053bc4260559"}
+{"target": {"record": "743ab557aabd5946d746"}, "kind": "effect_tags", "value": ["colour_shift", "blur", "washed_out", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "807e7b818d5b3bca"}
diff --git a/records/sd1/sweep/743ab557aabd5946d746/measurements.json b/records/sd1/sweep/743ab557aabd5946d746/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b6369c6be2c6a4cac35a5da3d617b6ea509e6c0
--- /dev/null
+++ b/records/sd1/sweep/743ab557aabd5946d746/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "743ab557aabd5946d746",
+ "values": {
+  "cosine_distance": {
+   "value": 0.2206,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.333058,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.074607,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.037033,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.516,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00327,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 35.47,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9819,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2306,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743ab557aabd5946d746/record.json b/records/sd1/sweep/743ab557aabd5946d746/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eeb55c15273c83b9a267222482b7d104fa2ae8df
--- /dev/null
+++ b/records/sd1/sweep/743ab557aabd5946d746/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "743ab557aabd5946d746",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/b2ea1ba5fe0468ff160f989f228f47b04646a712c086aa4cebc69131e81dccec.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b2ea1ba5fe0468ff160f989f228f47b04646a712c086aa4cebc69131e81dccec",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "98fa45991fb6a1362e66"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743ab557aabd5946d746/workflow.json b/records/sd1/sweep/743ab557aabd5946d746/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e25bae51f683ff6c113fc337397afab2b63279a1
--- /dev/null
+++ b/records/sd1/sweep/743ab557aabd5946d746/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 743ab557aabd5946d746)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743ed4d7632da9d3caa8/measurements.json b/records/sd1/sweep/743ed4d7632da9d3caa8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a45b30447099d0daee038db39a466a87d7dab65
--- /dev/null
+++ b/records/sd1/sweep/743ed4d7632da9d3caa8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "743ed4d7632da9d3caa8",
+ "values": {
+  "std": {
+   "value": 39.552,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00659,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.782,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.137388,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.281379,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.262298,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.075441,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0341,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0102,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743ed4d7632da9d3caa8/record.json b/records/sd1/sweep/743ed4d7632da9d3caa8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b9e0b249f225c04d539317795df4654f8cfe461
--- /dev/null
+++ b/records/sd1/sweep/743ed4d7632da9d3caa8/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "743ed4d7632da9d3caa8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:45:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "damp",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   0.5,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/743ed4d7632da9d3caa8/workflow.json b/records/sd1/sweep/743ed4d7632da9d3caa8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1163f26710577887b00ea1a3c41010f5ce610b08
--- /dev/null
+++ b/records/sd1/sweep/743ed4d7632da9d3caa8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 743ed4d7632da9d3caa8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/745eac3fb155dab9cb8d/measurements.json b/records/sd1/sweep/745eac3fb155dab9cb8d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c09181454d1a9d8874243f7a40e5b04237207faa
--- /dev/null
+++ b/records/sd1/sweep/745eac3fb155dab9cb8d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "745eac3fb155dab9cb8d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.147144,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.195186,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.10219,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.083232,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.15,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00485,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.355,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9683,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3805,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/745eac3fb155dab9cb8d/record.json b/records/sd1/sweep/745eac3fb155dab9cb8d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae7aaa57c8874e1a12fddd3b85581912b5945e1a
--- /dev/null
+++ b/records/sd1/sweep/745eac3fb155dab9cb8d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "745eac3fb155dab9cb8d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b77f8ab6a86fa6c14a8ef643f2bd8040cdaecbe7230400a9f70f243c0dac631e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b77f8ab6a86fa6c14a8ef643f2bd8040cdaecbe7230400a9f70f243c0dac631e",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "f55a5ad72a00e5bc09af"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/745eac3fb155dab9cb8d/workflow.json b/records/sd1/sweep/745eac3fb155dab9cb8d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a91dfebcf466e8505d69783f6f4f3e3a71bf21bb
--- /dev/null
+++ b/records/sd1/sweep/745eac3fb155dab9cb8d/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 745eac3fb155dab9cb8d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/746314b8a88d993ee25c/measurements.json b/records/sd1/sweep/746314b8a88d993ee25c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..65e7dbf768619e52e447130de785ed8c7f69a95b
--- /dev/null
+++ b/records/sd1/sweep/746314b8a88d993ee25c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "746314b8a88d993ee25c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.572197,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.722774,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.916875,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.414328,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 74.049,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0034,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 66.369,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8546,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5478,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/746314b8a88d993ee25c/record.json b/records/sd1/sweep/746314b8a88d993ee25c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b783f2be2820342b4e577ee58f69c42be6362f81
--- /dev/null
+++ b/records/sd1/sweep/746314b8a88d993ee25c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "746314b8a88d993ee25c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:19Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/6679a2996df712af7983c175c547defe7476d7fd2607542ea37ce097f0af2620.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6679a2996df712af7983c175c547defe7476d7fd2607542ea37ce097f0af2620",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "4521f64804213c309437"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/746314b8a88d993ee25c/workflow.json b/records/sd1/sweep/746314b8a88d993ee25c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..05b4cb71258610bdd2e4f744499a30bf12253d11
--- /dev/null
+++ b/records/sd1/sweep/746314b8a88d993ee25c/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 746314b8a88d993ee25c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.skip_connection\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/746392216076de39263b/measurements.json b/records/sd1/sweep/746392216076de39263b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..525759cf266add553c5ad610578ed6a1e3eee33d
--- /dev/null
+++ b/records/sd1/sweep/746392216076de39263b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "746392216076de39263b",
+ "values": {
+  "std": {
+   "value": 59.675,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02351,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 62.726,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.562747,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.633109,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.225276,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.104432,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9725,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2568,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/746392216076de39263b/record.json b/records/sd1/sweep/746392216076de39263b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7483ed0318e4f60f1646bf6b5d1080a9057a081b
--- /dev/null
+++ b/records/sd1/sweep/746392216076de39263b/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "746392216076de39263b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:09:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "zero",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "bf7b38adc35636cc6881"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/746392216076de39263b/workflow.json b/records/sd1/sweep/746392216076de39263b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6795a87c06ea5b80129bf6d6d3660f2fa6d3c180
--- /dev/null
+++ b/records/sd1/sweep/746392216076de39263b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 746392216076de39263b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7469a389c9690bb236c8/measurements.json b/records/sd1/sweep/7469a389c9690bb236c8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..455d85a2986924dfa8123c6202635f2f370c423e
--- /dev/null
+++ b/records/sd1/sweep/7469a389c9690bb236c8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7469a389c9690bb236c8",
+ "values": {
+  "std": {
+   "value": 44.435,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01699,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.939,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.2879,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.390547,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.08291,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.021495,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0079,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0184,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7469a389c9690bb236c8/record.json b/records/sd1/sweep/7469a389c9690bb236c8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbef004806fad1467384136dae14e0c4ecbc03b7
--- /dev/null
+++ b/records/sd1/sweep/7469a389c9690bb236c8/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "7469a389c9690bb236c8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:57:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "grow",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "scale",
+   1.4,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.7, 1.3] for out.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0519c296ea8083d5b483"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7469a389c9690bb236c8/workflow.json b/records/sd1/sweep/7469a389c9690bb236c8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c39fe991e95d325022dd235b7c67854e2b04ae66
--- /dev/null
+++ b/records/sd1/sweep/7469a389c9690bb236c8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7469a389c9690bb236c8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7475786922888486cb27/interpretations.jsonl b/records/sd1/sweep/7475786922888486cb27/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d4e07e890126ef3611302f94c968f88141a22393
--- /dev/null
+++ b/records/sd1/sweep/7475786922888486cb27/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7475786922888486cb27"}, "kind": "caption", "value": "Painted woman with tousled dark brown bob and blue eyes on a pale blue-grey background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7824226c3b6e9e81"}
+{"target": {"record": "7475786922888486cb27"}, "kind": "change", "value": "The hair turns into a darker, messy, loosely painted bob, the eyes become vivid blue, and the background shifts to a light blue-grey.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b439c5bb49640b0c"}
+{"target": {"record": "7475786922888486cb27"}, "kind": "keywords", "value": ["messy bob", "blue eyes", "loose brushwork", "light blue background", "dark hair"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cc8a0d90a8fa7f36"}
+{"target": {"record": "7475786922888486cb27"}, "kind": "effect_tags", "value": ["subject_kept", "painterly", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d5823280bb98c8c9"}
diff --git a/records/sd1/sweep/7475786922888486cb27/measurements.json b/records/sd1/sweep/7475786922888486cb27/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..301009c7d8e537f98c2d5f37b004523bffbcbd78
--- /dev/null
+++ b/records/sd1/sweep/7475786922888486cb27/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7475786922888486cb27",
+ "values": {
+  "cosine_distance": {
+   "value": 0.302635,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.359848,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.235109,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.077855,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.274,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0051,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 35.886,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9937,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0552,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7475786922888486cb27/record.json b/records/sd1/sweep/7475786922888486cb27/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6585caeb14c2e1eaa9e6ab342e3e41a673e04348
--- /dev/null
+++ b/records/sd1/sweep/7475786922888486cb27/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7475786922888486cb27",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.mid",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/aa0a62219208eea95f2bd21538d6cc402a5ad93b44802ca4c149c6bd9bb760b5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "aa0a62219208eea95f2bd21538d6cc402a5ad93b44802ca4c149c6bd9bb760b5",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "d1dd80297458a6b8825b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7475786922888486cb27/workflow.json b/records/sd1/sweep/7475786922888486cb27/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2718945bf9bfcdf7eec5362f72889e12aae27537
--- /dev/null
+++ b/records/sd1/sweep/7475786922888486cb27/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7475786922888486cb27)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7478ba810aab5674788b/measurements.json b/records/sd1/sweep/7478ba810aab5674788b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e448667c66969462706b3376a4a979b00cb55d66
--- /dev/null
+++ b/records/sd1/sweep/7478ba810aab5674788b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7478ba810aab5674788b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.489214,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.464951,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.816974,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.389925,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.47,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0105,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 53.623,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8189,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0588,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7478ba810aab5674788b/record.json b/records/sd1/sweep/7478ba810aab5674788b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a506a2ea341e1738cf6ee6e5a921595ee0fb94ab
--- /dev/null
+++ b/records/sd1/sweep/7478ba810aab5674788b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7478ba810aab5674788b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:51Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/96f0ae5f670a1a6729d3528d1acc5bfd98b49de32a22796bf677401511f41b0f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "96f0ae5f670a1a6729d3528d1acc5bfd98b49de32a22796bf677401511f41b0f",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "a76ae46ef8331e450b1d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7478ba810aab5674788b/workflow.json b/records/sd1/sweep/7478ba810aab5674788b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..adb8e37dc8fbcad1add32b1670d3ee09c160ff02
--- /dev/null
+++ b/records/sd1/sweep/7478ba810aab5674788b/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7478ba810aab5674788b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/747ee8001296819322d9/measurements.json b/records/sd1/sweep/747ee8001296819322d9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4f47f3f5b11d6b229096527bd725888e48148e1
--- /dev/null
+++ b/records/sd1/sweep/747ee8001296819322d9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "747ee8001296819322d9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.465884,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.452182,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.170685,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.16455,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.105,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0054,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.849,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0235,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2911,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/747ee8001296819322d9/record.json b/records/sd1/sweep/747ee8001296819322d9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2270fc64f678112fe266cbd7cf886f63aaa6bc4
--- /dev/null
+++ b/records/sd1/sweep/747ee8001296819322d9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "747ee8001296819322d9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/fecebf884799e3d1b5cc907de17a4fbc74e7e65f43b3facc174daaabddc50def.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "fecebf884799e3d1b5cc907de17a4fbc74e7e65f43b3facc174daaabddc50def",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "8da302cfa499f738134b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/747ee8001296819322d9/workflow.json b/records/sd1/sweep/747ee8001296819322d9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..669d4d11fe3b20fe83d4b4093d3464b7281c3c76
--- /dev/null
+++ b/records/sd1/sweep/747ee8001296819322d9/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 747ee8001296819322d9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.out_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74827528ffb45c07215f/measurements.json b/records/sd1/sweep/74827528ffb45c07215f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ab37083fdd00c8d803e35264d3a5e4c245596a0
--- /dev/null
+++ b/records/sd1/sweep/74827528ffb45c07215f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74827528ffb45c07215f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.515742,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.515523,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.385976,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.15972,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.722,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0056,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 50.519,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0236,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2558,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74827528ffb45c07215f/record.json b/records/sd1/sweep/74827528ffb45c07215f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5986ba69586c119994a612d04864fc7e5de9a639
--- /dev/null
+++ b/records/sd1/sweep/74827528ffb45c07215f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "74827528ffb45c07215f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/0f753bba052fe4f3a991cc8f066d16d0c1ca75ce54c77afd8ead635cae0c840d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0f753bba052fe4f3a991cc8f066d16d0c1ca75ce54c77afd8ead635cae0c840d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "2d024fad72521874028a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74827528ffb45c07215f/workflow.json b/records/sd1/sweep/74827528ffb45c07215f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4bb2f508aecd961e89cce4cc844f17f24c3504f
--- /dev/null
+++ b/records/sd1/sweep/74827528ffb45c07215f/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74827528ffb45c07215f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.proj_in\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74859eeb59b96029b94b/measurements.json b/records/sd1/sweep/74859eeb59b96029b94b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c43f2d254542124a5f615449e428cf32987b3976
--- /dev/null
+++ b/records/sd1/sweep/74859eeb59b96029b94b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74859eeb59b96029b94b",
+ "values": {
+  "std": {
+   "value": 54.618,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01371,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.592,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.202873,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.32115,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.391128,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.116867,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0404,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.02,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74859eeb59b96029b94b/record.json b/records/sd1/sweep/74859eeb59b96029b94b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e664da72552effa87257aaf09afb37bb2b6c1bd
--- /dev/null
+++ b/records/sd1/sweep/74859eeb59b96029b94b/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "74859eeb59b96029b94b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:01:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "heavy",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_noise",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74859eeb59b96029b94b/workflow.json b/records/sd1/sweep/74859eeb59b96029b94b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..068b33919b2408fbc03287f0200af23f15483cf3
--- /dev/null
+++ b/records/sd1/sweep/74859eeb59b96029b94b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74859eeb59b96029b94b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7499dae1a6d73fcb31bb/measurements.json b/records/sd1/sweep/7499dae1a6d73fcb31bb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..855adc3d71b8e6d99b50f92df22d0d92bf3612a5
--- /dev/null
+++ b/records/sd1/sweep/7499dae1a6d73fcb31bb/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7499dae1a6d73fcb31bb",
+ "values": {
+  "std": {
+   "value": 94.542,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0019,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 83.541,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.811985,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.825397,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.714648,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.352236,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.867,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8184,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7499dae1a6d73fcb31bb/record.json b/records/sd1/sweep/7499dae1a6d73fcb31bb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1b3960ddc103b3dbfaf4960f6f968bc8022895b
--- /dev/null
+++ b/records/sd1/sweep/7499dae1a6d73fcb31bb/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "7499dae1a6d73fcb31bb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:30:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "slight",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   30.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 30.0,
+    "applied": 30.0,
+    "reason": "outside safe range [0.0, 5.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "b5fb2fe34c8473a1bc70"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7499dae1a6d73fcb31bb/workflow.json b/records/sd1/sweep/7499dae1a6d73fcb31bb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..40cbe37456792b1b51019652c19b0511246627b4
--- /dev/null
+++ b/records/sd1/sweep/7499dae1a6d73fcb31bb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7499dae1a6d73fcb31bb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74a089484cb3e4b8fd7b/interpretations.jsonl b/records/sd1/sweep/74a089484cb3e4b8fd7b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e776ef5eaae33f4224ef5b6732f83ce23b105fe2
--- /dev/null
+++ b/records/sd1/sweep/74a089484cb3e4b8fd7b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "74a089484cb3e4b8fd7b"}, "kind": "caption", "value": "Blurry, smoothed woman with centre-parted dark hair and green sweater on flat mustard-yellow background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "65022f945eab251a"}
+{"target": {"record": "74a089484cb3e4b8fd7b"}, "kind": "change", "value": "The crisp grey portrait melts into a soft, airbrushed, symmetric frontal face with a sickly yellow-olive palette and an empty flat background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4542f33046b637ac"}
+{"target": {"record": "74a089484cb3e4b8fd7b"}, "kind": "keywords", "value": ["blurred portrait", "mustard yellow", "symmetric face", "airbrushed", "sepia tone"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f3ef8246e68d7362"}
+{"target": {"record": "74a089484cb3e4b8fd7b"}, "kind": "effect_tags", "value": ["blur", "colour_shift", "symmetry", "washed_out", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "647a02f920f3036b"}
diff --git a/records/sd1/sweep/74a089484cb3e4b8fd7b/measurements.json b/records/sd1/sweep/74a089484cb3e4b8fd7b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b0db30443be27e707e24a14f1ef7882207bcd27
--- /dev/null
+++ b/records/sd1/sweep/74a089484cb3e4b8fd7b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74a089484cb3e4b8fd7b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.681257,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.614714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.494978,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.215388,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.626,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00134,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 59.263,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.913,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5899,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74a089484cb3e4b8fd7b/record.json b/records/sd1/sweep/74a089484cb3e4b8fd7b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4da305d02909bdb08da45efadfe424225349859c
--- /dev/null
+++ b/records/sd1/sweep/74a089484cb3e4b8fd7b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "74a089484cb3e4b8fd7b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.out_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5b821f8e1a3d0a24902876595b674c3735034b4398570dc86ff305801a52a92d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5b821f8e1a3d0a24902876595b674c3735034b4398570dc86ff305801a52a92d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "58dd904f30d389e1fb90"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74a089484cb3e4b8fd7b/workflow.json b/records/sd1/sweep/74a089484cb3e4b8fd7b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b72f346190126476672f08b0be4250743c9b8768
--- /dev/null
+++ b/records/sd1/sweep/74a089484cb3e4b8fd7b/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74a089484cb3e4b8fd7b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74a8e04cfaa2f2e91e0a/measurements.json b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d85ac3a5a13009f5df820886edcdca67e7df1c9
--- /dev/null
+++ b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74a8e04cfaa2f2e91e0a",
+ "values": {
+  "std": {
+   "value": 51.025,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01285,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 8.51,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.040756,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.107584,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131578,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.037863,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0658,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0646,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74a8e04cfaa2f2e91e0a/record.json b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..026f8ce5a6adccd8c7a887726ce5946440861ce4
--- /dev/null
+++ b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "74a8e04cfaa2f2e91e0a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:01:38Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "mid",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_scalar",
+   0.8,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "cb81730797eb0334f04d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74a8e04cfaa2f2e91e0a/workflow.json b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..02d448eab9a818dd03805b19370693542450054e
--- /dev/null
+++ b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74a8e04cfaa2f2e91e0a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74aa245972c75dabe395/measurements.json b/records/sd1/sweep/74aa245972c75dabe395/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc21a758100f8be6d471da187e1e2c604f853d1d
--- /dev/null
+++ b/records/sd1/sweep/74aa245972c75dabe395/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74aa245972c75dabe395",
+ "values": {
+  "cosine_distance": {
+   "value": 0.018622,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.02876,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.009676,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004529,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.148,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00546,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.309,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0113,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1895,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74aa245972c75dabe395/record.json b/records/sd1/sweep/74aa245972c75dabe395/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..253d43d460409b420fb8f5a8cfb77e8bd78467be
--- /dev/null
+++ b/records/sd1/sweep/74aa245972c75dabe395/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74aa245972c75dabe395",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:38Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ef662ce2439f73a278d8c98c4caf1d2e8374e0e039ffb62482cd63043bee6db7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ef662ce2439f73a278d8c98c4caf1d2e8374e0e039ffb62482cd63043bee6db7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f14818aceaa24c1be648"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74aa245972c75dabe395/workflow.json b/records/sd1/sweep/74aa245972c75dabe395/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..14c2e6a4115da1ddbd86fd613ee15d56bce751f8
--- /dev/null
+++ b/records/sd1/sweep/74aa245972c75dabe395/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74aa245972c75dabe395)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74ba99159e0b8f7dd0c7/measurements.json b/records/sd1/sweep/74ba99159e0b8f7dd0c7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfb5e1900ceb59ab8f7eef8f6b3300250058e841
--- /dev/null
+++ b/records/sd1/sweep/74ba99159e0b8f7dd0c7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74ba99159e0b8f7dd0c7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.391532,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.518882,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.217277,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.088667,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 28.966,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0007,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.69,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0368,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2517,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74ba99159e0b8f7dd0c7/record.json b/records/sd1/sweep/74ba99159e0b8f7dd0c7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a2899fa1cc11bff0c120b4961f546fd673a13c08
--- /dev/null
+++ b/records/sd1/sweep/74ba99159e0b8f7dd0c7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74ba99159e0b8f7dd0c7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:25Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "time_embed.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "embed",
+   "kind": "time_embed",
+   "module_type": "SiLU",
+   "container": "Sequential",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/69bb9839745d2cf23ccd762deaecd281292134fb029fbfa8f10007bae5ca4007.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "69bb9839745d2cf23ccd762deaecd281292134fb029fbfa8f10007bae5ca4007",
+  "type_path": "UNetModel.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "bd635447578dc0bf684f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74ba99159e0b8f7dd0c7/workflow.json b/records/sd1/sweep/74ba99159e0b8f7dd0c7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..57c31fa62515a83877e96ac4c5d68924de3553d9
--- /dev/null
+++ b/records/sd1/sweep/74ba99159e0b8f7dd0c7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74ba99159e0b8f7dd0c7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"time_embed.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74bcad6e1bf961dbd754/interpretations.jsonl b/records/sd1/sweep/74bcad6e1bf961dbd754/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..3d4e289b3c1ff76a57def1d12933e9800ac6db72
--- /dev/null
+++ b/records/sd1/sweep/74bcad6e1bf961dbd754/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "74bcad6e1bf961dbd754"}, "kind": "caption", "value": "Waves crashing over a jagged orange-lit rock in a dark blue sea, no lighthouse.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a674977392cae44e"}
+{"target": {"record": "74bcad6e1bf961dbd754"}, "kind": "change", "value": "The lighthouse and sunset vanish, replaced by a churning sea scene with a rough, painterly rock outcrop and cold blue tones.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "62d099e870665cd0"}
+{"target": {"record": "74bcad6e1bf961dbd754"}, "kind": "keywords", "value": ["crashing waves", "rock in sea", "lighthouse removed", "stormy seascape", "painterly"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e7965ae891577275"}
+{"target": {"record": "74bcad6e1bf961dbd754"}, "kind": "effect_tags", "value": ["new_composition", "subject_lost", "painterly", "colour_shift", "texture"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6352f1f712d04493"}
diff --git a/records/sd1/sweep/74bcad6e1bf961dbd754/measurements.json b/records/sd1/sweep/74bcad6e1bf961dbd754/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5652aea7b1c7cb1040b1180c1afc1d27942b75f
--- /dev/null
+++ b/records/sd1/sweep/74bcad6e1bf961dbd754/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74bcad6e1bf961dbd754",
+ "values": {
+  "std": {
+   "value": 52.096,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01881,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 46.55,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.485651,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.629525,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.612715,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.250389,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8212,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2309,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74bcad6e1bf961dbd754/record.json b/records/sd1/sweep/74bcad6e1bf961dbd754/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d83163008aec80ecf4a397d7eb5caf7a85fa8e6
--- /dev/null
+++ b/records/sd1/sweep/74bcad6e1bf961dbd754/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "74bcad6e1bf961dbd754",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:04:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "multiply",
+   0.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for in.lo (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "3376f9d3a4d24355b45f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74bcad6e1bf961dbd754/workflow.json b/records/sd1/sweep/74bcad6e1bf961dbd754/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d77b37c561541e74b5b0240c15148a4d6ed959a
--- /dev/null
+++ b/records/sd1/sweep/74bcad6e1bf961dbd754/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74bcad6e1bf961dbd754)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74bd81d571aa43d22fb6/measurements.json b/records/sd1/sweep/74bd81d571aa43d22fb6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a42d764e7bc8365241dd453807d78878d4f9dad
--- /dev/null
+++ b/records/sd1/sweep/74bd81d571aa43d22fb6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74bd81d571aa43d22fb6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.600153,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.532136,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.353944,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.193893,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.393,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00768,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.374,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0111,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5085,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74bd81d571aa43d22fb6/record.json b/records/sd1/sweep/74bd81d571aa43d22fb6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bc63075c727d6e7b091b3cdb0de0fa8352fd9b8
--- /dev/null
+++ b/records/sd1/sweep/74bd81d571aa43d22fb6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "74bd81d571aa43d22fb6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ad2a2af0bf11c80f553d859a472d9945d859c6e6868f633934d415d3c7a2d9cc.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ad2a2af0bf11c80f553d859a472d9945d859c6e6868f633934d415d3c7a2d9cc",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "7488fe8c65fe8a98edf7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74bd81d571aa43d22fb6/workflow.json b/records/sd1/sweep/74bd81d571aa43d22fb6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..787f3a615183e32b9dd4483343fabe688a321791
--- /dev/null
+++ b/records/sd1/sweep/74bd81d571aa43d22fb6/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74bd81d571aa43d22fb6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74c41ebb67d808d6ff33/measurements.json b/records/sd1/sweep/74c41ebb67d808d6ff33/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2835a0a9817a6216912ea4702debc393b29699af
--- /dev/null
+++ b/records/sd1/sweep/74c41ebb67d808d6ff33/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74c41ebb67d808d6ff33",
+ "values": {
+  "cosine_distance": {
+   "value": 0.278,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.440601,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.144706,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.133807,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.304,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00316,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.637,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9588,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0822,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74c41ebb67d808d6ff33/record.json b/records/sd1/sweep/74c41ebb67d808d6ff33/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..37a19593ff8203bd09bf3f251de7e82accb407dc
--- /dev/null
+++ b/records/sd1/sweep/74c41ebb67d808d6ff33/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74c41ebb67d808d6ff33",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.transformer_blocks.0.ff.net.0.proj",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.mid",
+   "kind": "ff",
+   "module_type": "Linear",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ce3b60fc55623d5e59d5ccbdb30e664f8f5f5d26e6408589c6115cb9b30eba64.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ce3b60fc55623d5e59d5ccbdb30e664f8f5f5d26e6408589c6115cb9b30eba64",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.GEGLU.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "df245390e0dde01a5561"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74c41ebb67d808d6ff33/workflow.json b/records/sd1/sweep/74c41ebb67d808d6ff33/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..72f4a637adf759fcd8c16cba49aba4dd189a50ad
--- /dev/null
+++ b/records/sd1/sweep/74c41ebb67d808d6ff33/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74c41ebb67d808d6ff33)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.ff.net.0.proj\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74c51e1674389d1fac6c/measurements.json b/records/sd1/sweep/74c51e1674389d1fac6c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a788e1a8232230e7768b07337cd50df1674ca641
--- /dev/null
+++ b/records/sd1/sweep/74c51e1674389d1fac6c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74c51e1674389d1fac6c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.360976,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.42061,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.236412,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.165165,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.065,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00384,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.939,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9993,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2606,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74c51e1674389d1fac6c/record.json b/records/sd1/sweep/74c51e1674389d1fac6c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f6ccbb079939cae660d30373d023886a7a7a983
--- /dev/null
+++ b/records/sd1/sweep/74c51e1674389d1fac6c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "74c51e1674389d1fac6c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f37c7dae1d339c179e8ef154379d4d9a51665ebeb46f5e2da84ffcbd2b4f4c99.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f37c7dae1d339c179e8ef154379d4d9a51665ebeb46f5e2da84ffcbd2b4f4c99",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "0732e76a477dbb3a7709"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74c51e1674389d1fac6c/workflow.json b/records/sd1/sweep/74c51e1674389d1fac6c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..63193c247c4bc4b4a696761edecd8cb2b424de81
--- /dev/null
+++ b/records/sd1/sweep/74c51e1674389d1fac6c/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74c51e1674389d1fac6c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74ccc9325dbe264a870a/measurements.json b/records/sd1/sweep/74ccc9325dbe264a870a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fdd933e5a2b858387d576bc6ba4adb0b05e99ab
--- /dev/null
+++ b/records/sd1/sweep/74ccc9325dbe264a870a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74ccc9325dbe264a870a",
+ "values": {
+  "std": {
+   "value": 36.724,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02147,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.499,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.119037,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.214538,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.05884,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.026168,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.03,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0388,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74ccc9325dbe264a870a/record.json b/records/sd1/sweep/74ccc9325dbe264a870a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..26b0b83b66c4e320bc09461170076f337990a144
--- /dev/null
+++ b/records/sd1/sweep/74ccc9325dbe264a870a/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "74ccc9325dbe264a870a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:00:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_scalar",
+   0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e38413cd878adeabf6db"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74ccc9325dbe264a870a/workflow.json b/records/sd1/sweep/74ccc9325dbe264a870a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..30c0f07b2907b06d50f5532815acdcefee145c6c
--- /dev/null
+++ b/records/sd1/sweep/74ccc9325dbe264a870a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74ccc9325dbe264a870a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d233af02933d3913eb/interpretations.jsonl b/records/sd1/sweep/74d233af02933d3913eb/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..1d6fbc9b98524089be64fef10119d14361641970
--- /dev/null
+++ b/records/sd1/sweep/74d233af02933d3913eb/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "74d233af02933d3913eb"}, "kind": "caption", "value": "Coarse, blobby rainbow colour noise covering the frame with no subject", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:25:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f96763f7f4bf4a38"}
+{"target": {"record": "74d233af02933d3913eb"}, "kind": "change", "value": "The portrait collapses into larger-grained saturated colour blobs, with the subject fully gone.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:25:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d856463871b63945"}
+{"target": {"record": "74d233af02933d3913eb"}, "kind": "keywords", "value": ["coarse noise", "colour blobs", "static", "collapse"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:25:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d1664b63a0a9baa6"}
+{"target": {"record": "74d233af02933d3913eb"}, "kind": "effect_tags", "value": ["glitch", "degenerate", "saturated", "subject_lost", "pattern", "abstract"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:25:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7086f66b6bfafc6a"}
diff --git a/records/sd1/sweep/74d233af02933d3913eb/measurements.json b/records/sd1/sweep/74d233af02933d3913eb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e3c9fe328dd1753d990a59f090a5969e8b32b14
--- /dev/null
+++ b/records/sd1/sweep/74d233af02933d3913eb/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74d233af02933d3913eb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.954252,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.848061,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.884581,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.450764,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 88.993,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01999,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "clipped"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 101.537,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.903,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9749,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d233af02933d3913eb/record.json b/records/sd1/sweep/74d233af02933d3913eb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ab62fc0bb611cf29da7a2e80ab5b0bcbf5c2e04
--- /dev/null
+++ b/records/sd1/sweep/74d233af02933d3913eb/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74d233af02933d3913eb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:13Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "out.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.conv",
+   "kind": "out_conv",
+   "module_type": "Conv2d",
+   "container": "Sequential",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/84913a4dfecf323d99775a1d80d73951562a7e46a1c48c79c7d72b4130e30e3c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "84913a4dfecf323d99775a1d80d73951562a7e46a1c48c79c7d72b4130e30e3c",
+  "type_path": "UNetModel.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "98a1af8f9d75701b9bbc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d233af02933d3913eb/workflow.json b/records/sd1/sweep/74d233af02933d3913eb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..df86c84fc0d6e96ed0b50c2dcb0fba9a2523ffb3
--- /dev/null
+++ b/records/sd1/sweep/74d233af02933d3913eb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74d233af02933d3913eb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"out.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d27d2121e2c497c48f/measurements.json b/records/sd1/sweep/74d27d2121e2c497c48f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..65ef6a255e3364584ed161d303afcebc04489cba
--- /dev/null
+++ b/records/sd1/sweep/74d27d2121e2c497c48f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74d27d2121e2c497c48f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.432687,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.586076,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.33272,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.253243,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 32.3,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0034,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.605,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.004,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5077,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d27d2121e2c497c48f/record.json b/records/sd1/sweep/74d27d2121e2c497c48f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1841cd682d066725910a6829bb06270e1c308eed
--- /dev/null
+++ b/records/sd1/sweep/74d27d2121e2c497c48f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74d27d2121e2c497c48f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e5e8e67edc81b26e40633923fe7ab190a57b6afa6ba86857ec3847256dd892f7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e5e8e67edc81b26e40633923fe7ab190a57b6afa6ba86857ec3847256dd892f7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "aa8fb9bf1b8e29bfa942"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d27d2121e2c497c48f/workflow.json b/records/sd1/sweep/74d27d2121e2c497c48f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4409f83edae54c4ee20fc00fa5b61a3b4c08da6
--- /dev/null
+++ b/records/sd1/sweep/74d27d2121e2c497c48f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74d27d2121e2c497c48f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.in_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d34817a345e33aba0f/measurements.json b/records/sd1/sweep/74d34817a345e33aba0f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba398ed1a75c2120fa2d99d97c2f62e432d96f72
--- /dev/null
+++ b/records/sd1/sweep/74d34817a345e33aba0f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74d34817a345e33aba0f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.347749,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.387985,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.173812,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.089005,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.866,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00382,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 39.28,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9495,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.092,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d34817a345e33aba0f/record.json b/records/sd1/sweep/74d34817a345e33aba0f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f0edf62b35b02d603f6dca640a8802a08f14a03
--- /dev/null
+++ b/records/sd1/sweep/74d34817a345e33aba0f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74d34817a345e33aba0f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/77ebc821fb9fc8a248499a43906e45ec4822337f7ca1893f0dbf1a4990deda91.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "77ebc821fb9fc8a248499a43906e45ec4822337f7ca1893f0dbf1a4990deda91",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "14e19adbca2fab89d741"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74d34817a345e33aba0f/workflow.json b/records/sd1/sweep/74d34817a345e33aba0f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f8a42f5c8311b6b472dd2a68c774c62f1fbb2c7
--- /dev/null
+++ b/records/sd1/sweep/74d34817a345e33aba0f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74d34817a345e33aba0f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74db7eaba0d5521af51a/interpretations.jsonl b/records/sd1/sweep/74db7eaba0d5521af51a/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..1bf15dde6b258f2902d8df3b50d9d1d278700882
--- /dev/null
+++ b/records/sd1/sweep/74db7eaba0d5521af51a/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "74db7eaba0d5521af51a"}, "kind": "caption", "value": "Painted portrait of a blue-eyed woman with dark wavy hair before a lighter blue-grey backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "591ad454eefc68f4"}
+{"target": {"record": "74db7eaba0d5521af51a"}, "kind": "change", "value": "The same woman now has blue eyes and softer, more naturally lit hair, and the backdrop turns a lighter blue-grey, giving a cooler, more modern look while the pose stays the same.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "733db2022ebd70db"}
+{"target": {"record": "74db7eaba0d5521af51a"}, "kind": "keywords", "value": ["portrait", "blue eyes", "cool palette", "lighter background", "soft lighting"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "10e302fd0441b1ed"}
+{"target": {"record": "74db7eaba0d5521af51a"}, "kind": "effect_tags", "value": ["colour_shift", "subject_kept", "bright"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6df699283e7e6978"}
diff --git a/records/sd1/sweep/74db7eaba0d5521af51a/measurements.json b/records/sd1/sweep/74db7eaba0d5521af51a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec2810c3174daa38c2a7f788d8994c333a444168
--- /dev/null
+++ b/records/sd1/sweep/74db7eaba0d5521af51a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74db7eaba0d5521af51a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.098701,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.153268,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.048021,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.028707,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.901,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00546,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.543,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9904,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0936,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74db7eaba0d5521af51a/record.json b/records/sd1/sweep/74db7eaba0d5521af51a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d68c5bb895015409909c7a239bbb283ccd81be98
--- /dev/null
+++ b/records/sd1/sweep/74db7eaba0d5521af51a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74db7eaba0d5521af51a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/79b3a2206a8a8bb0c21dfcf76c6a222b470a57f1365bcc47252b525cfe86475b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "79b3a2206a8a8bb0c21dfcf76c6a222b470a57f1365bcc47252b525cfe86475b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "1028d4af78bcae34a28c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74db7eaba0d5521af51a/workflow.json b/records/sd1/sweep/74db7eaba0d5521af51a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f00b55a25e4ef1e557e1fd0a94b978f916ee2021
--- /dev/null
+++ b/records/sd1/sweep/74db7eaba0d5521af51a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74db7eaba0d5521af51a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74e4f048bba1ce2340b4/measurements.json b/records/sd1/sweep/74e4f048bba1ce2340b4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..48d3eae9b05bca41eca03eb157de66fecd03948d
--- /dev/null
+++ b/records/sd1/sweep/74e4f048bba1ce2340b4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74e4f048bba1ce2340b4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.191854,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.258571,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.14441,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.026496,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.053,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00572,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.428,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9945,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1009,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74e4f048bba1ce2340b4/record.json b/records/sd1/sweep/74e4f048bba1ce2340b4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e0c38f516eb3e3a88cc8c2f78d2bfc6a74a7dd8
--- /dev/null
+++ b/records/sd1/sweep/74e4f048bba1ce2340b4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74e4f048bba1ce2340b4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/4b034bdb26778696984cf36829d19c046236e1959cdb52a23c7424d835b0e68d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4b034bdb26778696984cf36829d19c046236e1959cdb52a23c7424d835b0e68d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "28f302047c75f137084f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74e4f048bba1ce2340b4/workflow.json b/records/sd1/sweep/74e4f048bba1ce2340b4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..32bb7db7cb6b07427e46375bdd10e4f036909ce2
--- /dev/null
+++ b/records/sd1/sweep/74e4f048bba1ce2340b4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74e4f048bba1ce2340b4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74eb40765b650595c75d/measurements.json b/records/sd1/sweep/74eb40765b650595c75d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7104624f35f2f80db509a052ba8377d517c90cb
--- /dev/null
+++ b/records/sd1/sweep/74eb40765b650595c75d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74eb40765b650595c75d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.120216,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.169448,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.066176,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.041542,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.851,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00415,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.379,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9616,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0627,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74eb40765b650595c75d/record.json b/records/sd1/sweep/74eb40765b650595c75d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9276c96a525710579a14ef673064ef42217a6c47
--- /dev/null
+++ b/records/sd1/sweep/74eb40765b650595c75d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74eb40765b650595c75d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/069e6dddf281f411a26df6c69ccbc915c730e514993824bb2fef3efe367369da.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "069e6dddf281f411a26df6c69ccbc915c730e514993824bb2fef3efe367369da",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8f856748170424973aeb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74eb40765b650595c75d/workflow.json b/records/sd1/sweep/74eb40765b650595c75d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c75e7583ac070da557512180a3fb29d05ba8c426
--- /dev/null
+++ b/records/sd1/sweep/74eb40765b650595c75d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74eb40765b650595c75d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f0326d60879943e257/measurements.json b/records/sd1/sweep/74f0326d60879943e257/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6664fd91128bc43d6d1e16a116e1d16fc6cf20f
--- /dev/null
+++ b/records/sd1/sweep/74f0326d60879943e257/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74f0326d60879943e257",
+ "values": {
+  "cosine_distance": {
+   "value": 0.389311,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.434494,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.278386,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.110572,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.461,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00531,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.375,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9743,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3127,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f0326d60879943e257/record.json b/records/sd1/sweep/74f0326d60879943e257/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..989b18233e89c3a50b7a7867ebf1391baf778c3f
--- /dev/null
+++ b/records/sd1/sweep/74f0326d60879943e257/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "74f0326d60879943e257",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.skip_connection",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "mid",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/68efbdb9ca59beb16247925c157b4ea67f1c3fb6226c458f2c4dfb75e8d4e968.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "68efbdb9ca59beb16247925c157b4ea67f1c3fb6226c458f2c4dfb75e8d4e968",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "df3498a4709a09c72f9c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f0326d60879943e257/workflow.json b/records/sd1/sweep/74f0326d60879943e257/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..81787e32261b8caf8e0ddd50d1f6e220e0941ea6
--- /dev/null
+++ b/records/sd1/sweep/74f0326d60879943e257/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74f0326d60879943e257)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.skip_connection\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f0940e4ab8127aef65/measurements.json b/records/sd1/sweep/74f0940e4ab8127aef65/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8724b453098f62199b0051c107c814e19bac3c59
--- /dev/null
+++ b/records/sd1/sweep/74f0940e4ab8127aef65/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "74f0940e4ab8127aef65",
+ "values": {
+  "std": {
+   "value": 44.045,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0165,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.415,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.01955,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.048105,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.018873,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010849,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9929,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0599,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f0940e4ab8127aef65/record.json b/records/sd1/sweep/74f0940e4ab8127aef65/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f36ef7a61de0767d5c4a1a172e916e74d00577c
--- /dev/null
+++ b/records/sd1/sweep/74f0940e4ab8127aef65/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "74f0940e4ab8127aef65",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:56:17Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "heavy",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "add_noise",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f0940e4ab8127aef65/workflow.json b/records/sd1/sweep/74f0940e4ab8127aef65/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..29b175cf4e731d753b1d4cf36df2743654db9b6c
--- /dev/null
+++ b/records/sd1/sweep/74f0940e4ab8127aef65/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74f0940e4ab8127aef65)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f1982961914c180316/measurements.json b/records/sd1/sweep/74f1982961914c180316/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b385ca7879cc7d87670a49ef4a371433e82ca9bd
--- /dev/null
+++ b/records/sd1/sweep/74f1982961914c180316/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "74f1982961914c180316",
+ "values": {
+  "cosine_distance": {
+   "value": 0.09913,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.109652,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038876,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.017957,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.229,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00508,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 16.369,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9528,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1354,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f1982961914c180316/record.json b/records/sd1/sweep/74f1982961914c180316/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f618767e1a9e3c8add9ac173cec63fe81c95bbd2
--- /dev/null
+++ b/records/sd1/sweep/74f1982961914c180316/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "74f1982961914c180316",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9601cf04cc7964ec5166eff68ebd4d3219d5827d59e0d0d27170503aa192503d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9601cf04cc7964ec5166eff68ebd4d3219d5827d59e0d0d27170503aa192503d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "1a64d9692a1287dac0c3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/74f1982961914c180316/workflow.json b/records/sd1/sweep/74f1982961914c180316/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4293d972f22f82875b07372c398386afdf5bc0f2
--- /dev/null
+++ b/records/sd1/sweep/74f1982961914c180316/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 74f1982961914c180316)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751110323e7b730885e4/interpretations.jsonl b/records/sd1/sweep/751110323e7b730885e4/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ec8f3aadf24664c3e5cbfc7b3ffef57c03977cdd
--- /dev/null
+++ b/records/sd1/sweep/751110323e7b730885e4/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "751110323e7b730885e4"}, "kind": "caption", "value": "Tilted, darker neon street with dense blue and orange signs and rainy reflections", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8ffaaeaa9b7ee3f5"}
+{"target": {"record": "751110323e7b730885e4"}, "kind": "change", "value": "The street is redrawn at a tilted, closer angle with deeper blue shadows and denser, more chaotic neon signage, so it feels darker and more grungy.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "75c6d47df0cc205f"}
+{"target": {"record": "751110323e7b730885e4"}, "kind": "keywords", "value": ["tilted neon street", "dark blue night", "dense signs", "grungy city", "dutch angle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e9f367854797eb57"}
+{"target": {"record": "751110323e7b730885e4"}, "kind": "effect_tags", "value": ["new_composition", "dark", "subject_kept", "high_contrast"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f86c70dcfba19857"}
diff --git a/records/sd1/sweep/751110323e7b730885e4/measurements.json b/records/sd1/sweep/751110323e7b730885e4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a9c0db8f32e6f88287a825743a28a7c8f1a691d
--- /dev/null
+++ b/records/sd1/sweep/751110323e7b730885e4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "751110323e7b730885e4",
+ "values": {
+  "std": {
+   "value": 27.476,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.04026,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 34.288,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.32532,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.570383,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.464638,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.169095,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.892,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6896,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751110323e7b730885e4/record.json b/records/sd1/sweep/751110323e7b730885e4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..00975ca5edd8edbd3cd392086a5913fb6c988844
--- /dev/null
+++ b/records/sd1/sweep/751110323e7b730885e4/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "751110323e7b730885e4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:53:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "slight",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "rotate",
+   30.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7c5bd625009750f3e854"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751110323e7b730885e4/workflow.json b/records/sd1/sweep/751110323e7b730885e4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a4f1f09dc3ce85aa2b51331414950296f69320d3
--- /dev/null
+++ b/records/sd1/sweep/751110323e7b730885e4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 751110323e7b730885e4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751556cab678a4cb7aee/measurements.json b/records/sd1/sweep/751556cab678a4cb7aee/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..02eefeaf8ff6fc2878e16007056bb223ed018e4e
--- /dev/null
+++ b/records/sd1/sweep/751556cab678a4cb7aee/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "751556cab678a4cb7aee",
+ "values": {
+  "std": {
+   "value": 46.774,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0134,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.993,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.009672,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.039221,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.006315,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007092,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9872,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3264,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751556cab678a4cb7aee/record.json b/records/sd1/sweep/751556cab678a4cb7aee/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..64b59379a36d299ad814ee6682e5dd237e2cb242
--- /dev/null
+++ b/records/sd1/sweep/751556cab678a4cb7aee/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "751556cab678a4cb7aee",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:25:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8dfba4287956274e352d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751556cab678a4cb7aee/workflow.json b/records/sd1/sweep/751556cab678a4cb7aee/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e261d7a7b8a0bbbf97f529e79f6b81d4a6a0a696
--- /dev/null
+++ b/records/sd1/sweep/751556cab678a4cb7aee/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 751556cab678a4cb7aee)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751f8261227aa1d68c26/measurements.json b/records/sd1/sweep/751f8261227aa1d68c26/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..507624b4f930c1b20dc9892ba3280794d5ed3f4e
--- /dev/null
+++ b/records/sd1/sweep/751f8261227aa1d68c26/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "751f8261227aa1d68c26",
+ "values": {
+  "std": {
+   "value": 50.849,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01494,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 2.612,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.005261,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.017543,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.009507,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003824,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0105,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0905,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751f8261227aa1d68c26/record.json b/records/sd1/sweep/751f8261227aa1d68c26/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..af1aab36279cb268860537f83476b1d5ce4a3401
--- /dev/null
+++ b/records/sd1/sweep/751f8261227aa1d68c26/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "751f8261227aa1d68c26",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:12:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "medium"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_noise",
+   0.2,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "abfbe6d50cd168793d67"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/751f8261227aa1d68c26/workflow.json b/records/sd1/sweep/751f8261227aa1d68c26/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7855d4b224975f21b217408487b947ab53d2526
--- /dev/null
+++ b/records/sd1/sweep/751f8261227aa1d68c26/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 751f8261227aa1d68c26)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/752486356d81dc018f27/interpretations.jsonl b/records/sd1/sweep/752486356d81dc018f27/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6c0e18d00c42c5c001635f44c9130ef7ed7fddb4
--- /dev/null
+++ b/records/sd1/sweep/752486356d81dc018f27/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "752486356d81dc018f27"}, "kind": "caption", "value": "Illustrated portrait of a woman with wavy brown hair on dark grey, nearly unchanged", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "adbb40aaab75d310"}
+{"target": {"record": "752486356d81dc018f27"}, "kind": "change", "value": "Almost no visible change, only tiny differences in hair strands and facial shading.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f75919d14f0a6bc2"}
+{"target": {"record": "752486356d81dc018f27"}, "kind": "keywords", "value": ["nearly identical", "minimal change", "portrait", "wavy hair"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "87a36f55f6537c1a"}
+{"target": {"record": "752486356d81dc018f27"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0d13515cd3df571c"}
diff --git a/records/sd1/sweep/752486356d81dc018f27/measurements.json b/records/sd1/sweep/752486356d81dc018f27/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ebba6e1357aec1432775cb4b77ee72087821682
--- /dev/null
+++ b/records/sd1/sweep/752486356d81dc018f27/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "752486356d81dc018f27",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.618,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00534,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1553,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/752486356d81dc018f27/record.json b/records/sd1/sweep/752486356d81dc018f27/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..babaa8245e9dbe9734a652fa2fdad975bdd220d2
--- /dev/null
+++ b/records/sd1/sweep/752486356d81dc018f27/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "752486356d81dc018f27",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1eb774744f0916d6e42923f5d02d8b213056b594e29f07ccbf8573adc314121b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1eb774744f0916d6e42923f5d02d8b213056b594e29f07ccbf8573adc314121b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "40f9943c4cb1ae31a2b8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/752486356d81dc018f27/workflow.json b/records/sd1/sweep/752486356d81dc018f27/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9bcca5b3102f59c78656fd2a13dcff96e5d84a13
--- /dev/null
+++ b/records/sd1/sweep/752486356d81dc018f27/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 752486356d81dc018f27)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.h_upd\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/752a3257da42b124b57a/measurements.json b/records/sd1/sweep/752a3257da42b124b57a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cecaae13803ca1204f04f86ab9c916c53492d9ff
--- /dev/null
+++ b/records/sd1/sweep/752a3257da42b124b57a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "752a3257da42b124b57a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.025372,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.047788,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.015292,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002805,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.196,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00665,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.369,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0196,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1076,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/752a3257da42b124b57a/record.json b/records/sd1/sweep/752a3257da42b124b57a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..67a51b63b54f434e96f3ed4a931ea198f2cb0369
--- /dev/null
+++ b/records/sd1/sweep/752a3257da42b124b57a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "752a3257da42b124b57a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:50Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/03d5a885b3018d4614d83b1983a7eee7563f5c4859296676d2e9d262952d6c28.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "03d5a885b3018d4614d83b1983a7eee7563f5c4859296676d2e9d262952d6c28",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ececa4550707166e0e6d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/752a3257da42b124b57a/workflow.json b/records/sd1/sweep/752a3257da42b124b57a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc7c4d861b805a19242e199bedf5b86b47f71685
--- /dev/null
+++ b/records/sd1/sweep/752a3257da42b124b57a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 752a3257da42b124b57a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.in_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7535b8f2f0c2a997740c/measurements.json b/records/sd1/sweep/7535b8f2f0c2a997740c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..309b95d1f4ddb4dd10589429f964a38653d53129
--- /dev/null
+++ b/records/sd1/sweep/7535b8f2f0c2a997740c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7535b8f2f0c2a997740c",
+ "values": {
+  "std": {
+   "value": 41.506,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01671,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.055,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.1048,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.336868,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.43859,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.167991,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0748,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0129,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7535b8f2f0c2a997740c/record.json b/records/sd1/sweep/7535b8f2f0c2a997740c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..36c9254ea119f261f378c3b2fc9bc9487c7a3623
--- /dev/null
+++ b/records/sd1/sweep/7535b8f2f0c2a997740c/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7535b8f2f0c2a997740c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:57:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "shrink",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "scale",
+   0.7,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5e462f42491d57d3fc9e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7535b8f2f0c2a997740c/workflow.json b/records/sd1/sweep/7535b8f2f0c2a997740c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e800103d22a4441cc1e5b0409a65da42d38f866e
--- /dev/null
+++ b/records/sd1/sweep/7535b8f2f0c2a997740c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7535b8f2f0c2a997740c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7542fd0536db25472e2a/measurements.json b/records/sd1/sweep/7542fd0536db25472e2a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c4bbdeaa8e2e46f4bf7f5c2a522d0b4b53fd743
--- /dev/null
+++ b/records/sd1/sweep/7542fd0536db25472e2a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7542fd0536db25472e2a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.131281,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.309437,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.22717,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.088425,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.37,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00351,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.456,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0269,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0672,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7542fd0536db25472e2a/record.json b/records/sd1/sweep/7542fd0536db25472e2a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..314fd129b0292776648eeacd73a84dae123e221f
--- /dev/null
+++ b/records/sd1/sweep/7542fd0536db25472e2a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7542fd0536db25472e2a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/6942244b740adf8cfb589f3852c5154c279287a64a206703ab06402539dc7650.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6942244b740adf8cfb589f3852c5154c279287a64a206703ab06402539dc7650",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "24e0f6cd3cacabeb8294"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7542fd0536db25472e2a/workflow.json b/records/sd1/sweep/7542fd0536db25472e2a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..146d76b39ebc3d66a2f23c01192b8db3e41fcb2f
--- /dev/null
+++ b/records/sd1/sweep/7542fd0536db25472e2a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7542fd0536db25472e2a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/755245009e763508bb4a/measurements.json b/records/sd1/sweep/755245009e763508bb4a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4ab57816ac98044d0cb2399db77cf291841f425
--- /dev/null
+++ b/records/sd1/sweep/755245009e763508bb4a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "755245009e763508bb4a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.615612,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.864452,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.848513,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.366472,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 17.749,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01093,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 53.719,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9009,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1881,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/755245009e763508bb4a/record.json b/records/sd1/sweep/755245009e763508bb4a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e217565a6ffd27cc2dea42275bf33b28592c422f
--- /dev/null
+++ b/records/sd1/sweep/755245009e763508bb4a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "755245009e763508bb4a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:38Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9c91507bb08c7f22eaac7e87d4c1da68d009070ff9ba4355295886609175d077.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9c91507bb08c7f22eaac7e87d4c1da68d009070ff9ba4355295886609175d077",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "bbd3ab7bce0d0f074c49"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/755245009e763508bb4a/workflow.json b/records/sd1/sweep/755245009e763508bb4a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cd54f7e4250a6234e05d7006e0eec853bbc0c87
--- /dev/null
+++ b/records/sd1/sweep/755245009e763508bb4a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 755245009e763508bb4a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/756274ffca019a89953c/measurements.json b/records/sd1/sweep/756274ffca019a89953c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..02b7d6d597482aa9801c3274b3a2aa12874b6003
--- /dev/null
+++ b/records/sd1/sweep/756274ffca019a89953c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "756274ffca019a89953c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.416438,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.505581,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.321385,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.138153,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.828,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00542,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 48.366,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9938,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2344,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/756274ffca019a89953c/record.json b/records/sd1/sweep/756274ffca019a89953c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a9d591981cda2c4ed667eb592c561f946daef6a
--- /dev/null
+++ b/records/sd1/sweep/756274ffca019a89953c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "756274ffca019a89953c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.out_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5f394a3977ad391224c79a92c7f387912fa5c77eec94c7747db38b5f97aeb8b7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5f394a3977ad391224c79a92c7f387912fa5c77eec94c7747db38b5f97aeb8b7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "90fa07c66895649bd8ca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/756274ffca019a89953c/workflow.json b/records/sd1/sweep/756274ffca019a89953c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4ea009c76bab124ed4cc97b23b73103ed0c432
--- /dev/null
+++ b/records/sd1/sweep/756274ffca019a89953c/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 756274ffca019a89953c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.out_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/756ea17f64ab2d98df8e/measurements.json b/records/sd1/sweep/756ea17f64ab2d98df8e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..925ac1917bc7e56510f10af7711cc9e2f7508508
--- /dev/null
+++ b/records/sd1/sweep/756ea17f64ab2d98df8e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "756ea17f64ab2d98df8e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.099042,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.098284,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.03911,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.071533,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.292,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00596,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.167,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9817,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2392,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/756ea17f64ab2d98df8e/record.json b/records/sd1/sweep/756ea17f64ab2d98df8e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc4bf56c9bd05b68f60012e7063e1c254483a1a3
--- /dev/null
+++ b/records/sd1/sweep/756ea17f64ab2d98df8e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "756ea17f64ab2d98df8e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:12Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/a0375dcfd7cd7f8b59c66b3a4c6bf346d04c3be9315e090e831af0880d803ebb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a0375dcfd7cd7f8b59c66b3a4c6bf346d04c3be9315e090e831af0880d803ebb",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "8305c381dd1af0fd8f56"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/756ea17f64ab2d98df8e/workflow.json b/records/sd1/sweep/756ea17f64ab2d98df8e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c08cc2f44c3fbc7305d6262c6431f8bf5d1c8320
--- /dev/null
+++ b/records/sd1/sweep/756ea17f64ab2d98df8e/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 756ea17f64ab2d98df8e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7570a32d193c99a53ccd/measurements.json b/records/sd1/sweep/7570a32d193c99a53ccd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f11d2573d41ae9b770f21f97701cbedd520bb668
--- /dev/null
+++ b/records/sd1/sweep/7570a32d193c99a53ccd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7570a32d193c99a53ccd",
+ "values": {
+  "std": {
+   "value": 40.027,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01734,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.616,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.281927,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.418452,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.128718,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.031967,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9912,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0513,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7570a32d193c99a53ccd/record.json b/records/sd1/sweep/7570a32d193c99a53ccd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fb0423e4a43404070c1948567529227946a073e
--- /dev/null
+++ b/records/sd1/sweep/7570a32d193c99a53ccd/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "7570a32d193c99a53ccd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:53:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7570a32d193c99a53ccd/workflow.json b/records/sd1/sweep/7570a32d193c99a53ccd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5767de05a972113b1d134f1f636bfaa02c1b7557
--- /dev/null
+++ b/records/sd1/sweep/7570a32d193c99a53ccd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7570a32d193c99a53ccd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7574bbafa84ae25245dd/measurements.json b/records/sd1/sweep/7574bbafa84ae25245dd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..59ea7ff920d8c5fdf1d50da2ac853c80258fd08d
--- /dev/null
+++ b/records/sd1/sweep/7574bbafa84ae25245dd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7574bbafa84ae25245dd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.097909,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.13989,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.061362,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025948,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.545,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00477,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.093,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9954,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2472,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7574bbafa84ae25245dd/record.json b/records/sd1/sweep/7574bbafa84ae25245dd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..299a5a80c61ba3a8f6d7ddd89fd9e60e2086804b
--- /dev/null
+++ b/records/sd1/sweep/7574bbafa84ae25245dd/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7574bbafa84ae25245dd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:13Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5c26441d04eba884ae2722210d4e74b76c469dd83e9f7e3af945a2005aa2cb22.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5c26441d04eba884ae2722210d4e74b76c469dd83e9f7e3af945a2005aa2cb22",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6dc09f8e2c462c9a04ed"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7574bbafa84ae25245dd/workflow.json b/records/sd1/sweep/7574bbafa84ae25245dd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fe8e3bc82540fb6d3130d4b9d597b3748a0f19a
--- /dev/null
+++ b/records/sd1/sweep/7574bbafa84ae25245dd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7574bbafa84ae25245dd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.transformer_blocks.0.attn1.to_out.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/757a0e6a310aa974f7e2/measurements.json b/records/sd1/sweep/757a0e6a310aa974f7e2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea3e25b67f48764e24a42c8253b491f2b5792f10
--- /dev/null
+++ b/records/sd1/sweep/757a0e6a310aa974f7e2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "757a0e6a310aa974f7e2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.179708,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.286168,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.282677,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.13681,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.298,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00266,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.317,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0003,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1405,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/757a0e6a310aa974f7e2/record.json b/records/sd1/sweep/757a0e6a310aa974f7e2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8de4fa92f093a6155e34ade1b9b6ad224c21cf2a
--- /dev/null
+++ b/records/sd1/sweep/757a0e6a310aa974f7e2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "757a0e6a310aa974f7e2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/24e7dad45b1f4e57b7c76e45148863f170b644c4e73ee5068b11140485bc6db6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "24e7dad45b1f4e57b7c76e45148863f170b644c4e73ee5068b11140485bc6db6",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "a027e943ec4bd37f037a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/757a0e6a310aa974f7e2/workflow.json b/records/sd1/sweep/757a0e6a310aa974f7e2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dddd886b0b4745df85c1b06159ba8e421d5fbbae
--- /dev/null
+++ b/records/sd1/sweep/757a0e6a310aa974f7e2/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 757a0e6a310aa974f7e2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7580d74d132c48a96d9d/interpretations.jsonl b/records/sd1/sweep/7580d74d132c48a96d9d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d020f0c19878b5f977e1f3f937fd102465ed21c3
--- /dev/null
+++ b/records/sd1/sweep/7580d74d132c48a96d9d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7580d74d132c48a96d9d"}, "kind": "caption", "value": "Blurry mottled grey and beige texture with no recognisable objects", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ba0ae087cfb622ec"}
+{"target": {"record": "7580d74d132c48a96d9d"}, "kind": "change", "value": "The still life completely dissolves into an abstract, camouflage-like blotchy noise texture in grey and tan.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e9048faaf946f335"}
+{"target": {"record": "7580d74d132c48a96d9d"}, "kind": "keywords", "value": ["abstract texture", "mottled noise", "camouflage pattern", "collapsed image"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a5c2aaef894b59c5"}
+{"target": {"record": "7580d74d132c48a96d9d"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "texture", "blur", "degenerate", "desaturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a0eabb3f5fc00a11"}
diff --git a/records/sd1/sweep/7580d74d132c48a96d9d/measurements.json b/records/sd1/sweep/7580d74d132c48a96d9d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b64b4609dee1b72a1cf84f193277ef0329c997e3
--- /dev/null
+++ b/records/sd1/sweep/7580d74d132c48a96d9d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7580d74d132c48a96d9d",
+ "values": {
+  "std": {
+   "value": 26.083,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00411,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.908,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.716121,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.807347,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 1.039455,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.431362,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.5038,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6129,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7580d74d132c48a96d9d/record.json b/records/sd1/sweep/7580d74d132c48a96d9d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45487f9d960fa32285d5927171cd3db662bdb618
--- /dev/null
+++ b/records/sd1/sweep/7580d74d132c48a96d9d/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "7580d74d132c48a96d9d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:21:13Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "3b7d00a9f8bb50029c8d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7580d74d132c48a96d9d/workflow.json b/records/sd1/sweep/7580d74d132c48a96d9d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..313fc43956504cb099127e3b8b819f9167bb7203
--- /dev/null
+++ b/records/sd1/sweep/7580d74d132c48a96d9d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7580d74d132c48a96d9d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75879c0efb7c5f841a18/measurements.json b/records/sd1/sweep/75879c0efb7c5f841a18/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..94d9f65547cd11490c040a5f90ee72e0dee60d81
--- /dev/null
+++ b/records/sd1/sweep/75879c0efb7c5f841a18/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75879c0efb7c5f841a18",
+ "values": {
+  "cosine_distance": {
+   "value": 0.333073,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.585197,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.370052,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.236108,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 43.878,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00825,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.644,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9951,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4172,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75879c0efb7c5f841a18/record.json b/records/sd1/sweep/75879c0efb7c5f841a18/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..07f8d2ca933be86bfa33f442919371b2ac74eaca
--- /dev/null
+++ b/records/sd1/sweep/75879c0efb7c5f841a18/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "75879c0efb7c5f841a18",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/9ab7da3185d39378b7e1554088611c705d48c6dc155e3844fc22e8646330bdd7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9ab7da3185d39378b7e1554088611c705d48c6dc155e3844fc22e8646330bdd7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "cb8adda58a1aee9e53bf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75879c0efb7c5f841a18/workflow.json b/records/sd1/sweep/75879c0efb7c5f841a18/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..34e296f43ab612ab09c44492c29f7784735a682d
--- /dev/null
+++ b/records/sd1/sweep/75879c0efb7c5f841a18/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75879c0efb7c5f841a18)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7587eef302e82ae6125b/interpretations.jsonl b/records/sd1/sweep/7587eef302e82ae6125b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..f4883403f6f171348cce4885cd89c3ee2d399fdf
--- /dev/null
+++ b/records/sd1/sweep/7587eef302e82ae6125b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7587eef302e82ae6125b"}, "kind": "caption", "value": "Painterly view of churning surf over brown rocks in muted earthy colours", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9630a3bc62ad12af"}
+{"target": {"record": "7587eef302e82ae6125b"}, "kind": "change", "value": "The lighthouse and sunset sky disappear, replaced by a close top-down-feeling impasto study of waves and rocks in muted browns, creams and blues.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a19fe5404ce19156"}
+{"target": {"record": "7587eef302e82ae6125b"}, "kind": "keywords", "value": ["impasto surf", "rocky shore", "muted earth tones", "oil painting", "no lighthouse"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e5ceba22594b1541"}
+{"target": {"record": "7587eef302e82ae6125b"}, "kind": "effect_tags", "value": ["painterly", "subject_lost", "new_composition", "texture", "desaturated", "zoom"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e3c626f3a62777a1"}
diff --git a/records/sd1/sweep/7587eef302e82ae6125b/measurements.json b/records/sd1/sweep/7587eef302e82ae6125b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ba10f61ee0a61539f9921e9860d7a38a6877f64
--- /dev/null
+++ b/records/sd1/sweep/7587eef302e82ae6125b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7587eef302e82ae6125b",
+ "values": {
+  "std": {
+   "value": 29.379,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02225,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.969,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.396037,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.597431,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.528052,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.273576,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9673,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0063,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7587eef302e82ae6125b/record.json b/records/sd1/sweep/7587eef302e82ae6125b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfc8273fdb7395ad3c37bc096559c788798ca512
--- /dev/null
+++ b/records/sd1/sweep/7587eef302e82ae6125b/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7587eef302e82ae6125b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:57:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "quarter",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   90.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5d06c4698546328bc3aa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7587eef302e82ae6125b/workflow.json b/records/sd1/sweep/7587eef302e82ae6125b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d387b9d6f9f2da7dcfcb29df94eab2be03db2df
--- /dev/null
+++ b/records/sd1/sweep/7587eef302e82ae6125b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7587eef302e82ae6125b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75a986b7999dcb2c492e/measurements.json b/records/sd1/sweep/75a986b7999dcb2c492e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9da00c5490ca148bb15a46b05aa6c379d086cb2d
--- /dev/null
+++ b/records/sd1/sweep/75a986b7999dcb2c492e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "75a986b7999dcb2c492e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.594011,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.60555,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.402238,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.306218,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 87.264,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00239,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 58.636,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9658,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7443,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75a986b7999dcb2c492e/record.json b/records/sd1/sweep/75a986b7999dcb2c492e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..761ad50d64335fc9daa2de77fcf7b60e6545a788
--- /dev/null
+++ b/records/sd1/sweep/75a986b7999dcb2c492e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "75a986b7999dcb2c492e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/3ddd66748d34c1cb39cde248f7661a301aa39311a290fe01d4ea689840f853f0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3ddd66748d34c1cb39cde248f7661a301aa39311a290fe01d4ea689840f853f0",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "5e44ba77f5f56791d04b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75a986b7999dcb2c492e/workflow.json b/records/sd1/sweep/75a986b7999dcb2c492e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7aab82e47aa65dcd9809091e3109f181e11eb3d7
--- /dev/null
+++ b/records/sd1/sweep/75a986b7999dcb2c492e/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75a986b7999dcb2c492e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75b4ab9b7a463a9f0df3/measurements.json b/records/sd1/sweep/75b4ab9b7a463a9f0df3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..37a24fe8948c06c60db4d86c8c45f234efeea416
--- /dev/null
+++ b/records/sd1/sweep/75b4ab9b7a463a9f0df3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75b4ab9b7a463a9f0df3",
+ "values": {
+  "std": {
+   "value": 27.513,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01565,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.938,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.461651,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.698913,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.663287,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.186708,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0122,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1108,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75b4ab9b7a463a9f0df3/record.json b/records/sd1/sweep/75b4ab9b7a463a9f0df3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f79493e3592101d9f467e70b8baec787bb1a9742
--- /dev/null
+++ b/records/sd1/sweep/75b4ab9b7a463a9f0df3/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "75b4ab9b7a463a9f0df3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:19:40Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "shrink",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "scale",
+   0.7,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75b4ab9b7a463a9f0df3/workflow.json b/records/sd1/sweep/75b4ab9b7a463a9f0df3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a86a4dacf4a5cd58280f43deb6573caf180ddd1
--- /dev/null
+++ b/records/sd1/sweep/75b4ab9b7a463a9f0df3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75b4ab9b7a463a9f0df3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75c059b0fa1465204dcd/measurements.json b/records/sd1/sweep/75c059b0fa1465204dcd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c6bd697b61331608ed1fb49210da75e9bfcd571
--- /dev/null
+++ b/records/sd1/sweep/75c059b0fa1465204dcd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75c059b0fa1465204dcd",
+ "values": {
+  "std": {
+   "value": 43.363,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00822,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.363,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.180035,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.322871,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.078348,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.048947,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0007,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.012,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75c059b0fa1465204dcd/record.json b/records/sd1/sweep/75c059b0fa1465204dcd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a8d0231e7a9f301d7b4076a958bfa6693dacd0e
--- /dev/null
+++ b/records/sd1/sweep/75c059b0fa1465204dcd/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "75c059b0fa1465204dcd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:15:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.5,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6398875e9030e4feda5c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75c059b0fa1465204dcd/workflow.json b/records/sd1/sweep/75c059b0fa1465204dcd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..96be3ed6681381200a7e91b42b037db46ed04ee3
--- /dev/null
+++ b/records/sd1/sweep/75c059b0fa1465204dcd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75c059b0fa1465204dcd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75c1ec8c5737fd06c8ff/interpretations.jsonl b/records/sd1/sweep/75c1ec8c5737fd06c8ff/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..f7f55c8d5c9a001ae155b0d1a8dd069852fcfc31
--- /dev/null
+++ b/records/sd1/sweep/75c1ec8c5737fd06c8ff/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "75c1ec8c5737fd06c8ff"}, "kind": "caption", "value": "Coarse saturated rainbow noise blobs covering the whole image, with no subject visible.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "57eadcd8440c5169"}
+{"target": {"record": "75c1ec8c5737fd06c8ff"}, "kind": "change", "value": "Like B but with larger, chunkier blobs: the photo collapses into vivid multicoloured noise and the figure is completely lost.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1f0ba3a4e49f51b0"}
+{"target": {"record": "75c1ec8c5737fd06c8ff"}, "kind": "keywords", "value": ["coarse colour noise", "rainbow blobs", "psychedelic static", "image collapse"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d9fd907eb1cb9cbf"}
+{"target": {"record": "75c1ec8c5737fd06c8ff"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "saturated", "pattern", "texture", "glitch", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e76797c593fbcd27"}
diff --git a/records/sd1/sweep/75c1ec8c5737fd06c8ff/measurements.json b/records/sd1/sweep/75c1ec8c5737fd06c8ff/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..41cd68a981544b81840b82b1c6ff5873fc606382
--- /dev/null
+++ b/records/sd1/sweep/75c1ec8c5737fd06c8ff/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "75c1ec8c5737fd06c8ff",
+ "values": {
+  "cosine_distance": {
+   "value": 0.768813,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.914479,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.949949,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.543435,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 85.905,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01364,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "clipped"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "clip_degenerate"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 101.988,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8086,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9837,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75c1ec8c5737fd06c8ff/record.json b/records/sd1/sweep/75c1ec8c5737fd06c8ff/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5cdef32704be4ebdf3bb2d0380f15ceac6feae79
--- /dev/null
+++ b/records/sd1/sweep/75c1ec8c5737fd06c8ff/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "75c1ec8c5737fd06c8ff",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.skip_connection",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.hi",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/3fbbdfdca1c0e6f60d91d78edb3e3f4f0f97fad01bd4bba4b4dac5afa499ff2b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3fbbdfdca1c0e6f60d91d78edb3e3f4f0f97fad01bd4bba4b4dac5afa499ff2b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c2aa1e419c086d149206"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75c1ec8c5737fd06c8ff/workflow.json b/records/sd1/sweep/75c1ec8c5737fd06c8ff/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a498b563fbdba8ce71f8b7a2a24039df2db135a5
--- /dev/null
+++ b/records/sd1/sweep/75c1ec8c5737fd06c8ff/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75c1ec8c5737fd06c8ff)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.skip_connection\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75ce512ff8ac7e0c8b24/measurements.json b/records/sd1/sweep/75ce512ff8ac7e0c8b24/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a17dc551c78e07e76e00cc1aa6d552028a9ab7bb
--- /dev/null
+++ b/records/sd1/sweep/75ce512ff8ac7e0c8b24/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75ce512ff8ac7e0c8b24",
+ "values": {
+  "std": {
+   "value": 46.746,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01122,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.425,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.159282,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.25591,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.085547,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.057879,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0177,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0393,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75ce512ff8ac7e0c8b24/record.json b/records/sd1/sweep/75ce512ff8ac7e0c8b24/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..248f3ff710080684e64a5d9249fce40aeb93d9f8
--- /dev/null
+++ b/records/sd1/sweep/75ce512ff8ac7e0c8b24/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "75ce512ff8ac7e0c8b24",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:42:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "boost",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   2.0,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "bfb25b3e0274c50ea387"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75ce512ff8ac7e0c8b24/workflow.json b/records/sd1/sweep/75ce512ff8ac7e0c8b24/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8e2f3f5f44a0e18286968c9d8960b48250f5175
--- /dev/null
+++ b/records/sd1/sweep/75ce512ff8ac7e0c8b24/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75ce512ff8ac7e0c8b24)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75d3cc2243862c0bcf0e/measurements.json b/records/sd1/sweep/75d3cc2243862c0bcf0e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec1a0e963bb7e00b5fae55ecd9ad824d5c0f2ea6
--- /dev/null
+++ b/records/sd1/sweep/75d3cc2243862c0bcf0e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "75d3cc2243862c0bcf0e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.076194,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.110444,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.035607,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025637,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.528,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00401,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.176,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0488,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1188,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75d3cc2243862c0bcf0e/record.json b/records/sd1/sweep/75d3cc2243862c0bcf0e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..841a5b21df99d8c011decbdeab98c3df3ceb139d
--- /dev/null
+++ b/records/sd1/sweep/75d3cc2243862c0bcf0e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "75d3cc2243862c0bcf0e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.hi",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7eb7fb41fc19dd9b489265c6dbfd291e917c9ff0a158a314babe885386e23789.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7eb7fb41fc19dd9b489265c6dbfd291e917c9ff0a158a314babe885386e23789",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5a77beecaaae85c8af65"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75d3cc2243862c0bcf0e/workflow.json b/records/sd1/sweep/75d3cc2243862c0bcf0e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..afab21cf662d9048f5959ee14cc7b5bdb43220bb
--- /dev/null
+++ b/records/sd1/sweep/75d3cc2243862c0bcf0e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75d3cc2243862c0bcf0e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.transformer_blocks.0.attn2.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75d479a1f49f4c2ba3de/measurements.json b/records/sd1/sweep/75d479a1f49f4c2ba3de/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4106e2188d2d204db9d91de7469c210777237ed
--- /dev/null
+++ b/records/sd1/sweep/75d479a1f49f4c2ba3de/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "75d479a1f49f4c2ba3de",
+ "values": {
+  "cosine_distance": {
+   "value": 0.057317,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.082131,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.021048,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.015702,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.597,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00503,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.677,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0149,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1891,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75d479a1f49f4c2ba3de/record.json b/records/sd1/sweep/75d479a1f49f4c2ba3de/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e965299bdc6e8d0473dc9dd3ad6deb5c6bce4619
--- /dev/null
+++ b/records/sd1/sweep/75d479a1f49f4c2ba3de/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "75d479a1f49f4c2ba3de",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/148dc4affada0ef8ab3918080f9b4f8d634c73a7080a7137484021541254e1ec.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "148dc4affada0ef8ab3918080f9b4f8d634c73a7080a7137484021541254e1ec",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "13a3ddef9c2829aa8247"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75d479a1f49f4c2ba3de/workflow.json b/records/sd1/sweep/75d479a1f49f4c2ba3de/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0437083d57afc21c7631346347b19102d655ed9f
--- /dev/null
+++ b/records/sd1/sweep/75d479a1f49f4c2ba3de/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75d479a1f49f4c2ba3de)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.attn1.to_out.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e3bd4bc0791981f505/measurements.json b/records/sd1/sweep/75e3bd4bc0791981f505/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..23f036e9fef3d7daa40ffb688355c74937ab5de6
--- /dev/null
+++ b/records/sd1/sweep/75e3bd4bc0791981f505/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75e3bd4bc0791981f505",
+ "values": {
+  "cosine_distance": {
+   "value": 0.398934,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.456382,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.576954,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.389187,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.06,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00569,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.721,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8646,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9241,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e3bd4bc0791981f505/record.json b/records/sd1/sweep/75e3bd4bc0791981f505/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8b198e058aaef5cb07763deceb38bce22a20caf
--- /dev/null
+++ b/records/sd1/sweep/75e3bd4bc0791981f505/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "75e3bd4bc0791981f505",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.proj_out",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/fa553d62f3b737d460958da144ca82c376891c1b933fbb8f44b77a4002f02462.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "fa553d62f3b737d460958da144ca82c376891c1b933fbb8f44b77a4002f02462",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "2f752586f54f82e98b56"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e3bd4bc0791981f505/workflow.json b/records/sd1/sweep/75e3bd4bc0791981f505/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6392723652ea0ee92c56a8d9e15164b81ad52d6e
--- /dev/null
+++ b/records/sd1/sweep/75e3bd4bc0791981f505/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75e3bd4bc0791981f505)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.proj_out\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e7474411f18b773382/interpretations.jsonl b/records/sd1/sweep/75e7474411f18b773382/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..b6d771c0774fc53143eef94840522d037ce5c09d
--- /dev/null
+++ b/records/sd1/sweep/75e7474411f18b773382/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "75e7474411f18b773382"}, "kind": "caption", "value": "Smiling woman with short dark hair in a black turtleneck, a black bar across her mouth", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "754b7b1a8a0962a7"}
+{"target": {"record": "75e7474411f18b773382"}, "kind": "change", "value": "The pose becomes frontal with a broad smile and a turtleneck on a teal-grey backdrop, and a strange black horizontal bar cuts across the lower face.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e963b2fe6b7ae9c0"}
+{"target": {"record": "75e7474411f18b773382"}, "kind": "keywords", "value": ["smiling portrait", "black bar", "censored mouth", "turtleneck", "teal backdrop"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ce545742c7d87a15"}
+{"target": {"record": "75e7474411f18b773382"}, "kind": "effect_tags", "value": ["subject_kept", "surreal", "glitch", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e480e4a5951fcb88"}
diff --git a/records/sd1/sweep/75e7474411f18b773382/measurements.json b/records/sd1/sweep/75e7474411f18b773382/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..98b70ccccd7445692878931489a51793033a44a8
--- /dev/null
+++ b/records/sd1/sweep/75e7474411f18b773382/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "75e7474411f18b773382",
+ "values": {
+  "cosine_distance": {
+   "value": 0.30067,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.333766,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.493017,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.153514,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 75.925,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00233,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.12,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0029,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1786,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e7474411f18b773382/record.json b/records/sd1/sweep/75e7474411f18b773382/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf538d26d9600f8c64363ee7b43c4a844993dc52
--- /dev/null
+++ b/records/sd1/sweep/75e7474411f18b773382/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "75e7474411f18b773382",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.10.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/57ae950f7e782cb01c415609cdc2a36517c9cc04f382effe4959df0f3b48d01a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "57ae950f7e782cb01c415609cdc2a36517c9cc04f382effe4959df0f3b48d01a",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "13e86ecda6fd2ee5ab04"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e7474411f18b773382/workflow.json b/records/sd1/sweep/75e7474411f18b773382/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d59ba4471c577ad2aed0b6ebc7d96b926ad6828
--- /dev/null
+++ b/records/sd1/sweep/75e7474411f18b773382/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75e7474411f18b773382)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.skip_connection\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e7cb4119b49780d8d2/measurements.json b/records/sd1/sweep/75e7cb4119b49780d8d2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1438be8959ca5e6006547b559d2ff3445c5627e
--- /dev/null
+++ b/records/sd1/sweep/75e7cb4119b49780d8d2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75e7cb4119b49780d8d2",
+ "values": {
+  "std": {
+   "value": 32.415,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0253,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.922,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.118671,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.246363,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.093134,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.057839,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9086,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2622,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e7cb4119b49780d8d2/record.json b/records/sd1/sweep/75e7cb4119b49780d8d2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..71d4aabe198f9822c9896c4f01588cac02351b66
--- /dev/null
+++ b/records/sd1/sweep/75e7cb4119b49780d8d2/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "75e7cb4119b49780d8d2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:09:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "boost",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   2.0,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8f9cdcc1901af8458899"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75e7cb4119b49780d8d2/workflow.json b/records/sd1/sweep/75e7cb4119b49780d8d2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9daf1e90982aaae3e14f3e95d21cba9c1c8462d8
--- /dev/null
+++ b/records/sd1/sweep/75e7cb4119b49780d8d2/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75e7cb4119b49780d8d2)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75f38de5297e1989463e/measurements.json b/records/sd1/sweep/75f38de5297e1989463e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec4ab051c70a38d0d1fe72e3976b475bc96cd2e1
--- /dev/null
+++ b/records/sd1/sweep/75f38de5297e1989463e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75f38de5297e1989463e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.036154,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.068293,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.05739,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007712,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.334,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00624,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 8.142,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0103,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.094,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75f38de5297e1989463e/record.json b/records/sd1/sweep/75f38de5297e1989463e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e6c4d51b17d90c1c701f6c63d6c149bfdc55090
--- /dev/null
+++ b/records/sd1/sweep/75f38de5297e1989463e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "75f38de5297e1989463e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/df3a1175811d88bfbd4ddc3a59e25672d6e0309f32d5a1d458cc94212cee351c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "df3a1175811d88bfbd4ddc3a59e25672d6e0309f32d5a1d458cc94212cee351c",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "bc714ef5d387fa156d2e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75f38de5297e1989463e/workflow.json b/records/sd1/sweep/75f38de5297e1989463e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6b7aaebf6098bb8d60b8e2ca2447afe386f5c73
--- /dev/null
+++ b/records/sd1/sweep/75f38de5297e1989463e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75f38de5297e1989463e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75f5ff5179d4387dd92d/measurements.json b/records/sd1/sweep/75f5ff5179d4387dd92d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..26db8cabfffbaf616e62afa0895414471e03dc89
--- /dev/null
+++ b/records/sd1/sweep/75f5ff5179d4387dd92d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "75f5ff5179d4387dd92d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.180181,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.286207,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.283487,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.137552,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.199,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00264,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.347,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0415,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1877,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75f5ff5179d4387dd92d/record.json b/records/sd1/sweep/75f5ff5179d4387dd92d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..47b065ebdc81526bc245dbf0d11d5df4e7bf6908
--- /dev/null
+++ b/records/sd1/sweep/75f5ff5179d4387dd92d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "75f5ff5179d4387dd92d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:22Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/770155748334effd4e11187cac5af0534185662a3f7458c720e1257825907bcb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "770155748334effd4e11187cac5af0534185662a3f7458c720e1257825907bcb",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "4ca6638d78d00d65e2a2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75f5ff5179d4387dd92d/workflow.json b/records/sd1/sweep/75f5ff5179d4387dd92d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..995724bf0cac22e9089f9c7d5aca8ccf9f20caec
--- /dev/null
+++ b/records/sd1/sweep/75f5ff5179d4387dd92d/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75f5ff5179d4387dd92d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75fe25b64b48b44eb4aa/interpretations.jsonl b/records/sd1/sweep/75fe25b64b48b44eb4aa/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a9c3e89943bd328c87733fcd46b71b1e35b6c272
--- /dev/null
+++ b/records/sd1/sweep/75fe25b64b48b44eb4aa/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "75fe25b64b48b44eb4aa"}, "kind": "caption", "value": "Painterly hazy figure of an old man wrapped in shaggy straw-like cloak by the sea", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:27Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ce76fa7ac38bca5d"}
+{"target": {"record": "75fe25b64b48b44eb4aa"}, "kind": "change", "value": "The sharp portrait turns into a soft, faded painterly scene: the man becomes a small head on a huge shaggy, fibrous cloaked body, with washed-out muted colours and a surreal feel.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:27Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1e1b313c783c5108"}
+{"target": {"record": "75fe25b64b48b44eb4aa"}, "kind": "keywords", "value": ["shaggy cloak", "painterly", "faded", "surreal figure", "old man"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:27Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ad4662d86e6b54a4"}
+{"target": {"record": "75fe25b64b48b44eb4aa"}, "kind": "effect_tags", "value": ["new_composition", "painterly", "washed_out", "surreal", "distorted", "texture"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:27Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d5fb37ef5778b293"}
diff --git a/records/sd1/sweep/75fe25b64b48b44eb4aa/measurements.json b/records/sd1/sweep/75fe25b64b48b44eb4aa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..718ddbf9f4d40aa67d52dee0bf33e6b4f2324c4c
--- /dev/null
+++ b/records/sd1/sweep/75fe25b64b48b44eb4aa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "75fe25b64b48b44eb4aa",
+ "values": {
+  "std": {
+   "value": 37.04,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01478,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.838,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.400265,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.494185,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.53701,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.142506,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8782,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.192,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75fe25b64b48b44eb4aa/record.json b/records/sd1/sweep/75fe25b64b48b44eb4aa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..13a66da19c9bb39b082aa53b60f32c1dbc39d7ef
--- /dev/null
+++ b/records/sd1/sweep/75fe25b64b48b44eb4aa/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "75fe25b64b48b44eb4aa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:23:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "zero",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "multiply",
+   0.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "65e1630076f095ed8e4f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/75fe25b64b48b44eb4aa/workflow.json b/records/sd1/sweep/75fe25b64b48b44eb4aa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe7279cb8aa26d39362f7eaa5d476ba3d7e5e706
--- /dev/null
+++ b/records/sd1/sweep/75fe25b64b48b44eb4aa/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 75fe25b64b48b44eb4aa)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7607e37d955e05ab9a7f/measurements.json b/records/sd1/sweep/7607e37d955e05ab9a7f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..633975505ef0b59d3be74eacfcc83517778e02df
--- /dev/null
+++ b/records/sd1/sweep/7607e37d955e05ab9a7f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7607e37d955e05ab9a7f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.285189,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.323992,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.129075,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.059198,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.269,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00563,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.815,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9678,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0823,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7607e37d955e05ab9a7f/record.json b/records/sd1/sweep/7607e37d955e05ab9a7f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..27e6c7949e8a0d202a2bb0f0805ed0bac0df50a0
--- /dev/null
+++ b/records/sd1/sweep/7607e37d955e05ab9a7f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7607e37d955e05ab9a7f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.attn1.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/bcbe223db966f2fd48e5f2d17ebeb01cdaf5eae960d615409e2ceb5e0510fff0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "bcbe223db966f2fd48e5f2d17ebeb01cdaf5eae960d615409e2ceb5e0510fff0",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c27a9bbd1792eed1e1e1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7607e37d955e05ab9a7f/workflow.json b/records/sd1/sweep/7607e37d955e05ab9a7f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..23297cfd85e3b9b81f50699eb2b2f8b8b6f20a12
--- /dev/null
+++ b/records/sd1/sweep/7607e37d955e05ab9a7f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7607e37d955e05ab9a7f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.attn1.to_v\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760bda95ecd9ca137c50/interpretations.jsonl b/records/sd1/sweep/760bda95ecd9ca137c50/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ea46af028ca9321ebd54a336ac4ce64f38444465
--- /dev/null
+++ b/records/sd1/sweep/760bda95ecd9ca137c50/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "760bda95ecd9ca137c50"}, "kind": "caption", "value": "Bob-haired young woman, three-quarter view, black collar, muted grey-green frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:22:24Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3a802008747547c0"}
+{"target": {"record": "760bda95ecd9ca137c50"}, "kind": "change", "value": "Almost no visible change; only very slight shifts in facial micro-detail.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:22:24Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b85a57b37d1dc8b0"}
+{"target": {"record": "760bda95ecd9ca137c50"}, "kind": "keywords", "value": ["portrait", "no change", "muted tones", "photo"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:22:24Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "72f15337cab1223f"}
+{"target": {"record": "760bda95ecd9ca137c50"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:22:24Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e4841463a661400e"}
diff --git a/records/sd1/sweep/760bda95ecd9ca137c50/measurements.json b/records/sd1/sweep/760bda95ecd9ca137c50/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..647d1a2bea160be9792785f395ae699b250467bc
--- /dev/null
+++ b/records/sd1/sweep/760bda95ecd9ca137c50/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "760bda95ecd9ca137c50",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760bda95ecd9ca137c50/record.json b/records/sd1/sweep/760bda95ecd9ca137c50/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a969cd799406b6032066099344edd6950135306
--- /dev/null
+++ b/records/sd1/sweep/760bda95ecd9ca137c50/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "760bda95ecd9ca137c50",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5605f6841586b9b45322f158c1cd99be1df8f66fe828ff1ff7f12cb5e39e6e64.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5605f6841586b9b45322f158c1cd99be1df8f66fe828ff1ff7f12cb5e39e6e64",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "29ebb74bf2a732a1588d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760bda95ecd9ca137c50/workflow.json b/records/sd1/sweep/760bda95ecd9ca137c50/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..22feb24b0606c722f54228167a7665dd99dfa2f2
--- /dev/null
+++ b/records/sd1/sweep/760bda95ecd9ca137c50/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 760bda95ecd9ca137c50)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.h_upd\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760d7c0efc493f771daa/measurements.json b/records/sd1/sweep/760d7c0efc493f771daa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dab8aea6772eab0e24dd142378c16b797f74a209
--- /dev/null
+++ b/records/sd1/sweep/760d7c0efc493f771daa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "760d7c0efc493f771daa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.17463,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.240837,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.119714,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.061293,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.825,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01659,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.87,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9938,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0589,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760d7c0efc493f771daa/record.json b/records/sd1/sweep/760d7c0efc493f771daa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..20f82371965bad78de6bb9fe1109d1aad238fc2d
--- /dev/null
+++ b/records/sd1/sweep/760d7c0efc493f771daa/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "760d7c0efc493f771daa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/daeb1eca6bc1490584920dbcae32e75ff2bc031f47672f1ef55e2270343126a7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "daeb1eca6bc1490584920dbcae32e75ff2bc031f47672f1ef55e2270343126a7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "975bae8bca5b0e8971b4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760d7c0efc493f771daa/workflow.json b/records/sd1/sweep/760d7c0efc493f771daa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..05eabd5468e5616fe8521fec6fe5176118cfe79c
--- /dev/null
+++ b/records/sd1/sweep/760d7c0efc493f771daa/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 760d7c0efc493f771daa)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.out_layers.3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760f3d1c9d55514d8160/measurements.json b/records/sd1/sweep/760f3d1c9d55514d8160/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c950562293dba220f606e44673e6e67692f2789d
--- /dev/null
+++ b/records/sd1/sweep/760f3d1c9d55514d8160/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "760f3d1c9d55514d8160",
+ "values": {
+  "std": {
+   "value": 51.152,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0164,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.242,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.359748,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.403802,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.058466,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.038441,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0485,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2164,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760f3d1c9d55514d8160/record.json b/records/sd1/sweep/760f3d1c9d55514d8160/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..736793b3cd278d613607f534940cf1124a60ffae
--- /dev/null
+++ b/records/sd1/sweep/760f3d1c9d55514d8160/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "760f3d1c9d55514d8160",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:26:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "grow",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "scale",
+   1.4,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2e7dfaf93a6ee40cf595"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760f3d1c9d55514d8160/workflow.json b/records/sd1/sweep/760f3d1c9d55514d8160/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0777bd737ba13576edf0e80bc229603e4477350
--- /dev/null
+++ b/records/sd1/sweep/760f3d1c9d55514d8160/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 760f3d1c9d55514d8160)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760f67a43f188d356a1e/measurements.json b/records/sd1/sweep/760f67a43f188d356a1e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0fc22c71cc55c57b1d4094f7adbfe7f1003a424
--- /dev/null
+++ b/records/sd1/sweep/760f67a43f188d356a1e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "760f67a43f188d356a1e",
+ "values": {
+  "std": {
+   "value": 24.903,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.023,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.332,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.295494,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.565995,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.282884,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.099825,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9924,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1137,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760f67a43f188d356a1e/record.json b/records/sd1/sweep/760f67a43f188d356a1e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6745c1290517a2f43b41849686ae8998fa553df
--- /dev/null
+++ b/records/sd1/sweep/760f67a43f188d356a1e/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "760f67a43f188d356a1e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:16:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "quarter",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "rotate",
+   90.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [5.0, 45.0] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "63e8ef785a923d0d679c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/760f67a43f188d356a1e/workflow.json b/records/sd1/sweep/760f67a43f188d356a1e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b33fd479561e5d97a0cd0c70e7f4fc72875da4e0
--- /dev/null
+++ b/records/sd1/sweep/760f67a43f188d356a1e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 760f67a43f188d356a1e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/761022169d673379fdbd/measurements.json b/records/sd1/sweep/761022169d673379fdbd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a2961f3fd6bf565326eb3893185f85b5dbbec5d
--- /dev/null
+++ b/records/sd1/sweep/761022169d673379fdbd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "761022169d673379fdbd",
+ "values": {
+  "std": {
+   "value": 37.623,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01424,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.818,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.306723,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.460252,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.136484,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.080289,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9272,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1143,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/761022169d673379fdbd/record.json b/records/sd1/sweep/761022169d673379fdbd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..27802f05ac397b9252d374d3db60f980a2b13b1b
--- /dev/null
+++ b/records/sd1/sweep/761022169d673379fdbd/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "761022169d673379fdbd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:09:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/761022169d673379fdbd/workflow.json b/records/sd1/sweep/761022169d673379fdbd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..71d2f6dc406db292bd3750904fe3a81e2b0c0782
--- /dev/null
+++ b/records/sd1/sweep/761022169d673379fdbd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 761022169d673379fdbd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7614eeafca06dd585cc5/measurements.json b/records/sd1/sweep/7614eeafca06dd585cc5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c3431f73fc1cdeac776f83fb54e7be7c4b3b151
--- /dev/null
+++ b/records/sd1/sweep/7614eeafca06dd585cc5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7614eeafca06dd585cc5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.247796,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.295226,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.14209,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.039464,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.985,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00471,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.647,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9461,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1167,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7614eeafca06dd585cc5/record.json b/records/sd1/sweep/7614eeafca06dd585cc5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f543eafb7f13dc795d1241e99eb182a297ab1fc
--- /dev/null
+++ b/records/sd1/sweep/7614eeafca06dd585cc5/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7614eeafca06dd585cc5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.ff.net.0.proj",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "kind": "ff",
+   "module_type": "Linear",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/eeee485b1fff943743b42e2b3b6db50e36cda1962afb35d31992ff82aefc1ac1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "eeee485b1fff943743b42e2b3b6db50e36cda1962afb35d31992ff82aefc1ac1",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.GEGLU.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "68527b172dc4d3a749eb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7614eeafca06dd585cc5/workflow.json b/records/sd1/sweep/7614eeafca06dd585cc5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4a1158a190f4857c04d6e6f7b139b6b2369622a
--- /dev/null
+++ b/records/sd1/sweep/7614eeafca06dd585cc5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7614eeafca06dd585cc5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.transformer_blocks.0.ff.net.0.proj\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7615a556e5c9ac0aee04/measurements.json b/records/sd1/sweep/7615a556e5c9ac0aee04/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7becfae2c919cc35a3e189645b2f249df909a586
--- /dev/null
+++ b/records/sd1/sweep/7615a556e5c9ac0aee04/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7615a556e5c9ac0aee04",
+ "values": {
+  "cosine_distance": {
+   "value": 0.376823,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.65276,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.676872,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.334799,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.426,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00304,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 46.379,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0603,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3246,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7615a556e5c9ac0aee04/record.json b/records/sd1/sweep/7615a556e5c9ac0aee04/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a47ac709d39267e23b56cf94d00f2d0cf01d2abb
--- /dev/null
+++ b/records/sd1/sweep/7615a556e5c9ac0aee04/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7615a556e5c9ac0aee04",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/db128aff22c8477eb911890bad144ba63401d7e96efb086db526479f818bb168.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "db128aff22c8477eb911890bad144ba63401d7e96efb086db526479f818bb168",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "8df791eae4009a369797"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7615a556e5c9ac0aee04/workflow.json b/records/sd1/sweep/7615a556e5c9ac0aee04/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..07332186a16c8f98aad95b5eae5aca4b00419013
--- /dev/null
+++ b/records/sd1/sweep/7615a556e5c9ac0aee04/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7615a556e5c9ac0aee04)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76263c0ebfa26080e9bd/interpretations.jsonl b/records/sd1/sweep/76263c0ebfa26080e9bd/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..88227b0a27bbc464aeaa560316e689eaef7400ac
--- /dev/null
+++ b/records/sd1/sweep/76263c0ebfa26080e9bd/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "76263c0ebfa26080e9bd"}, "kind": "caption", "value": "Short-haired woman in black top, three-quarter view, grey framed studio backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ad9d83ab9d1580a0"}
+{"target": {"record": "76263c0ebfa26080e9bd"}, "kind": "change", "value": "Almost no visible change, just slightly redder lips and a tiny tonal shift.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1978bfc0c475aef7"}
+{"target": {"record": "76263c0ebfa26080e9bd"}, "kind": "keywords", "value": ["near identical", "grey studio portrait", "subtle change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b0ff01ab3cce8d91"}
+{"target": {"record": "76263c0ebfa26080e9bd"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7ea7f517f67e39fe"}
diff --git a/records/sd1/sweep/76263c0ebfa26080e9bd/measurements.json b/records/sd1/sweep/76263c0ebfa26080e9bd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..da8a1703a79b718b5de25aedfad3b392d9d7f473
--- /dev/null
+++ b/records/sd1/sweep/76263c0ebfa26080e9bd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76263c0ebfa26080e9bd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76263c0ebfa26080e9bd/record.json b/records/sd1/sweep/76263c0ebfa26080e9bd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..184689b9bcbd822c573a16361c9a8fcc466e0990
--- /dev/null
+++ b/records/sd1/sweep/76263c0ebfa26080e9bd/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "76263c0ebfa26080e9bd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f42c8301378d93877b074a50d1c6cebcc231f0d3521fbc20c38faca9ab7c6850.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f42c8301378d93877b074a50d1c6cebcc231f0d3521fbc20c38faca9ab7c6850",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "8d91f4831bef1f58c1a1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76263c0ebfa26080e9bd/workflow.json b/records/sd1/sweep/76263c0ebfa26080e9bd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4949b0378ca72c9cc6b08c28c335df974ae9a6d4
--- /dev/null
+++ b/records/sd1/sweep/76263c0ebfa26080e9bd/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76263c0ebfa26080e9bd)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/763681832d5629b5b429/interpretations.jsonl b/records/sd1/sweep/763681832d5629b5b429/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..40180efeb3cbd9db00e773a3c7a16cfd8d45831d
--- /dev/null
+++ b/records/sd1/sweep/763681832d5629b5b429/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "763681832d5629b5b429"}, "kind": "caption", "value": "Neon street dominated by red signs and red reflections on wet pavement", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:10Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b60420e82aac4dfd"}
+{"target": {"record": "763681832d5629b5b429"}, "kind": "change", "value": "The palette shifts strongly to red and pink and the signs and buildings are rearranged, but it is still clearly a neon street seen down its length.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:10Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "07c832cc4e2bafe8"}
+{"target": {"record": "763681832d5629b5b429"}, "kind": "keywords", "value": ["red neon street", "colour shift red", "night city", "redrawn signs", "wet reflections"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:10Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "120716ec21b20f4c"}
+{"target": {"record": "763681832d5629b5b429"}, "kind": "effect_tags", "value": ["colour_shift", "saturated", "subject_kept", "new_composition"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:10Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c2aefad7c0897a60"}
diff --git a/records/sd1/sweep/763681832d5629b5b429/measurements.json b/records/sd1/sweep/763681832d5629b5b429/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0c24af5788f029c5ac4c65c29fd2c0ecbb8435f
--- /dev/null
+++ b/records/sd1/sweep/763681832d5629b5b429/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "763681832d5629b5b429",
+ "values": {
+  "std": {
+   "value": 37.088,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01524,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.342,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.15745,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.309996,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.143343,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.030922,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0105,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2771,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/763681832d5629b5b429/record.json b/records/sd1/sweep/763681832d5629b5b429/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..83849837560e7e0503e28a19d269ba8a4ea5e5b5
--- /dev/null
+++ b/records/sd1/sweep/763681832d5629b5b429/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "763681832d5629b5b429",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:24:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   2.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "d5df3ee3ec2639885735"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/763681832d5629b5b429/workflow.json b/records/sd1/sweep/763681832d5629b5b429/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c8d24f64d98ad92f44e04eb6a4edf7f75b3f18a
--- /dev/null
+++ b/records/sd1/sweep/763681832d5629b5b429/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 763681832d5629b5b429)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76421b672b13f1812ea6/measurements.json b/records/sd1/sweep/76421b672b13f1812ea6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..add1d5fd09972ca487d06f1e559a5f30e05fd1a5
--- /dev/null
+++ b/records/sd1/sweep/76421b672b13f1812ea6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76421b672b13f1812ea6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.197105,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.256522,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.137981,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027986,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.376,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00616,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.4,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9892,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0962,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76421b672b13f1812ea6/record.json b/records/sd1/sweep/76421b672b13f1812ea6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4afa50011d9da5e2a359419eba5e4534ad3a0011
--- /dev/null
+++ b/records/sd1/sweep/76421b672b13f1812ea6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76421b672b13f1812ea6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/a2618f1e208ab251bf7d27e1a906467bf1397edae4fd39e1d041febd05e51ff9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a2618f1e208ab251bf7d27e1a906467bf1397edae4fd39e1d041febd05e51ff9",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "9aa7162152970040e645"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76421b672b13f1812ea6/workflow.json b/records/sd1/sweep/76421b672b13f1812ea6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..81b63ab5a1e6dc7c517bc6577731a599b4188d40
--- /dev/null
+++ b/records/sd1/sweep/76421b672b13f1812ea6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76421b672b13f1812ea6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76465717f0d7d62cfa8f/measurements.json b/records/sd1/sweep/76465717f0d7d62cfa8f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee543b80af7732eac671a64f60b8bb43095566aa
--- /dev/null
+++ b/records/sd1/sweep/76465717f0d7d62cfa8f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76465717f0d7d62cfa8f",
+ "values": {
+  "std": {
+   "value": 49.856,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.016,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.048,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.279768,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.400439,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.436247,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.089329,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9876,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0351,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76465717f0d7d62cfa8f/record.json b/records/sd1/sweep/76465717f0d7d62cfa8f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..941a4e246bc6a3a0032e0fe20f95ca487cb22cc9
--- /dev/null
+++ b/records/sd1/sweep/76465717f0d7d62cfa8f/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "76465717f0d7d62cfa8f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:01:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "mid",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_scalar",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76465717f0d7d62cfa8f/workflow.json b/records/sd1/sweep/76465717f0d7d62cfa8f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d92e4ef703cbd7f19d6d88aabc421e05356ae819
--- /dev/null
+++ b/records/sd1/sweep/76465717f0d7d62cfa8f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76465717f0d7d62cfa8f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7649464d5a645064e65e/measurements.json b/records/sd1/sweep/7649464d5a645064e65e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..965cd720740dacce8c3dcc36bbe8feb79ae12c1c
--- /dev/null
+++ b/records/sd1/sweep/7649464d5a645064e65e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7649464d5a645064e65e",
+ "values": {
+  "std": {
+   "value": 50.268,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01465,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 6.16,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.035849,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.065698,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.040935,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.015846,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9925,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1039,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7649464d5a645064e65e/record.json b/records/sd1/sweep/7649464d5a645064e65e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a74b5e3d04436f9ff66bb3ba12ee6c7beea540c
--- /dev/null
+++ b/records/sd1/sweep/7649464d5a645064e65e/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "7649464d5a645064e65e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:20:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "mid",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_scalar",
+   -0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7649464d5a645064e65e/workflow.json b/records/sd1/sweep/7649464d5a645064e65e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a286345f11886a81ea0712e855adb768faf56674
--- /dev/null
+++ b/records/sd1/sweep/7649464d5a645064e65e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7649464d5a645064e65e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/764f9a0e4cadbd6cf26d/measurements.json b/records/sd1/sweep/764f9a0e4cadbd6cf26d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1f4c1b8a6588bb389a8a1ded743d361ff5a29d1
--- /dev/null
+++ b/records/sd1/sweep/764f9a0e4cadbd6cf26d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "764f9a0e4cadbd6cf26d",
+ "values": {
+  "std": {
+   "value": 43.035,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01642,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 6.425,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013564,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.059201,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.027853,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.020174,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0109,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0186,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/764f9a0e4cadbd6cf26d/record.json b/records/sd1/sweep/764f9a0e4cadbd6cf26d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..13ce218cfe1573d01644e52032deeda6f78d3a62
--- /dev/null
+++ b/records/sd1/sweep/764f9a0e4cadbd6cf26d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "764f9a0e4cadbd6cf26d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:56:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8bab3bb8d906fa80ce18"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/764f9a0e4cadbd6cf26d/workflow.json b/records/sd1/sweep/764f9a0e4cadbd6cf26d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0bf62a42437dbfac27424227cce20420dcc2787
--- /dev/null
+++ b/records/sd1/sweep/764f9a0e4cadbd6cf26d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 764f9a0e4cadbd6cf26d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7650c0a8d541b0010070/interpretations.jsonl b/records/sd1/sweep/7650c0a8d541b0010070/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..0b63949058b8cc08a8b1062f36482dd655459c06
--- /dev/null
+++ b/records/sd1/sweep/7650c0a8d541b0010070/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7650c0a8d541b0010070"}, "kind": "caption", "value": "Close-up of glass bowl with strawberries and yellow fruit beside a glass carafe", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2d14dacb1386c48c"}
+{"target": {"record": "7650c0a8d541b0010070"}, "kind": "change", "value": "The camera zooms in dramatically on a glass dish of fruit and a partial carafe, dropping the wider table arrangement, with a darker, more photographic close-up feel.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "33927bbb06d63cf6"}
+{"target": {"record": "7650c0a8d541b0010070"}, "kind": "keywords", "value": ["close-up still life", "glass bowl", "strawberries", "zoomed in", "carafe"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2c155a127ad590ab"}
+{"target": {"record": "7650c0a8d541b0010070"}, "kind": "effect_tags", "value": ["zoom", "new_composition", "framing", "subject_kept", "dark"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5600ab6a6ec40330"}
diff --git a/records/sd1/sweep/7650c0a8d541b0010070/measurements.json b/records/sd1/sweep/7650c0a8d541b0010070/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4b730d6edee55cb3265bc80e92d5b3d436dd765
--- /dev/null
+++ b/records/sd1/sweep/7650c0a8d541b0010070/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7650c0a8d541b0010070",
+ "values": {
+  "std": {
+   "value": 40.458,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00941,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.186,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.439889,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.646457,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.326793,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.175016,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9302,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1567,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7650c0a8d541b0010070/record.json b/records/sd1/sweep/7650c0a8d541b0010070/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcf61b0f8ebdbd8a3d34f7dbf1b87746550eca56
--- /dev/null
+++ b/records/sd1/sweep/7650c0a8d541b0010070/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "7650c0a8d541b0010070",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:37:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "scale",
+   1.4,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2d430f249e59f7bde007"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7650c0a8d541b0010070/workflow.json b/records/sd1/sweep/7650c0a8d541b0010070/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ac31a23664adfd5bf3b20aae2ece2c0b46d658e
--- /dev/null
+++ b/records/sd1/sweep/7650c0a8d541b0010070/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7650c0a8d541b0010070)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/766a2f066556da8c0b94/measurements.json b/records/sd1/sweep/766a2f066556da8c0b94/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..abfc248b3e03fedd6b738864bb50d72fd43e499c
--- /dev/null
+++ b/records/sd1/sweep/766a2f066556da8c0b94/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "766a2f066556da8c0b94",
+ "values": {
+  "std": {
+   "value": 30.264,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02331,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.715,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.319649,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.591934,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.33584,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.119747,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.978,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3559,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/766a2f066556da8c0b94/record.json b/records/sd1/sweep/766a2f066556da8c0b94/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..427471b6cf090cccb4b687da418d02fc668f63ea
--- /dev/null
+++ b/records/sd1/sweep/766a2f066556da8c0b94/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "766a2f066556da8c0b94",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:16:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.7, 1.3] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/766a2f066556da8c0b94/workflow.json b/records/sd1/sweep/766a2f066556da8c0b94/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ea79a217be9a7ea46c918f67890aa88d06d71a0
--- /dev/null
+++ b/records/sd1/sweep/766a2f066556da8c0b94/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 766a2f066556da8c0b94)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/766c5d31a465fa3f5263/measurements.json b/records/sd1/sweep/766c5d31a465fa3f5263/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a1f522bac3172e9447f655d80575f6f0d894e1e
--- /dev/null
+++ b/records/sd1/sweep/766c5d31a465fa3f5263/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "766c5d31a465fa3f5263",
+ "values": {
+  "cosine_distance": {
+   "value": 0.503599,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.494492,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.282103,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.155295,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.439,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00711,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 47.008,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0705,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3579,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/766c5d31a465fa3f5263/record.json b/records/sd1/sweep/766c5d31a465fa3f5263/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f91f0b0575b6ab4ff5b30880d6cac7a83de976a
--- /dev/null
+++ b/records/sd1/sweep/766c5d31a465fa3f5263/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "766c5d31a465fa3f5263",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.1.norm",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/bac29f94ab9c2c2d932ae40cacad3deea32938f2f1c6753c0bd83bf0e05cd313.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "bac29f94ab9c2c2d932ae40cacad3deea32938f2f1c6753c0bd83bf0e05cd313",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "0a66ca18e3c0f34068cc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/766c5d31a465fa3f5263/workflow.json b/records/sd1/sweep/766c5d31a465fa3f5263/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1408656d35370bda707c896f972c0e634b9154e4
--- /dev/null
+++ b/records/sd1/sweep/766c5d31a465fa3f5263/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 766c5d31a465fa3f5263)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.norm\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/767437988df570d061d0/measurements.json b/records/sd1/sweep/767437988df570d061d0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd374af7a81aa74c639ef2aa5083c399705d6b93
--- /dev/null
+++ b/records/sd1/sweep/767437988df570d061d0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "767437988df570d061d0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.083799,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.127352,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.075613,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025711,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.899,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00556,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.611,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0448,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1043,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/767437988df570d061d0/record.json b/records/sd1/sweep/767437988df570d061d0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4db5337d04a6910f0b6a19ff5822f40139a6affe
--- /dev/null
+++ b/records/sd1/sweep/767437988df570d061d0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "767437988df570d061d0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.hi",
+   "kind": "ff",
+   "module_type": "Linear",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/3b357e9ebde666155ce100f273f00b731d70c0c2fb74b9f33c2d76820f3f9cde.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3b357e9ebde666155ce100f273f00b731d70c0c2fb74b9f33c2d76820f3f9cde",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "906b2f87819403ad7d96"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/767437988df570d061d0/workflow.json b/records/sd1/sweep/767437988df570d061d0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..90e10ab9e83eeeecb6cf163935c09dac23cb148f
--- /dev/null
+++ b/records/sd1/sweep/767437988df570d061d0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 767437988df570d061d0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.transformer_blocks.0.ff.net.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7686a1e354e8d12ee8e1/interpretations.jsonl b/records/sd1/sweep/7686a1e354e8d12ee8e1/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..721c3b5b3681c1b0b5a4704c14385a3c68f839d2
--- /dev/null
+++ b/records/sd1/sweep/7686a1e354e8d12ee8e1/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7686a1e354e8d12ee8e1"}, "kind": "caption", "value": "Narrow neon alley at night with orange shop signs, blue sky and walkers", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3dff08e1868ae247"}
+{"target": {"record": "7686a1e354e8d12ee8e1"}, "kind": "change", "value": "The signs and lettering are rearranged a little and the alley gets slightly warmer and more blurred at the edges, while the layout stays similar.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0cf09c0c806d0b7d"}
+{"target": {"record": "7686a1e354e8d12ee8e1"}, "kind": "keywords", "value": ["neon alley", "orange and blue", "night street", "minor change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a6d4db78e23aadb7"}
+{"target": {"record": "7686a1e354e8d12ee8e1"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "05501bdaf499b969"}
diff --git a/records/sd1/sweep/7686a1e354e8d12ee8e1/measurements.json b/records/sd1/sweep/7686a1e354e8d12ee8e1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc304454dbd8060723a9dad711d85a18122dd40f
--- /dev/null
+++ b/records/sd1/sweep/7686a1e354e8d12ee8e1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7686a1e354e8d12ee8e1",
+ "values": {
+  "std": {
+   "value": 25.919,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02248,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.188,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.073687,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.19867,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.118842,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.030234,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9839,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2035,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7686a1e354e8d12ee8e1/record.json b/records/sd1/sweep/7686a1e354e8d12ee8e1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec33edf31a4deccdc12544a55db57968f84497c1
--- /dev/null
+++ b/records/sd1/sweep/7686a1e354e8d12ee8e1/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7686a1e354e8d12ee8e1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:11:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "heavy",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_noise",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7686a1e354e8d12ee8e1/workflow.json b/records/sd1/sweep/7686a1e354e8d12ee8e1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3afc4968555904b187536c5be5cd17b2cb8397b8
--- /dev/null
+++ b/records/sd1/sweep/7686a1e354e8d12ee8e1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7686a1e354e8d12ee8e1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/768f9c8a0aaf6fb6e690/measurements.json b/records/sd1/sweep/768f9c8a0aaf6fb6e690/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a53257b7cf11c49ff73706d6e08f19432bd003a8
--- /dev/null
+++ b/records/sd1/sweep/768f9c8a0aaf6fb6e690/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "768f9c8a0aaf6fb6e690",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/768f9c8a0aaf6fb6e690/record.json b/records/sd1/sweep/768f9c8a0aaf6fb6e690/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..506325e4b24a4c9a0e7fc762b80df962fc71d3e9
--- /dev/null
+++ b/records/sd1/sweep/768f9c8a0aaf6fb6e690/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "768f9c8a0aaf6fb6e690",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/fb7a14b8973e132f7a3e821b0d04a54e3aa353bc52207de5d5113db7495d2ec6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "fb7a14b8973e132f7a3e821b0d04a54e3aa353bc52207de5d5113db7495d2ec6",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "e06f303a6b9d9cdb4445"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/768f9c8a0aaf6fb6e690/workflow.json b/records/sd1/sweep/768f9c8a0aaf6fb6e690/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..708bd7a358f51fa3f0bf46cf8488dc7c9e2873ce
--- /dev/null
+++ b/records/sd1/sweep/768f9c8a0aaf6fb6e690/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 768f9c8a0aaf6fb6e690)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/769fa3c9bc07097531e9/interpretations.jsonl b/records/sd1/sweep/769fa3c9bc07097531e9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..469d5b6a13235e17cbeaf6a1339f918b0962c014
--- /dev/null
+++ b/records/sd1/sweep/769fa3c9bc07097531e9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "769fa3c9bc07097531e9"}, "kind": "caption", "value": "Smeared portrait with a blurred, partly erased face surrounded by dark curly hair", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:34:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bdfc202865d98310"}
+{"target": {"record": "769fa3c9bc07097531e9"}, "kind": "change", "value": "The face dissolves into a pale smudged blur with missing features while the hair and sweater remain as loose painted masses, breaking down the portrait.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:34:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2c2e695b84d4929c"}
+{"target": {"record": "769fa3c9bc07097531e9"}, "kind": "keywords", "value": ["erased face", "smeared features", "blurred portrait", "dissolving subject", "painterly breakdown"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:34:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a1fa484a0548a1f2"}
+{"target": {"record": "769fa3c9bc07097531e9"}, "kind": "effect_tags", "value": ["blur", "distorted", "painterly", "degenerate", "washed_out", "surreal", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:34:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0900a8071c53c4a6"}
diff --git a/records/sd1/sweep/769fa3c9bc07097531e9/measurements.json b/records/sd1/sweep/769fa3c9bc07097531e9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9e9064e820246a1d2822ce9b174110e221c0039
--- /dev/null
+++ b/records/sd1/sweep/769fa3c9bc07097531e9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "769fa3c9bc07097531e9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.201959,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.419686,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.314125,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.151516,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.079,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00453,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.62,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9858,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3077,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/769fa3c9bc07097531e9/record.json b/records/sd1/sweep/769fa3c9bc07097531e9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6d418e83084fe9801a0a60916e114aa10f52877
--- /dev/null
+++ b/records/sd1/sweep/769fa3c9bc07097531e9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "769fa3c9bc07097531e9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:16Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/4414dc0a4b630408303ce900a493ce1778c9268b05348a5a9c3c504a184c1740.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4414dc0a4b630408303ce900a493ce1778c9268b05348a5a9c3c504a184c1740",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "58eb37588a2b87600a6e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/769fa3c9bc07097531e9/workflow.json b/records/sd1/sweep/769fa3c9bc07097531e9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d7d6d65cab15d7403f069c4b60a5717ac707d07
--- /dev/null
+++ b/records/sd1/sweep/769fa3c9bc07097531e9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 769fa3c9bc07097531e9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76a5ac11f615a02d85a1/measurements.json b/records/sd1/sweep/76a5ac11f615a02d85a1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec2277c416e3582e2a2d796cfabaa2bd4aa756f8
--- /dev/null
+++ b/records/sd1/sweep/76a5ac11f615a02d85a1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "76a5ac11f615a02d85a1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.011805,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.02416,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.005567,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001817,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.318,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00501,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.576,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0221,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1427,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76a5ac11f615a02d85a1/record.json b/records/sd1/sweep/76a5ac11f615a02d85a1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..df52ed69e9e5dce40bb7917b04b7bf3dbce8b438
--- /dev/null
+++ b/records/sd1/sweep/76a5ac11f615a02d85a1/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76a5ac11f615a02d85a1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/773f9622f9c6c3b4528e8e85fa9c2edb7ff9e4bc2295319cf46539b7edb19fab.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "773f9622f9c6c3b4528e8e85fa9c2edb7ff9e4bc2295319cf46539b7edb19fab",
+  "type_path": "UNetModel.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "aeeed6fe2f30cccefed4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76a5ac11f615a02d85a1/workflow.json b/records/sd1/sweep/76a5ac11f615a02d85a1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..98a45d84be5b4e418369a3d81351fe7388dfa935
--- /dev/null
+++ b/records/sd1/sweep/76a5ac11f615a02d85a1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76a5ac11f615a02d85a1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.norm2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c31564fe769336dc7f/measurements.json b/records/sd1/sweep/76c31564fe769336dc7f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a7cb70b97691c4c43b1e1bef2f8591a380e0f09
--- /dev/null
+++ b/records/sd1/sweep/76c31564fe769336dc7f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76c31564fe769336dc7f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.180806,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.273453,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.125621,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.041768,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.983,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00668,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.45,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9615,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0514,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c31564fe769336dc7f/record.json b/records/sd1/sweep/76c31564fe769336dc7f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ad2f3e1fe03b770658af0af30cd08b86b19a23e
--- /dev/null
+++ b/records/sd1/sweep/76c31564fe769336dc7f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76c31564fe769336dc7f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:16Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/4714cea64d4c3e5fc57f9c42ad9a129e697ce984f6cf9ce9cdf5556e0c4d4d6c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4714cea64d4c3e5fc57f9c42ad9a129e697ce984f6cf9ce9cdf5556e0c4d4d6c",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f0ce4478f8b5159130f7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c31564fe769336dc7f/workflow.json b/records/sd1/sweep/76c31564fe769336dc7f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..469ef155ba2c0fc21075008983c3ff2c76a040e5
--- /dev/null
+++ b/records/sd1/sweep/76c31564fe769336dc7f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76c31564fe769336dc7f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c6f8813df950fc44a9/measurements.json b/records/sd1/sweep/76c6f8813df950fc44a9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..43250ed08e483f1779b393665ff0c4dbe12be0ca
--- /dev/null
+++ b/records/sd1/sweep/76c6f8813df950fc44a9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76c6f8813df950fc44a9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.032968,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.056281,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.017562,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.011364,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.647,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00646,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.99,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.019,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0947,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c6f8813df950fc44a9/record.json b/records/sd1/sweep/76c6f8813df950fc44a9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..648c76054a358d2ed409492cac3fec00949e7efb
--- /dev/null
+++ b/records/sd1/sweep/76c6f8813df950fc44a9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76c6f8813df950fc44a9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/8611261da310d59a37ea4273f16c03f428c79f95581614b5cef9275024d926a0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8611261da310d59a37ea4273f16c03f428c79f95581614b5cef9275024d926a0",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "70069b039e147a295411"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c6f8813df950fc44a9/workflow.json b/records/sd1/sweep/76c6f8813df950fc44a9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9088f84b853097c1ac81c6020a6f4e49d8652b63
--- /dev/null
+++ b/records/sd1/sweep/76c6f8813df950fc44a9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76c6f8813df950fc44a9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c96b2115238f16e764/measurements.json b/records/sd1/sweep/76c96b2115238f16e764/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..eead4c2dac26b77cc57c9e6379b3d5ee0349a0cd
--- /dev/null
+++ b/records/sd1/sweep/76c96b2115238f16e764/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "76c96b2115238f16e764",
+ "values": {
+  "cosine_distance": {
+   "value": 0.260594,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.33204,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.245167,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.096941,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.549,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00528,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 32.219,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9861,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0221,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c96b2115238f16e764/record.json b/records/sd1/sweep/76c96b2115238f16e764/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f39773bc25f42f35ea710a470406f4543c2c8e16
--- /dev/null
+++ b/records/sd1/sweep/76c96b2115238f16e764/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76c96b2115238f16e764",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e9b06b933c638537a4bb9f2865bee3af32044c9c71d7ceb6734f61afcfc379a4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e9b06b933c638537a4bb9f2865bee3af32044c9c71d7ceb6734f61afcfc379a4",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0758c90e1b675db24eb5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76c96b2115238f16e764/workflow.json b/records/sd1/sweep/76c96b2115238f16e764/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b84075485124f4cf46d25ee5dfe1aed212e94330
--- /dev/null
+++ b/records/sd1/sweep/76c96b2115238f16e764/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76c96b2115238f16e764)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ccabcc92e22173932a/measurements.json b/records/sd1/sweep/76ccabcc92e22173932a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8dbca2e00f6e6045d1217d9fe661614e4ea9ecbd
--- /dev/null
+++ b/records/sd1/sweep/76ccabcc92e22173932a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76ccabcc92e22173932a",
+ "values": {
+  "std": {
+   "value": 48.838,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01326,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.171,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.093762,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.240107,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.435068,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.085702,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0461,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0165,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ccabcc92e22173932a/record.json b/records/sd1/sweep/76ccabcc92e22173932a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bb0dee4afa1b719089baef66f59e25138f2ae1d
--- /dev/null
+++ b/records/sd1/sweep/76ccabcc92e22173932a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "76ccabcc92e22173932a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:19Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "quarter",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "59e4aee8d2b80f2b7c61"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ccabcc92e22173932a/workflow.json b/records/sd1/sweep/76ccabcc92e22173932a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d769cdf2c296d0faef12baa3cc7f42b82c11994b
--- /dev/null
+++ b/records/sd1/sweep/76ccabcc92e22173932a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76ccabcc92e22173932a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ce621ac63011ab87a0/measurements.json b/records/sd1/sweep/76ce621ac63011ab87a0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5964fcb37af3f46b42141ea9d4fab75a8d720330
--- /dev/null
+++ b/records/sd1/sweep/76ce621ac63011ab87a0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76ce621ac63011ab87a0",
+ "values": {
+  "std": {
+   "value": 16.805,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03189,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.133,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.27433,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.571027,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.292331,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.084086,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0026,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3744,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ce621ac63011ab87a0/record.json b/records/sd1/sweep/76ce621ac63011ab87a0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..60b66ef09c36636b2ff72460fac80db8ff5e8032
--- /dev/null
+++ b/records/sd1/sweep/76ce621ac63011ab87a0/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "76ce621ac63011ab87a0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:54:50Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.7, 1.3] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ce621ac63011ab87a0/workflow.json b/records/sd1/sweep/76ce621ac63011ab87a0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f716b8a3a26873f6567dae126f53238922344880
--- /dev/null
+++ b/records/sd1/sweep/76ce621ac63011ab87a0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76ce621ac63011ab87a0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76e5311849a8ead02a49/measurements.json b/records/sd1/sweep/76e5311849a8ead02a49/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb116f9c8a2c701ac4c2960c2bc9f6f4b6655413
--- /dev/null
+++ b/records/sd1/sweep/76e5311849a8ead02a49/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76e5311849a8ead02a49",
+ "values": {
+  "cosine_distance": {
+   "value": 0.067295,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.169411,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.063887,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.034694,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 38.584,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0032,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.747,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0336,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0641,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76e5311849a8ead02a49/record.json b/records/sd1/sweep/76e5311849a8ead02a49/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b8a516d05ef96c0e72053a71039f525ffb7688f
--- /dev/null
+++ b/records/sd1/sweep/76e5311849a8ead02a49/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76e5311849a8ead02a49",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/df32ea64d3c63604504d5bae3fc445267ece1889f36f8b552bb6710bdf49054f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "df32ea64d3c63604504d5bae3fc445267ece1889f36f8b552bb6710bdf49054f",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "27ae699c55d2180eeaca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76e5311849a8ead02a49/workflow.json b/records/sd1/sweep/76e5311849a8ead02a49/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0236b67b64ef9ddcf1bc7884d179e60b2f21c03b
--- /dev/null
+++ b/records/sd1/sweep/76e5311849a8ead02a49/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76e5311849a8ead02a49)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.norm1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76e9a7eee645551f0df1/measurements.json b/records/sd1/sweep/76e9a7eee645551f0df1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ce166adcd8342aa53b203baa804662ef7b614a0
--- /dev/null
+++ b/records/sd1/sweep/76e9a7eee645551f0df1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76e9a7eee645551f0df1",
+ "values": {
+  "std": {
+   "value": 36.472,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00709,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 39.914,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.435278,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.656506,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.662027,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.291263,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9659,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0245,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76e9a7eee645551f0df1/record.json b/records/sd1/sweep/76e9a7eee645551f0df1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8437e1c9384254ca01c25cfc6d23f6ff8cd10afa
--- /dev/null
+++ b/records/sd1/sweep/76e9a7eee645551f0df1/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "76e9a7eee645551f0df1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:45:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   30.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f6b92adf190c45174b5e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76e9a7eee645551f0df1/workflow.json b/records/sd1/sweep/76e9a7eee645551f0df1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..515a496a6bd55365a30018b17cae49298e2c7ee5
--- /dev/null
+++ b/records/sd1/sweep/76e9a7eee645551f0df1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76e9a7eee645551f0df1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ed281908a353f7860b/measurements.json b/records/sd1/sweep/76ed281908a353f7860b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f017689e162c9196728acae6d683f72c91f610ef
--- /dev/null
+++ b/records/sd1/sweep/76ed281908a353f7860b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76ed281908a353f7860b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.419343,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.467168,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.254976,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.117976,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 77.463,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00784,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 56.458,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0847,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1319,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ed281908a353f7860b/record.json b/records/sd1/sweep/76ed281908a353f7860b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..348e908a409b28196619bc08ab744eba3b1d7146
--- /dev/null
+++ b/records/sd1/sweep/76ed281908a353f7860b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "76ed281908a353f7860b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.out_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/53db329b0d16e63ced62405262e8377307520c2f094a18587b5cc417360162cb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "53db329b0d16e63ced62405262e8377307520c2f094a18587b5cc417360162cb",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c6406802dc5c46251b14"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76ed281908a353f7860b/workflow.json b/records/sd1/sweep/76ed281908a353f7860b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e35f5c36145f5406228aa8607a5668f8bc3218d7
--- /dev/null
+++ b/records/sd1/sweep/76ed281908a353f7860b/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76ed281908a353f7860b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76f06d3d3896548e1c36/measurements.json b/records/sd1/sweep/76f06d3d3896548e1c36/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfaa81e8020df1764a39654f364bc898ff481bbb
--- /dev/null
+++ b/records/sd1/sweep/76f06d3d3896548e1c36/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76f06d3d3896548e1c36",
+ "values": {
+  "std": {
+   "value": 57.952,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00981,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.582,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.111146,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.320623,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.481691,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.171815,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.944,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3243,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76f06d3d3896548e1c36/record.json b/records/sd1/sweep/76f06d3d3896548e1c36/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2b5869604e477571789e3f8fe51dac4a91a73a5
--- /dev/null
+++ b/records/sd1/sweep/76f06d3d3896548e1c36/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "76f06d3d3896548e1c36",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:35:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "quarter",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [5.0, 45.0] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "3afe2798ce31f753c857"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76f06d3d3896548e1c36/workflow.json b/records/sd1/sweep/76f06d3d3896548e1c36/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0abd0f41e0968803e3b1889966d71ebef4ef47ec
--- /dev/null
+++ b/records/sd1/sweep/76f06d3d3896548e1c36/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76f06d3d3896548e1c36)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76f09b12c08703304709/measurements.json b/records/sd1/sweep/76f09b12c08703304709/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fad9015646caf8f159e763596cc04abb9619c6d4
--- /dev/null
+++ b/records/sd1/sweep/76f09b12c08703304709/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "76f09b12c08703304709",
+ "values": {
+  "cosine_distance": {
+   "value": 0.40015,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.37246,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.269037,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.17731,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 43.439,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0047,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.218,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9617,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0318,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76f09b12c08703304709/record.json b/records/sd1/sweep/76f09b12c08703304709/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8660d2b374ab30daa9e38af624d80d2c960f1ff4
--- /dev/null
+++ b/records/sd1/sweep/76f09b12c08703304709/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "76f09b12c08703304709",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/2b7725ce4afd2bb77b67e27d3d2fc988bc545c2efbe521ab87ff44e4370529ad.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2b7725ce4afd2bb77b67e27d3d2fc988bc545c2efbe521ab87ff44e4370529ad",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "a6b098a4ee6f125b989e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/76f09b12c08703304709/workflow.json b/records/sd1/sweep/76f09b12c08703304709/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..233b7daa050301b42e71e5de8fee1a0d475b36b1
--- /dev/null
+++ b/records/sd1/sweep/76f09b12c08703304709/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 76f09b12c08703304709)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7704ed862c810e1b9f5d/measurements.json b/records/sd1/sweep/7704ed862c810e1b9f5d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08bc38b8508843a82e18eb98b74667d0c64ee779
--- /dev/null
+++ b/records/sd1/sweep/7704ed862c810e1b9f5d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7704ed862c810e1b9f5d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.194591,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.27362,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.122142,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040626,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.279,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0053,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.942,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9614,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1784,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7704ed862c810e1b9f5d/record.json b/records/sd1/sweep/7704ed862c810e1b9f5d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b2cb604cea2bd8e7f31d54407efd656e396004a
--- /dev/null
+++ b/records/sd1/sweep/7704ed862c810e1b9f5d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7704ed862c810e1b9f5d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5078b045575ee684f44ede0df4dc0a052929e5559b51f639a1f6d68c6e1d4ce9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5078b045575ee684f44ede0df4dc0a052929e5559b51f639a1f6d68c6e1d4ce9",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "717898e2146248d72529"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7704ed862c810e1b9f5d/workflow.json b/records/sd1/sweep/7704ed862c810e1b9f5d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..424b10a3fe44230ac2552955a969962e4263b9e0
--- /dev/null
+++ b/records/sd1/sweep/7704ed862c810e1b9f5d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7704ed862c810e1b9f5d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7711d93b53cd2103d978/measurements.json b/records/sd1/sweep/7711d93b53cd2103d978/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6135ea2c2e2c9218de3850efd3974c4fc1109eac
--- /dev/null
+++ b/records/sd1/sweep/7711d93b53cd2103d978/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7711d93b53cd2103d978",
+ "values": {
+  "cosine_distance": {
+   "value": 0.083413,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.148675,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.054861,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025314,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.878,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00502,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.112,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0198,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1141,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7711d93b53cd2103d978/record.json b/records/sd1/sweep/7711d93b53cd2103d978/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d512b38acdc804c0d8b34558a8e46765bab295d
--- /dev/null
+++ b/records/sd1/sweep/7711d93b53cd2103d978/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7711d93b53cd2103d978",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.transformer_blocks.0.ff.net.0.proj",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.mid",
+   "kind": "ff",
+   "module_type": "Linear",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/748936ae39406ad28d83a3e4c0248b46fd21215c738de8d25f6a200e25e03ba3.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "748936ae39406ad28d83a3e4c0248b46fd21215c738de8d25f6a200e25e03ba3",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.GEGLU.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e37ef1a0ac60e0923b07"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7711d93b53cd2103d978/workflow.json b/records/sd1/sweep/7711d93b53cd2103d978/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..036c4aa8f94ddf0302018e01f53ae718cd386248
--- /dev/null
+++ b/records/sd1/sweep/7711d93b53cd2103d978/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7711d93b53cd2103d978)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.ff.net.0.proj\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/771ab44400857ecaa31c/measurements.json b/records/sd1/sweep/771ab44400857ecaa31c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c0e61ccd7568cedf0165eefa02043f8459e3765
--- /dev/null
+++ b/records/sd1/sweep/771ab44400857ecaa31c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "771ab44400857ecaa31c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/771ab44400857ecaa31c/record.json b/records/sd1/sweep/771ab44400857ecaa31c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..44fb2f341c59125b3bf9d7bda796ad3597bcf1bd
--- /dev/null
+++ b/records/sd1/sweep/771ab44400857ecaa31c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "771ab44400857ecaa31c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:30Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8b2ac7cc0cbd7707a822efef7d0617d116efa5ec36594464627051a06bcc08ef.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8b2ac7cc0cbd7707a822efef7d0617d116efa5ec36594464627051a06bcc08ef",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "ca95d0a72b4b6e6160e2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/771ab44400857ecaa31c/workflow.json b/records/sd1/sweep/771ab44400857ecaa31c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d78892b60c1dcfa340249e7aff0ab11c31a5ba7
--- /dev/null
+++ b/records/sd1/sweep/771ab44400857ecaa31c/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 771ab44400857ecaa31c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.h_upd\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7721fd671efa916092cd/measurements.json b/records/sd1/sweep/7721fd671efa916092cd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1db1ac7034a200c9b06486e5a940d0c7b1bf2883
--- /dev/null
+++ b/records/sd1/sweep/7721fd671efa916092cd/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7721fd671efa916092cd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.618,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00534,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1553,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7721fd671efa916092cd/record.json b/records/sd1/sweep/7721fd671efa916092cd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c713d8eb44e881a0b0d3b66f5b7e6d012ee561b3
--- /dev/null
+++ b/records/sd1/sweep/7721fd671efa916092cd/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7721fd671efa916092cd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/99d6c80120ca7114f97be52dd12706be4ffeb6ecd812bfb34c43b241168ed07a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "99d6c80120ca7114f97be52dd12706be4ffeb6ecd812bfb34c43b241168ed07a",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ab8230757807e3e1805d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7721fd671efa916092cd/workflow.json b/records/sd1/sweep/7721fd671efa916092cd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9bc565a157744f45578f5e40fd284248a9675cd0
--- /dev/null
+++ b/records/sd1/sweep/7721fd671efa916092cd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7721fd671efa916092cd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7723a7a55c2419799386/measurements.json b/records/sd1/sweep/7723a7a55c2419799386/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8c0405eea36ae55cea52ba747ca6bf9363bae42
--- /dev/null
+++ b/records/sd1/sweep/7723a7a55c2419799386/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7723a7a55c2419799386",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7723a7a55c2419799386/record.json b/records/sd1/sweep/7723a7a55c2419799386/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fd75c58103d51b5c8b9c9fe0e4c3756f2374a65
--- /dev/null
+++ b/records/sd1/sweep/7723a7a55c2419799386/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7723a7a55c2419799386",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/44f5de68a7223402809b63ddc3f886eb8507e7b31d5e1217b3dd284332c3daac.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "44f5de68a7223402809b63ddc3f886eb8507e7b31d5e1217b3dd284332c3daac",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "f21931577ae85be5937f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7723a7a55c2419799386/workflow.json b/records/sd1/sweep/7723a7a55c2419799386/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..927a32d7731509d0fc76e9369cb2f098bae3d1ca
--- /dev/null
+++ b/records/sd1/sweep/7723a7a55c2419799386/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7723a7a55c2419799386)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7727f8c6478ab69fed00/measurements.json b/records/sd1/sweep/7727f8c6478ab69fed00/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ee3f98105f3f6fa34b7f703ba69b27db38b8437
--- /dev/null
+++ b/records/sd1/sweep/7727f8c6478ab69fed00/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7727f8c6478ab69fed00",
+ "values": {
+  "cosine_distance": {
+   "value": 0.202365,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.294831,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.121789,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.065917,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 43.775,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00571,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 27.287,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9397,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1121,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7727f8c6478ab69fed00/record.json b/records/sd1/sweep/7727f8c6478ab69fed00/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..61c697879e81460c272eee3ed9e9cfef96215d14
--- /dev/null
+++ b/records/sd1/sweep/7727f8c6478ab69fed00/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7727f8c6478ab69fed00",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/307b9210aafc5296280c116bce8dc8cc910539156f6b2f467c9eee3fd601846d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "307b9210aafc5296280c116bce8dc8cc910539156f6b2f467c9eee3fd601846d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "526262f7dde1723ec6c5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7727f8c6478ab69fed00/workflow.json b/records/sd1/sweep/7727f8c6478ab69fed00/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..37874087b3945e0e54c56c07cc7334d18cdaf5d2
--- /dev/null
+++ b/records/sd1/sweep/7727f8c6478ab69fed00/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7727f8c6478ab69fed00)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_layers.3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7729864c3cc7ced84036/measurements.json b/records/sd1/sweep/7729864c3cc7ced84036/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..87d12e5206551d069c229a9f5a84e06656c8dba1
--- /dev/null
+++ b/records/sd1/sweep/7729864c3cc7ced84036/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7729864c3cc7ced84036",
+ "values": {
+  "cosine_distance": {
+   "value": 0.075784,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.116737,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.028566,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01249,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.975,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00598,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.106,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9948,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1272,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7729864c3cc7ced84036/record.json b/records/sd1/sweep/7729864c3cc7ced84036/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb117b832e8d88e7beffd66df7861f5ca8100df0
--- /dev/null
+++ b/records/sd1/sweep/7729864c3cc7ced84036/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7729864c3cc7ced84036",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/df4e1fcca8f293086aa1c8b704b155c2ca2fc2cd6a9129a72b2df5825081ac52.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "df4e1fcca8f293086aa1c8b704b155c2ca2fc2cd6a9129a72b2df5825081ac52",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ef8178e059345bb345cf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7729864c3cc7ced84036/workflow.json b/records/sd1/sweep/7729864c3cc7ced84036/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b333acfb0838bf79537e542d245f1d71dcc1cdeb
--- /dev/null
+++ b/records/sd1/sweep/7729864c3cc7ced84036/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7729864c3cc7ced84036)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77309306157d0e3ba826/measurements.json b/records/sd1/sweep/77309306157d0e3ba826/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..73480f7be6140fcc9d9688e540f1f372a3657759
--- /dev/null
+++ b/records/sd1/sweep/77309306157d0e3ba826/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "77309306157d0e3ba826",
+ "values": {
+  "cosine_distance": {
+   "value": 0.01863,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.040325,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.014317,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.006782,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.779,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0061,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.911,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0186,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1051,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77309306157d0e3ba826/record.json b/records/sd1/sweep/77309306157d0e3ba826/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..af9ee51b14e364a203b5fade708ab551090f97d6
--- /dev/null
+++ b/records/sd1/sweep/77309306157d0e3ba826/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "77309306157d0e3ba826",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:07Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/2f9a29adae187eee9e387510c9f94982b8039f550653558adc955bd1e1c209e0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2f9a29adae187eee9e387510c9f94982b8039f550653558adc955bd1e1c209e0",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f85a1ebee2e0285000ee"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77309306157d0e3ba826/workflow.json b/records/sd1/sweep/77309306157d0e3ba826/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e57a1b0a510d4f1c0f319a78b40ee204a63696ea
--- /dev/null
+++ b/records/sd1/sweep/77309306157d0e3ba826/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77309306157d0e3ba826)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.attn2.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77360af8089199cf2a10/interpretations.jsonl b/records/sd1/sweep/77360af8089199cf2a10/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..08c1048601e305c764ffc782be2ace5953d722cb
--- /dev/null
+++ b/records/sd1/sweep/77360af8089199cf2a10/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "77360af8089199cf2a10"}, "kind": "caption", "value": "Two stacked small portraits of a pale woman with long dark hair and red lips, off-centre", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d18c32db92d9c021"}
+{"target": {"record": "77360af8089199cf2a10"}, "kind": "change", "value": "The single framed portrait breaks into two small, repeated, stacked copies of a pale, red-lipped woman in the upper-left, with blank grey space and a mismatched layout around them.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ce8d73a980ebbc17"}
+{"target": {"record": "77360af8089199cf2a10"}, "kind": "keywords", "value": ["duplicated subject", "stacked portraits", "red lips", "split image", "tiling glitch"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "19d6fdfe25e0dbec"}
+{"target": {"record": "77360af8089199cf2a10"}, "kind": "effect_tags", "value": ["new_composition", "fragmented", "glitch", "surreal", "high_contrast", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "95585a5aab88b3ad"}
diff --git a/records/sd1/sweep/77360af8089199cf2a10/measurements.json b/records/sd1/sweep/77360af8089199cf2a10/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfbdd8fad40a52d1da12b44f05db4ef90bd1f43b
--- /dev/null
+++ b/records/sd1/sweep/77360af8089199cf2a10/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "77360af8089199cf2a10",
+ "values": {
+  "cosine_distance": {
+   "value": 0.627388,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.588115,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.570967,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.283823,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.554,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00642,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 52.173,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8692,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7911,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77360af8089199cf2a10/record.json b/records/sd1/sweep/77360af8089199cf2a10/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..127fa937f4e9e37f924c5511357a7722a7ff609f
--- /dev/null
+++ b/records/sd1/sweep/77360af8089199cf2a10/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "77360af8089199cf2a10",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "self_attn",
+   "module_type": "Dropout",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b1e86ba176fa96df8b2d9fd17e6ed30f062085e2b06cc55263f4fe4fcbd30e2d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b1e86ba176fa96df8b2d9fd17e6ed30f062085e2b06cc55263f4fe4fcbd30e2d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "e14235285816db8459ff"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77360af8089199cf2a10/workflow.json b/records/sd1/sweep/77360af8089199cf2a10/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7707f42bc0491d506490df42302c1efc953e2434
--- /dev/null
+++ b/records/sd1/sweep/77360af8089199cf2a10/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77360af8089199cf2a10)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.attn1.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773a8c6bf9fd24479fe1/measurements.json b/records/sd1/sweep/773a8c6bf9fd24479fe1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d39e84ab423d08793a6e1baed42dec1be6e421d
--- /dev/null
+++ b/records/sd1/sweep/773a8c6bf9fd24479fe1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "773a8c6bf9fd24479fe1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.386181,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.606432,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.457062,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.349771,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.304,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03198,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 45.801,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9282,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4821,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773a8c6bf9fd24479fe1/record.json b/records/sd1/sweep/773a8c6bf9fd24479fe1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cef7452abeb973226950c700a11e0d9f1cc7060
--- /dev/null
+++ b/records/sd1/sweep/773a8c6bf9fd24479fe1/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "773a8c6bf9fd24479fe1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:00Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/7746551d850a5f8029767f5a11fa85afb7294a004a32f35d3ad61db50d035fc7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7746551d850a5f8029767f5a11fa85afb7294a004a32f35d3ad61db50d035fc7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c1b529dc1706f4fa4218"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773a8c6bf9fd24479fe1/workflow.json b/records/sd1/sweep/773a8c6bf9fd24479fe1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb0d4a9f2c36e3c1d9a50c570ec80cb5f992a3a3
--- /dev/null
+++ b/records/sd1/sweep/773a8c6bf9fd24479fe1/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 773a8c6bf9fd24479fe1)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.proj_in\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773c6662fee5b68b9579/measurements.json b/records/sd1/sweep/773c6662fee5b68b9579/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ddc115128eb685e54d578678b8702880059dc17f
--- /dev/null
+++ b/records/sd1/sweep/773c6662fee5b68b9579/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "773c6662fee5b68b9579",
+ "values": {
+  "cosine_distance": {
+   "value": 0.262676,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.311136,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.192099,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.06796,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.112,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00474,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 27.96,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0072,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2069,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773c6662fee5b68b9579/record.json b/records/sd1/sweep/773c6662fee5b68b9579/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5bbc34122c1b4b80664edb64102a22d1a8e1ed1
--- /dev/null
+++ b/records/sd1/sweep/773c6662fee5b68b9579/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "773c6662fee5b68b9579",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/cf86c710ca20e4dea9154af0bf510a32ebdee7f0f804f3250999fe7a7d19814d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "cf86c710ca20e4dea9154af0bf510a32ebdee7f0f804f3250999fe7a7d19814d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "caf8713faf4bbe7f971c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773c6662fee5b68b9579/workflow.json b/records/sd1/sweep/773c6662fee5b68b9579/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d35003c6e46b1ba4ec6c85e8c47141ae59a113b
--- /dev/null
+++ b/records/sd1/sweep/773c6662fee5b68b9579/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 773c6662fee5b68b9579)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773df460d23c71c022f4/measurements.json b/records/sd1/sweep/773df460d23c71c022f4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..467c477854655081ad014a9000ca031648da6ba8
--- /dev/null
+++ b/records/sd1/sweep/773df460d23c71c022f4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "773df460d23c71c022f4",
+ "values": {
+  "std": {
+   "value": 49.664,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01383,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.018,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.081195,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.192241,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.252042,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.108123,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0448,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0154,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773df460d23c71c022f4/record.json b/records/sd1/sweep/773df460d23c71c022f4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..42f0f0090dbc6127329aa6037faaee294129bed1
--- /dev/null
+++ b/records/sd1/sweep/773df460d23c71c022f4/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "773df460d23c71c022f4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:07:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "slight",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "rotate",
+   30.0,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ce72ec291b679917b62a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773df460d23c71c022f4/workflow.json b/records/sd1/sweep/773df460d23c71c022f4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c958f58b73fcb066525aefdd4872b1edb87c22c7
--- /dev/null
+++ b/records/sd1/sweep/773df460d23c71c022f4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 773df460d23c71c022f4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773e185bdcc32839431e/measurements.json b/records/sd1/sweep/773e185bdcc32839431e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f23dfbfad37e6019a6c18a82124ae7c722755b8
--- /dev/null
+++ b/records/sd1/sweep/773e185bdcc32839431e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "773e185bdcc32839431e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.021175,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.034208,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.011851,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.006694,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.251,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00571,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.268,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0145,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1094,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773e185bdcc32839431e/record.json b/records/sd1/sweep/773e185bdcc32839431e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a94c352be98b4c8e35c9e8088de31d2618a327d7
--- /dev/null
+++ b/records/sd1/sweep/773e185bdcc32839431e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "773e185bdcc32839431e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/af9a88966c6b66ff8e23254451c81a55de13b7f7bcfbef2d580d545085914182.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "af9a88966c6b66ff8e23254451c81a55de13b7f7bcfbef2d580d545085914182",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "d6cb30a0165fc7e71ce6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/773e185bdcc32839431e/workflow.json b/records/sd1/sweep/773e185bdcc32839431e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0eb21ca016c1fd640335d8051a50079b3cf287fa
--- /dev/null
+++ b/records/sd1/sweep/773e185bdcc32839431e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 773e185bdcc32839431e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7745a95002895fbd48f0/measurements.json b/records/sd1/sweep/7745a95002895fbd48f0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..91fb0702275beb0026f507eeb3063b2da020db34
--- /dev/null
+++ b/records/sd1/sweep/7745a95002895fbd48f0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7745a95002895fbd48f0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.181195,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.25863,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131661,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.046824,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.166,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00772,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.431,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9644,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0631,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7745a95002895fbd48f0/record.json b/records/sd1/sweep/7745a95002895fbd48f0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad4d721208bd352a34ca3265c98323242ce9da5
--- /dev/null
+++ b/records/sd1/sweep/7745a95002895fbd48f0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7745a95002895fbd48f0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:43Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e71039ea67cf9f21144c966a75c6612ac9ca7b939b77e242bed031f547777105.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e71039ea67cf9f21144c966a75c6612ac9ca7b939b77e242bed031f547777105",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c3694e8f784cd2f085ef"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7745a95002895fbd48f0/workflow.json b/records/sd1/sweep/7745a95002895fbd48f0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..75809b214590f612bcab46213f0a3d191423756d
--- /dev/null
+++ b/records/sd1/sweep/7745a95002895fbd48f0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7745a95002895fbd48f0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/774d92fb612576c939cf/measurements.json b/records/sd1/sweep/774d92fb612576c939cf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a7a094606ee792bb899a8b9f0d5b167f917bb3b
--- /dev/null
+++ b/records/sd1/sweep/774d92fb612576c939cf/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "774d92fb612576c939cf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.022585,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.04341,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.013943,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.006156,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.574,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00552,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 6.829,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0134,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.167,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/774d92fb612576c939cf/record.json b/records/sd1/sweep/774d92fb612576c939cf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b8007abb6206360d1b971eeed46ef020d2d7498
--- /dev/null
+++ b/records/sd1/sweep/774d92fb612576c939cf/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "774d92fb612576c939cf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.1.transformer_blocks.0.attn2.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Dropout",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/987aba88224b76cf9410434135887fc7b49d314123a46b440070e431ef047b0d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "987aba88224b76cf9410434135887fc7b49d314123a46b440070e431ef047b0d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5c5d834af3a88b991cf5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/774d92fb612576c939cf/workflow.json b/records/sd1/sweep/774d92fb612576c939cf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..468905665ce50d436a6fc943b85649448dbf8abc
--- /dev/null
+++ b/records/sd1/sweep/774d92fb612576c939cf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 774d92fb612576c939cf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.transformer_blocks.0.attn2.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/774f902671c21085e6a4/measurements.json b/records/sd1/sweep/774f902671c21085e6a4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b68b1fa94c5a99ef7659181cd7bd124f6366cd62
--- /dev/null
+++ b/records/sd1/sweep/774f902671c21085e6a4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "774f902671c21085e6a4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.342768,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.412244,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.257069,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.074484,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 39.588,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00446,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.949,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9575,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0481,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/774f902671c21085e6a4/record.json b/records/sd1/sweep/774f902671c21085e6a4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ac26e1a074772f371c1b49dcedb9f0bb757cc82
--- /dev/null
+++ b/records/sd1/sweep/774f902671c21085e6a4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "774f902671c21085e6a4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:22Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e733a7c87f474503a3fde994d1d4687a47a1b9fb495e52ae6f4c601bbe17f3b1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e733a7c87f474503a3fde994d1d4687a47a1b9fb495e52ae6f4c601bbe17f3b1",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "006b07670ac3d8dc1809"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/774f902671c21085e6a4/workflow.json b/records/sd1/sweep/774f902671c21085e6a4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..59d97bc8f46df4292b86c1bd789789d22873ad00
--- /dev/null
+++ b/records/sd1/sweep/774f902671c21085e6a4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 774f902671c21085e6a4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776165928da293234a58/measurements.json b/records/sd1/sweep/776165928da293234a58/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1fcf20b2ea32d1828cf99511c8b80daca2ec4b51
--- /dev/null
+++ b/records/sd1/sweep/776165928da293234a58/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "776165928da293234a58",
+ "values": {
+  "cosine_distance": {
+   "value": 0.445752,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.59106,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.433138,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.274145,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 31.977,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00247,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 42.52,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9872,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2748,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776165928da293234a58/record.json b/records/sd1/sweep/776165928da293234a58/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..358794a0d28e10a53fde1fd72ee30f7682b024b6
--- /dev/null
+++ b/records/sd1/sweep/776165928da293234a58/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "776165928da293234a58",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/c51351906b63cbc3e011aeb53cd87e2f6836510a699efcbdfd15d9d97e2029d3.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c51351906b63cbc3e011aeb53cd87e2f6836510a699efcbdfd15d9d97e2029d3",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "56d71d66049aae8b320d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776165928da293234a58/workflow.json b/records/sd1/sweep/776165928da293234a58/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e537d0c32bc5e51428a1e9d8d22029f511fc910c
--- /dev/null
+++ b/records/sd1/sweep/776165928da293234a58/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 776165928da293234a58)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7766ec77d9b1c0c14020/measurements.json b/records/sd1/sweep/7766ec77d9b1c0c14020/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..06868cb5da2452941e52707ca91b1c10486b32bc
--- /dev/null
+++ b/records/sd1/sweep/7766ec77d9b1c0c14020/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7766ec77d9b1c0c14020",
+ "values": {
+  "cosine_distance": {
+   "value": 0.300606,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.304241,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.3474,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.204412,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.878,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0037,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.479,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0362,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1826,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7766ec77d9b1c0c14020/record.json b/records/sd1/sweep/7766ec77d9b1c0c14020/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c9a9927646a77ab541dcd2a4e6abcf01e4ade96
--- /dev/null
+++ b/records/sd1/sweep/7766ec77d9b1c0c14020/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7766ec77d9b1c0c14020",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/7ddb773e4638c9e3c03d1ef635125f2fb9330c0888e5f96bb354389f4b5d5115.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7ddb773e4638c9e3c03d1ef635125f2fb9330c0888e5f96bb354389f4b5d5115",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c08644b6691db4371418"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7766ec77d9b1c0c14020/workflow.json b/records/sd1/sweep/7766ec77d9b1c0c14020/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f49890d0bf0c21646929624c030c7870f0c887f5
--- /dev/null
+++ b/records/sd1/sweep/7766ec77d9b1c0c14020/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7766ec77d9b1c0c14020)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77684dfd84da99e04c03/interpretations.jsonl b/records/sd1/sweep/77684dfd84da99e04c03/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..18f1acbc62e958c94619d37bdf865334fc0b104b
--- /dev/null
+++ b/records/sd1/sweep/77684dfd84da99e04c03/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "77684dfd84da99e04c03"}, "kind": "caption", "value": "Windswept old man with long white hair in grey jacket before stormy sea", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:47:19Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2ef65ad64a6d2b7c"}
+{"target": {"record": "77684dfd84da99e04c03"}, "kind": "change", "value": "The knit cap is gone and the old man is re-posed in a weathered work jacket with wind-blown hair, against a wider, stormier sea with cooler, greyer tones.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:47:19Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c672639794c76a0e"}
+{"target": {"record": "77684dfd84da99e04c03"}, "kind": "keywords", "value": ["old fisherman", "windswept hair", "stormy sea", "work jacket", "new pose"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:47:19Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3754ea2cb0a01868"}
+{"target": {"record": "77684dfd84da99e04c03"}, "kind": "effect_tags", "value": ["new_composition", "subject_kept", "desaturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:47:19Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4552b64fcf8672da"}
diff --git a/records/sd1/sweep/77684dfd84da99e04c03/measurements.json b/records/sd1/sweep/77684dfd84da99e04c03/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..923bb4574a3be21673fe0b541cd632fc2ba2263e
--- /dev/null
+++ b/records/sd1/sweep/77684dfd84da99e04c03/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "77684dfd84da99e04c03",
+ "values": {
+  "std": {
+   "value": 43.899,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01253,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.955,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.433539,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.468497,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.279252,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.118612,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9667,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3346,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77684dfd84da99e04c03/record.json b/records/sd1/sweep/77684dfd84da99e04c03/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fbf10042da695e484357460984b550920391a43
--- /dev/null
+++ b/records/sd1/sweep/77684dfd84da99e04c03/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "77684dfd84da99e04c03",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:19:07Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "multiply",
+   2.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "398309af507755f6444c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77684dfd84da99e04c03/workflow.json b/records/sd1/sweep/77684dfd84da99e04c03/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8708851c3abfd3e312429ffd3320cbd85aa4433f
--- /dev/null
+++ b/records/sd1/sweep/77684dfd84da99e04c03/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77684dfd84da99e04c03)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7769ff941bb4634787e2/measurements.json b/records/sd1/sweep/7769ff941bb4634787e2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..527f9aeb5ac11e352c4ac68bb502f0befca54613
--- /dev/null
+++ b/records/sd1/sweep/7769ff941bb4634787e2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7769ff941bb4634787e2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.215384,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.253262,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.129143,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.051052,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.321,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00508,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 27.186,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9421,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1097,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7769ff941bb4634787e2/record.json b/records/sd1/sweep/7769ff941bb4634787e2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..82721f5a778e24f73b859396137e4f2da89d1452
--- /dev/null
+++ b/records/sd1/sweep/7769ff941bb4634787e2/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7769ff941bb4634787e2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e1aa573559ef9a70d39dd3584eef6324a428ee51a2212365754e9e656dd73bff.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e1aa573559ef9a70d39dd3584eef6324a428ee51a2212365754e9e656dd73bff",
+  "type_path": "UNetModel.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f7465c9dcb66570f7523"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7769ff941bb4634787e2/workflow.json b/records/sd1/sweep/7769ff941bb4634787e2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..765572d50a533e230a58d0a9908abe97d404268b
--- /dev/null
+++ b/records/sd1/sweep/7769ff941bb4634787e2/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7769ff941bb4634787e2)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.norm2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776a1ff454943ee0151f/interpretations.jsonl b/records/sd1/sweep/776a1ff454943ee0151f/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c7152922700f6f8d880f49fb35532c524648aa75
--- /dev/null
+++ b/records/sd1/sweep/776a1ff454943ee0151f/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "776a1ff454943ee0151f"}, "kind": "caption", "value": "Blurry abstract patchwork of blue, cream and brown blobs with no clear subject", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "db3fa9aead8573fe"}
+{"target": {"record": "776a1ff454943ee0151f"}, "kind": "change", "value": "The portrait dissolves into a soft, blotchy patchwork of blue, cream and dark tones; no face remains.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a2b5c0e49b020766"}
+{"target": {"record": "776a1ff454943ee0151f"}, "kind": "keywords", "value": ["abstract blobs", "blurry patchwork", "lost portrait", "degenerate output"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "84cfc5e3ed0cff74"}
+{"target": {"record": "776a1ff454943ee0151f"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "blur", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "267b21852309a209"}
diff --git a/records/sd1/sweep/776a1ff454943ee0151f/measurements.json b/records/sd1/sweep/776a1ff454943ee0151f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9146f4886c0bc7be6e73f8e2cf27c7605f0dc00e
--- /dev/null
+++ b/records/sd1/sweep/776a1ff454943ee0151f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "776a1ff454943ee0151f",
+ "values": {
+  "std": {
+   "value": 55.029,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00312,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 56.002,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.664677,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.681184,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.953746,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.394865,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7271,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2977,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776a1ff454943ee0151f/record.json b/records/sd1/sweep/776a1ff454943ee0151f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce2caf1c7d704d4b91c4b4e282d529107887ae1e
--- /dev/null
+++ b/records/sd1/sweep/776a1ff454943ee0151f/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "776a1ff454943ee0151f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:17:36Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "rotate",
+   180.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 180.0,
+    "applied": 180.0,
+    "reason": "outside safe range [0.0, 5.0] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "bb349204ea085253ff46"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776a1ff454943ee0151f/workflow.json b/records/sd1/sweep/776a1ff454943ee0151f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b57e112c451e6ebf04cc1f7802c05c86d8b04451
--- /dev/null
+++ b/records/sd1/sweep/776a1ff454943ee0151f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 776a1ff454943ee0151f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776f1756e2360bff46d3/measurements.json b/records/sd1/sweep/776f1756e2360bff46d3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfa456b20b9f49c8d79d68fb3dd351740093e197
--- /dev/null
+++ b/records/sd1/sweep/776f1756e2360bff46d3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "776f1756e2360bff46d3",
+ "values": {
+  "std": {
+   "value": 46.029,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01308,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 6.744,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.01886,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.043167,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.00792,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007468,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0567,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1512,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776f1756e2360bff46d3/record.json b/records/sd1/sweep/776f1756e2360bff46d3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f8cb7eaa2b640c0f3dcf9c4202d4ec1b236b072
--- /dev/null
+++ b/records/sd1/sweep/776f1756e2360bff46d3/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "776f1756e2360bff46d3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:29:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "heavy"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_noise",
+   0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f993bace8a7e5ba49eb1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/776f1756e2360bff46d3/workflow.json b/records/sd1/sweep/776f1756e2360bff46d3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..af19b47eae8a5aae77bb93935a456bad6814c0dc
--- /dev/null
+++ b/records/sd1/sweep/776f1756e2360bff46d3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 776f1756e2360bff46d3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/777eb2bbeb0450a00126/measurements.json b/records/sd1/sweep/777eb2bbeb0450a00126/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b977367342cad91b8eaa97d65a7de1811b2eccd3
--- /dev/null
+++ b/records/sd1/sweep/777eb2bbeb0450a00126/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "777eb2bbeb0450a00126",
+ "values": {
+  "cosine_distance": {
+   "value": 0.145942,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.182752,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.138892,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.072961,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.12,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00355,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.642,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9427,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.058,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/777eb2bbeb0450a00126/record.json b/records/sd1/sweep/777eb2bbeb0450a00126/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a39e89dddcc86f816e31ed88905f75896e95aead
--- /dev/null
+++ b/records/sd1/sweep/777eb2bbeb0450a00126/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "777eb2bbeb0450a00126",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:44Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/16597f923528983bda2149e68fa6f370421593a87c38d5fb8452d950cfaa2da8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "16597f923528983bda2149e68fa6f370421593a87c38d5fb8452d950cfaa2da8",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "daa5d853fd6a4f99638d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/777eb2bbeb0450a00126/workflow.json b/records/sd1/sweep/777eb2bbeb0450a00126/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7ba404b9034e78bd43ca6f09876c1bbb2ce7cb8
--- /dev/null
+++ b/records/sd1/sweep/777eb2bbeb0450a00126/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 777eb2bbeb0450a00126)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77bb6e1ea3a67fd45780/measurements.json b/records/sd1/sweep/77bb6e1ea3a67fd45780/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..40af4c973a730695d3984283cbf46abbb4a27190
--- /dev/null
+++ b/records/sd1/sweep/77bb6e1ea3a67fd45780/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "77bb6e1ea3a67fd45780",
+ "values": {
+  "std": {
+   "value": 51.624,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00656,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.542,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.493204,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.666754,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.664257,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.385313,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7069,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7057,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77bb6e1ea3a67fd45780/record.json b/records/sd1/sweep/77bb6e1ea3a67fd45780/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a572329d265d976c624f7126370ceb7cdd3a4d0b
--- /dev/null
+++ b/records/sd1/sweep/77bb6e1ea3a67fd45780/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "77bb6e1ea3a67fd45780",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:01:12Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "slight",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "rotate",
+   30.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 30.0,
+    "applied": 30.0,
+    "reason": "outside safe range [0.0, 5.0] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "1d13cc43854e2a0fa2c2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77bb6e1ea3a67fd45780/workflow.json b/records/sd1/sweep/77bb6e1ea3a67fd45780/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f962ab0759c462ba77dc227a30dd3197e4e7c6d
--- /dev/null
+++ b/records/sd1/sweep/77bb6e1ea3a67fd45780/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77bb6e1ea3a67fd45780)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77c0e9e77219ac5eeb94/measurements.json b/records/sd1/sweep/77c0e9e77219ac5eeb94/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..291b4788adeb8627c6e768e746d5f4d05f079702
--- /dev/null
+++ b/records/sd1/sweep/77c0e9e77219ac5eeb94/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "77c0e9e77219ac5eeb94",
+ "values": {
+  "cosine_distance": {
+   "value": 0.557532,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.528341,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.401735,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.134421,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 83.312,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00609,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 63.571,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9349,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5192,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77c0e9e77219ac5eeb94/record.json b/records/sd1/sweep/77c0e9e77219ac5eeb94/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5bd5cdd2392ab1bfeabb2844ebbf568e9ae78f6
--- /dev/null
+++ b/records/sd1/sweep/77c0e9e77219ac5eeb94/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "77c0e9e77219ac5eeb94",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:51Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/fe912556c5e83de0b21d81399536db6bd6c73c27c8826c94c8838ccbff13daf8.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "fe912556c5e83de0b21d81399536db6bd6c73c27c8826c94c8838ccbff13daf8",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "7c0d00f24341ff4444e8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77c0e9e77219ac5eeb94/workflow.json b/records/sd1/sweep/77c0e9e77219ac5eeb94/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bb0488201bd03d532dbc6248f69ba6daec00140
--- /dev/null
+++ b/records/sd1/sweep/77c0e9e77219ac5eeb94/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77c0e9e77219ac5eeb94)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.norm1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77c850588b59d524ea36/measurements.json b/records/sd1/sweep/77c850588b59d524ea36/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..31044e3b25514b6ea0085c425ab58391604fa7b1
--- /dev/null
+++ b/records/sd1/sweep/77c850588b59d524ea36/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "77c850588b59d524ea36",
+ "values": {
+  "cosine_distance": {
+   "value": 0.052885,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.081996,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.041731,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01594,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.173,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00532,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.417,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.984,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.188,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77c850588b59d524ea36/record.json b/records/sd1/sweep/77c850588b59d524ea36/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d777948eda226e3cdf153e182ba1b9013f4d679
--- /dev/null
+++ b/records/sd1/sweep/77c850588b59d524ea36/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "77c850588b59d524ea36",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:15Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/cb792f3d60d2d15e63ef4e9490f04c26b798b589f52d9a03a330dc9d79777748.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "cb792f3d60d2d15e63ef4e9490f04c26b798b589f52d9a03a330dc9d79777748",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6bb9d434692302acda5e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77c850588b59d524ea36/workflow.json b/records/sd1/sweep/77c850588b59d524ea36/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d41010626f6c06c6e85b34c004f23a161abfcc9a
--- /dev/null
+++ b/records/sd1/sweep/77c850588b59d524ea36/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77c850588b59d524ea36)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77ced9b2372f268fdc8f/interpretations.jsonl b/records/sd1/sweep/77ced9b2372f268fdc8f/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a6d2c609cb515273fa05b68842605014e04e6064
--- /dev/null
+++ b/records/sd1/sweep/77ced9b2372f268fdc8f/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "77ced9b2372f268fdc8f"}, "kind": "caption", "value": "Portrait of a woman with a dark brown shoulder-length bob and turtleneck on grey", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6dc37ea3167ab4bf"}
+{"target": {"record": "77ced9b2372f268fdc8f"}, "kind": "change", "value": "Restyled into a cleaner illustration: a dark brown shoulder-length bob, rounder face, larger eyes and a flatter grey-green background with no border.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "faeb60127ca9953b"}
+{"target": {"record": "77ced9b2372f268fdc8f"}, "kind": "keywords", "value": ["portrait", "dark bob", "illustration", "round face"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "05e1042a520001f9"}
+{"target": {"record": "77ced9b2372f268fdc8f"}, "kind": "effect_tags", "value": ["subject_kept", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c70569979e536cb8"}
diff --git a/records/sd1/sweep/77ced9b2372f268fdc8f/measurements.json b/records/sd1/sweep/77ced9b2372f268fdc8f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0bd374905d570e2dcdc7defd73ad4298d67aca9
--- /dev/null
+++ b/records/sd1/sweep/77ced9b2372f268fdc8f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "77ced9b2372f268fdc8f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.26271,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.306574,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.142186,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.055539,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.795,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0052,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 32.122,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9862,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0742,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77ced9b2372f268fdc8f/record.json b/records/sd1/sweep/77ced9b2372f268fdc8f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f34e9b8eece5a18a685d016029b73100207889c
--- /dev/null
+++ b/records/sd1/sweep/77ced9b2372f268fdc8f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "77ced9b2372f268fdc8f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/eb29d6c4d2ee5fca34faf804b8fb9896b6a4d692746fdea17b2a89c1bc055216.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "eb29d6c4d2ee5fca34faf804b8fb9896b6a4d692746fdea17b2a89c1bc055216",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4464d3847ab7b1048e73"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77ced9b2372f268fdc8f/workflow.json b/records/sd1/sweep/77ced9b2372f268fdc8f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..75ebf46ecb04b56fdb68583a934958587a5748ae
--- /dev/null
+++ b/records/sd1/sweep/77ced9b2372f268fdc8f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77ced9b2372f268fdc8f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77d7e83376bc888e94ff/measurements.json b/records/sd1/sweep/77d7e83376bc888e94ff/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c780be01ac3f2aa21ee951b49ba5fe2aa49d03b
--- /dev/null
+++ b/records/sd1/sweep/77d7e83376bc888e94ff/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "77d7e83376bc888e94ff",
+ "values": {
+  "cosine_distance": {
+   "value": 0.380264,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.414075,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.255997,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.122207,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.003,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00497,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.794,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0041,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1033,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77d7e83376bc888e94ff/record.json b/records/sd1/sweep/77d7e83376bc888e94ff/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..93a64283d900e63e8d8c053ed3ae46febc622e66
--- /dev/null
+++ b/records/sd1/sweep/77d7e83376bc888e94ff/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "77d7e83376bc888e94ff",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:08Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/026ff51146039ad231f26990b535e986d1f083dbdbd66d4d39856489f1e1a737.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "026ff51146039ad231f26990b535e986d1f083dbdbd66d4d39856489f1e1a737",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "d2af1a70579416d8a697"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77d7e83376bc888e94ff/workflow.json b/records/sd1/sweep/77d7e83376bc888e94ff/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..139acf9ded3cdb9df2571f0858051ca50f9733dd
--- /dev/null
+++ b/records/sd1/sweep/77d7e83376bc888e94ff/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77d7e83376bc888e94ff)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77e054db3c219b1f0838/measurements.json b/records/sd1/sweep/77e054db3c219b1f0838/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a3eb09ba3234a946696d5acd4cd7acf0fb8d289
--- /dev/null
+++ b/records/sd1/sweep/77e054db3c219b1f0838/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "77e054db3c219b1f0838",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.618,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00534,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1553,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77e054db3c219b1f0838/record.json b/records/sd1/sweep/77e054db3c219b1f0838/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e5118d5d66ee38cac4a2d18c8e056388b787791
--- /dev/null
+++ b/records/sd1/sweep/77e054db3c219b1f0838/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "77e054db3c219b1f0838",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/352c6ef9ac7f6dea9843805ed3ce1a5171b482bad0d2131b786cf59443e4e1e8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "352c6ef9ac7f6dea9843805ed3ce1a5171b482bad0d2131b786cf59443e4e1e8",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "576ca3c69c30e27c523c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77e054db3c219b1f0838/workflow.json b/records/sd1/sweep/77e054db3c219b1f0838/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9accfd7f1c50fd1e0f3b64d1231245f663cad790
--- /dev/null
+++ b/records/sd1/sweep/77e054db3c219b1f0838/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77e054db3c219b1f0838)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77ecf594e6e71c88f48a/interpretations.jsonl b/records/sd1/sweep/77ecf594e6e71c88f48a/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..806347350912a5fc900e2e3075eb94042b6cfd9a
--- /dev/null
+++ b/records/sd1/sweep/77ecf594e6e71c88f48a/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "77ecf594e6e71c88f48a"}, "kind": "caption", "value": "Smooth, doll-like young man with blue eyes and maroon helmet hair on diagonal-striped background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b7f2e5e16dea8c16"}
+{"target": {"record": "77ecf594e6e71c88f48a"}, "kind": "change", "value": "The photo becomes a soft, plasticky, airbrushed frontal male face with a rounded maroon hair cap, warm pink skin and faint diagonal stripes in the backdrop.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2ea4e06bbe0c5432"}
+{"target": {"record": "77ecf594e6e71c88f48a"}, "kind": "keywords", "value": ["airbrushed face", "doll-like man", "diagonal stripes", "smooth render", "frontal portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3ddc55a523f4a833"}
+{"target": {"record": "77ecf594e6e71c88f48a"}, "kind": "effect_tags", "value": ["blur", "new_composition", "symmetry", "pattern", "style_change", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ec07a5f70786ce4b"}
diff --git a/records/sd1/sweep/77ecf594e6e71c88f48a/measurements.json b/records/sd1/sweep/77ecf594e6e71c88f48a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..50fb752ca169c625c9531193e6611a08964aa818
--- /dev/null
+++ b/records/sd1/sweep/77ecf594e6e71c88f48a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "77ecf594e6e71c88f48a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.549589,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.586715,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.424091,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.379264,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 82.932,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00276,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 53.367,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9312,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8672,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77ecf594e6e71c88f48a/record.json b/records/sd1/sweep/77ecf594e6e71c88f48a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..744cd17c9ba97f12732e80389c6382ab926b92a8
--- /dev/null
+++ b/records/sd1/sweep/77ecf594e6e71c88f48a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "77ecf594e6e71c88f48a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/19d07b67cbbd262ba7d8f0cdc7e9b420cb03b5bdbdf2eb03f8297fbdb84263cd.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "19d07b67cbbd262ba7d8f0cdc7e9b420cb03b5bdbdf2eb03f8297fbdb84263cd",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "baae716b05fbf03c176d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77ecf594e6e71c88f48a/workflow.json b/records/sd1/sweep/77ecf594e6e71c88f48a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e00d5c8c1e36161beaad4c44b3e305e1425fee1
--- /dev/null
+++ b/records/sd1/sweep/77ecf594e6e71c88f48a/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77ecf594e6e71c88f48a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77f38075fa85ac261fcd/measurements.json b/records/sd1/sweep/77f38075fa85ac261fcd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea01a3e64de4e09228d7a9ad26c306c6a8f16918
--- /dev/null
+++ b/records/sd1/sweep/77f38075fa85ac261fcd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "77f38075fa85ac261fcd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.134647,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.208367,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.10771,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023626,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.452,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0077,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.439,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9752,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1277,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77f38075fa85ac261fcd/record.json b/records/sd1/sweep/77f38075fa85ac261fcd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d89f8067ab2f3fb1170b17738bc4d7ce30d32afa
--- /dev/null
+++ b/records/sd1/sweep/77f38075fa85ac261fcd/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "77f38075fa85ac261fcd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:51Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/be105084362e193f36581cb50e7e19fdd2a41337615242f8269e2d4cae8b01ea.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "be105084362e193f36581cb50e7e19fdd2a41337615242f8269e2d4cae8b01ea",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4ef11d4860d7d96e08c2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/77f38075fa85ac261fcd/workflow.json b/records/sd1/sweep/77f38075fa85ac261fcd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a6ac809dfdcd0fe46872a3136a52f25754a80c9
--- /dev/null
+++ b/records/sd1/sweep/77f38075fa85ac261fcd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 77f38075fa85ac261fcd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780491e63d6aa6950a0f/measurements.json b/records/sd1/sweep/780491e63d6aa6950a0f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dff82db32006b8c728092e6eafa0fce051b916dc
--- /dev/null
+++ b/records/sd1/sweep/780491e63d6aa6950a0f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "780491e63d6aa6950a0f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.490143,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.519659,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.466822,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.219346,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 83.448,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00699,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 56.02,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9459,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2955,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780491e63d6aa6950a0f/record.json b/records/sd1/sweep/780491e63d6aa6950a0f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba67a598e32f466b7669998a54481fb320fb1345
--- /dev/null
+++ b/records/sd1/sweep/780491e63d6aa6950a0f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "780491e63d6aa6950a0f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/dc41d6f514bfe9ffd4cd469f248eb9786bc3c784773d3b62c9fa7a1417333d02.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "dc41d6f514bfe9ffd4cd469f248eb9786bc3c784773d3b62c9fa7a1417333d02",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "10c86b06aa6d1c4c5e1d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780491e63d6aa6950a0f/workflow.json b/records/sd1/sweep/780491e63d6aa6950a0f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f5ac964d84a4533c4811865884e6a0d8848bf5b
--- /dev/null
+++ b/records/sd1/sweep/780491e63d6aa6950a0f/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 780491e63d6aa6950a0f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780ca188aa4fc6b233f1/measurements.json b/records/sd1/sweep/780ca188aa4fc6b233f1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0ed87b6b1c128600c3b1e65eca7f0d2657f8542
--- /dev/null
+++ b/records/sd1/sweep/780ca188aa4fc6b233f1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "780ca188aa4fc6b233f1",
+ "values": {
+  "std": {
+   "value": 32.298,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02719,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.169,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.161199,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.291128,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.157495,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.049158,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0527,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3928,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780ca188aa4fc6b233f1/record.json b/records/sd1/sweep/780ca188aa4fc6b233f1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f6916eec3b01a30f6dc061e87e8068c4e500d36
--- /dev/null
+++ b/records/sd1/sweep/780ca188aa4fc6b233f1/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "780ca188aa4fc6b233f1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:18:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "damp",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "multiply",
+   0.5,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0c189d5929567f1ff97f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780ca188aa4fc6b233f1/workflow.json b/records/sd1/sweep/780ca188aa4fc6b233f1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b17e87f90583b116ff8e7b104f85b51e71fdd330
--- /dev/null
+++ b/records/sd1/sweep/780ca188aa4fc6b233f1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 780ca188aa4fc6b233f1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780fe68ea2b0185dd348/measurements.json b/records/sd1/sweep/780fe68ea2b0185dd348/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..309862eb340878b9c013e3c326277596e346649a
--- /dev/null
+++ b/records/sd1/sweep/780fe68ea2b0185dd348/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "780fe68ea2b0185dd348",
+ "values": {
+  "cosine_distance": {
+   "value": 0.100412,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.120424,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.041169,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02105,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.94,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00517,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.181,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.963,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1118,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780fe68ea2b0185dd348/record.json b/records/sd1/sweep/780fe68ea2b0185dd348/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0dedf4fb18044afe16b3ff3daa5d280bd6a4ecd
--- /dev/null
+++ b/records/sd1/sweep/780fe68ea2b0185dd348/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "780fe68ea2b0185dd348",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:50Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/4c92d4e7238ef05f61128ff1db8b0d955ee77fc85b2a15a37ede772c79d0eea1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4c92d4e7238ef05f61128ff1db8b0d955ee77fc85b2a15a37ede772c79d0eea1",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "716da96c7a62037f388c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/780fe68ea2b0185dd348/workflow.json b/records/sd1/sweep/780fe68ea2b0185dd348/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9356147178236aa67a02c86b1bd2e019497390d
--- /dev/null
+++ b/records/sd1/sweep/780fe68ea2b0185dd348/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 780fe68ea2b0185dd348)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/782e0d34306dd8c45f2a/measurements.json b/records/sd1/sweep/782e0d34306dd8c45f2a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c91e886b6a1634bf2025313cfa88c7dc6b03d66
--- /dev/null
+++ b/records/sd1/sweep/782e0d34306dd8c45f2a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "782e0d34306dd8c45f2a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.067964,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.095589,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.061871,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.022848,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.519,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0057,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.734,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.031,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1215,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/782e0d34306dd8c45f2a/record.json b/records/sd1/sweep/782e0d34306dd8c45f2a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae0cc03f0acd8c2bfc1681a0914633abdedc1b78
--- /dev/null
+++ b/records/sd1/sweep/782e0d34306dd8c45f2a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "782e0d34306dd8c45f2a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/586f7e9791827952eaaefc6805b27b9e3a94ca7246f6fdb536536b3b9c1cf3ab.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "586f7e9791827952eaaefc6805b27b9e3a94ca7246f6fdb536536b3b9c1cf3ab",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8f805c0a5e6b287cef08"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/782e0d34306dd8c45f2a/workflow.json b/records/sd1/sweep/782e0d34306dd8c45f2a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3cdda35194dcfcc0b535fea4c4f1f996655cab5a
--- /dev/null
+++ b/records/sd1/sweep/782e0d34306dd8c45f2a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 782e0d34306dd8c45f2a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7835219479fa3eb99912/measurements.json b/records/sd1/sweep/7835219479fa3eb99912/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6967d961c1ba00f25893c1a84543c397a5481cde
--- /dev/null
+++ b/records/sd1/sweep/7835219479fa3eb99912/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7835219479fa3eb99912",
+ "values": {
+  "std": {
+   "value": 44.28,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01762,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 50.204,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.58102,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.681158,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.877996,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.382004,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.715,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0321,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7835219479fa3eb99912/record.json b/records/sd1/sweep/7835219479fa3eb99912/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9441ec166dec7a2c411ce1aaa1e191d2688c1164
--- /dev/null
+++ b/records/sd1/sweep/7835219479fa3eb99912/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "7835219479fa3eb99912",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:28:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   90.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "1a2985e885a5c9f62fe6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7835219479fa3eb99912/workflow.json b/records/sd1/sweep/7835219479fa3eb99912/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..33ae29f9791b7bd966ea6a5d23c7f7f0da2512b1
--- /dev/null
+++ b/records/sd1/sweep/7835219479fa3eb99912/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7835219479fa3eb99912)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/783f07e402f3192ce7e1/measurements.json b/records/sd1/sweep/783f07e402f3192ce7e1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d15771e1a1412623e0531f499988d06297e5faa1
--- /dev/null
+++ b/records/sd1/sweep/783f07e402f3192ce7e1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "783f07e402f3192ce7e1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.178161,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.271898,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.148593,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.054804,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.444,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00523,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.698,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9914,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.136,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/783f07e402f3192ce7e1/record.json b/records/sd1/sweep/783f07e402f3192ce7e1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..be59d56209f8f97ca83c4039ed1eee647719e412
--- /dev/null
+++ b/records/sd1/sweep/783f07e402f3192ce7e1/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "783f07e402f3192ce7e1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/c075fd46cac9c79456e27c00068db663171a27a34e3726574e408850709271e2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c075fd46cac9c79456e27c00068db663171a27a34e3726574e408850709271e2",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "1321f140babadd79fcd8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/783f07e402f3192ce7e1/workflow.json b/records/sd1/sweep/783f07e402f3192ce7e1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c5710d761a0e3ace8bc9e76cadfd01c541dc40a
--- /dev/null
+++ b/records/sd1/sweep/783f07e402f3192ce7e1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 783f07e402f3192ce7e1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.norm1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78402d2816e50c0ec106/interpretations.jsonl b/records/sd1/sweep/78402d2816e50c0ec106/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..be239abbaa29697d1d86dfa10b9877500d736540
--- /dev/null
+++ b/records/sd1/sweep/78402d2816e50c0ec106/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "78402d2816e50c0ec106"}, "kind": "caption", "value": "Turbulent swirl of grey-white brushy strokes with a faint reddish face shape", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:31Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4906693a8ad048c3"}
+{"target": {"record": "78402d2816e50c0ec106"}, "kind": "change", "value": "The portrait is swept into a stormy, wave-like swirl of white and grey brushwork that merges with the sea, leaving only a faint hint of a face.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:31Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0efeb4aa28c36221"}
+{"target": {"record": "78402d2816e50c0ec106"}, "kind": "keywords", "value": ["swirling brushwork", "stormy abstraction", "dissolved portrait", "sea foam", "turbulent"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:31Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fc0229440337430f"}
+{"target": {"record": "78402d2816e50c0ec106"}, "kind": "effect_tags", "value": ["abstract", "painterly", "subject_lost", "texture", "distorted"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:31Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6418558b1a48b457"}
diff --git a/records/sd1/sweep/78402d2816e50c0ec106/measurements.json b/records/sd1/sweep/78402d2816e50c0ec106/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f119b1cefdc601586335fd81acb503a42bca868c
--- /dev/null
+++ b/records/sd1/sweep/78402d2816e50c0ec106/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78402d2816e50c0ec106",
+ "values": {
+  "std": {
+   "value": 38.466,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01577,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.295,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.500089,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.642664,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.917686,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.406276,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.5956,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78402d2816e50c0ec106/record.json b/records/sd1/sweep/78402d2816e50c0ec106/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..072437374c1078cd17bf881dde469325d35597b9
--- /dev/null
+++ b/records/sd1/sweep/78402d2816e50c0ec106/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "78402d2816e50c0ec106",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:35:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "rotate",
+   90.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [5.0, 45.0] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e8e03a2cf1f834ac9845"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78402d2816e50c0ec106/workflow.json b/records/sd1/sweep/78402d2816e50c0ec106/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5d4cb22d75425c849388de4f249a07be28a85f1
--- /dev/null
+++ b/records/sd1/sweep/78402d2816e50c0ec106/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78402d2816e50c0ec106)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/784a55c6fdb74082f404/interpretations.jsonl b/records/sd1/sweep/784a55c6fdb74082f404/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..694e1a9e64d5bcebcf2e672883ed216b78a4cd05
--- /dev/null
+++ b/records/sd1/sweep/784a55c6fdb74082f404/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "784a55c6fdb74082f404"}, "kind": "caption", "value": "Chaotic mosaic of dark blobs with glowing cyan, pink and violet patches", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e8f8dd940c53dedc"}
+{"target": {"record": "784a55c6fdb74082f404"}, "kind": "change", "value": "The neon street fragments into a dark, patchy mosaic; the central cyan glow and scattered pink-purple highlights remain but no structure survives.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2d556f32dadb09b4"}
+{"target": {"record": "784a55c6fdb74082f404"}, "kind": "keywords", "value": ["neon mosaic", "fragmented street", "cyan glow", "abstract blotches"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "81d43c3672432659"}
+{"target": {"record": "784a55c6fdb74082f404"}, "kind": "effect_tags", "value": ["abstract", "fragmented", "subject_lost", "texture", "dark", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a3069de563e2a780"}
diff --git a/records/sd1/sweep/784a55c6fdb74082f404/measurements.json b/records/sd1/sweep/784a55c6fdb74082f404/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5701c5b9787144b386aebf144c2c93197c29841b
--- /dev/null
+++ b/records/sd1/sweep/784a55c6fdb74082f404/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "784a55c6fdb74082f404",
+ "values": {
+  "std": {
+   "value": 68.483,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00449,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 58.493,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.445679,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.655566,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.924995,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.3042,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7107,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8212,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/784a55c6fdb74082f404/record.json b/records/sd1/sweep/784a55c6fdb74082f404/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f06023de1f65a90f5c7c3dba591a777950a5932
--- /dev/null
+++ b/records/sd1/sweep/784a55c6fdb74082f404/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "784a55c6fdb74082f404",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:11:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.95, 1.05] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/784a55c6fdb74082f404/workflow.json b/records/sd1/sweep/784a55c6fdb74082f404/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9419b2b5df2626ad83ea54decb9e3cb0d02bb67
--- /dev/null
+++ b/records/sd1/sweep/784a55c6fdb74082f404/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 784a55c6fdb74082f404)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/785f1310d9709172abec/measurements.json b/records/sd1/sweep/785f1310d9709172abec/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..86ff7d5be5cc8bd519ccf33566fa553fd6463083
--- /dev/null
+++ b/records/sd1/sweep/785f1310d9709172abec/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "785f1310d9709172abec",
+ "values": {
+  "std": {
+   "value": 32.735,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02498,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.156,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.156964,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.307008,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.140394,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.05461,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9525,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3788,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/785f1310d9709172abec/record.json b/records/sd1/sweep/785f1310d9709172abec/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..24c0701d29b194469752e57fa82c718004e021d4
--- /dev/null
+++ b/records/sd1/sweep/785f1310d9709172abec/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "785f1310d9709172abec",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:53:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "damp",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "multiply",
+   0.5,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/785f1310d9709172abec/workflow.json b/records/sd1/sweep/785f1310d9709172abec/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a20d8157eb1073dd75edcbc710d72596eec084f3
--- /dev/null
+++ b/records/sd1/sweep/785f1310d9709172abec/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 785f1310d9709172abec)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/786030205ecfe7b6b827/measurements.json b/records/sd1/sweep/786030205ecfe7b6b827/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..63e4266c102df146e8585a92fa1c838ac414fbc7
--- /dev/null
+++ b/records/sd1/sweep/786030205ecfe7b6b827/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "786030205ecfe7b6b827",
+ "values": {
+  "cosine_distance": {
+   "value": 0.25502,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.363929,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.182607,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.165862,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.82,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00694,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.199,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9137,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5921,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/786030205ecfe7b6b827/record.json b/records/sd1/sweep/786030205ecfe7b6b827/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e274f74fcab61b4adfd40444c5828fcfdc3aa63
--- /dev/null
+++ b/records/sd1/sweep/786030205ecfe7b6b827/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "786030205ecfe7b6b827",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5e74817c23d5b2d7759d84a62c2c4db46d7878f8e5ae1dd866ce87c569ee5fb6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5e74817c23d5b2d7759d84a62c2c4db46d7878f8e5ae1dd866ce87c569ee5fb6",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c6d8379c8f8707e090cb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/786030205ecfe7b6b827/workflow.json b/records/sd1/sweep/786030205ecfe7b6b827/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c06d77481f8b3383e1aa636bbd69c30da1ba001b
--- /dev/null
+++ b/records/sd1/sweep/786030205ecfe7b6b827/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 786030205ecfe7b6b827)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7864e87287f500f7e560/interpretations.jsonl b/records/sd1/sweep/7864e87287f500f7e560/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..8e9b4131ad6a055264919aa65ee79523563a3f02
--- /dev/null
+++ b/records/sd1/sweep/7864e87287f500f7e560/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7864e87287f500f7e560"}, "kind": "caption", "value": "Short-haired woman in black shirt, vintage portrait nearly identical to the original", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "de1693b557f97f46"}
+{"target": {"record": "7864e87287f500f7e560"}, "kind": "change", "value": "Almost no visible change; small differences in lip tint and skin texture.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "49feee8408ba7a9e"}
+{"target": {"record": "7864e87287f500f7e560"}, "kind": "keywords", "value": ["portrait", "near identical", "vintage photo", "minimal change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "102e8401ae4574a1"}
+{"target": {"record": "7864e87287f500f7e560"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4d10c9b5d9d34f51"}
diff --git a/records/sd1/sweep/7864e87287f500f7e560/measurements.json b/records/sd1/sweep/7864e87287f500f7e560/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a872476970e981aac5a13485d6f65f49aad0731
--- /dev/null
+++ b/records/sd1/sweep/7864e87287f500f7e560/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7864e87287f500f7e560",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7864e87287f500f7e560/record.json b/records/sd1/sweep/7864e87287f500f7e560/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..79dcb190291cabbc603cf1b2e294e81d4c2c16a3
--- /dev/null
+++ b/records/sd1/sweep/7864e87287f500f7e560/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7864e87287f500f7e560",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/eab368d1ea7ae6a599c483e930117a1fcdcd0bd9233108cb193d53427266242a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "eab368d1ea7ae6a599c483e930117a1fcdcd0bd9233108cb193d53427266242a",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "2e538a67a8948485f5c9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7864e87287f500f7e560/workflow.json b/records/sd1/sweep/7864e87287f500f7e560/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d904b663a5b2df7729c0e1aa2a9ab199aca56c2
--- /dev/null
+++ b/records/sd1/sweep/7864e87287f500f7e560/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7864e87287f500f7e560)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.h_upd\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/786b72edfc5b8969003c/measurements.json b/records/sd1/sweep/786b72edfc5b8969003c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..151e31cc4e249490610fecd4a55023d7fc181afa
--- /dev/null
+++ b/records/sd1/sweep/786b72edfc5b8969003c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "786b72edfc5b8969003c",
+ "values": {
+  "std": {
+   "value": 34.71,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01187,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.745,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.352748,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.562025,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.462835,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.110975,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9873,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.009,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/786b72edfc5b8969003c/record.json b/records/sd1/sweep/786b72edfc5b8969003c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..991f572daad9b01d39a4c9adc3cd49ba91d88d5d
--- /dev/null
+++ b/records/sd1/sweep/786b72edfc5b8969003c/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "786b72edfc5b8969003c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:37:16Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "half",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "rotate",
+   180.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c07f1d3518d6530030c9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/786b72edfc5b8969003c/workflow.json b/records/sd1/sweep/786b72edfc5b8969003c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e1abc2f85eb98e9dcd30b1488f78806de1ac7c0
--- /dev/null
+++ b/records/sd1/sweep/786b72edfc5b8969003c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 786b72edfc5b8969003c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7870ef8ff992c2660db4/measurements.json b/records/sd1/sweep/7870ef8ff992c2660db4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0210d3fa869c61f140ddcaa47939697dcecfaaf
--- /dev/null
+++ b/records/sd1/sweep/7870ef8ff992c2660db4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7870ef8ff992c2660db4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.110643,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.138426,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.112126,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.088391,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.615,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00584,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.972,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.943,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.286,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7870ef8ff992c2660db4/record.json b/records/sd1/sweep/7870ef8ff992c2660db4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f1ba4ec82219653c35f90941ca18ed21a69a25f
--- /dev/null
+++ b/records/sd1/sweep/7870ef8ff992c2660db4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7870ef8ff992c2660db4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/50d6942df624d69b1e941d9a7c9e37cd20802f3db9148b31b666fdf52deba083.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "50d6942df624d69b1e941d9a7c9e37cd20802f3db9148b31b666fdf52deba083",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "8bf45687e5b07ef339e1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7870ef8ff992c2660db4/workflow.json b/records/sd1/sweep/7870ef8ff992c2660db4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6f7acb08af7394efba5f9c76ce1b061a4b3725f
--- /dev/null
+++ b/records/sd1/sweep/7870ef8ff992c2660db4/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7870ef8ff992c2660db4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.proj_out\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78801c1607303074208a/measurements.json b/records/sd1/sweep/78801c1607303074208a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..135a3e209c25e401ba3675abad017377933311e6
--- /dev/null
+++ b/records/sd1/sweep/78801c1607303074208a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78801c1607303074208a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0809,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.094137,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.086852,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.018921,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.857,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00463,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.719,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9892,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1076,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78801c1607303074208a/record.json b/records/sd1/sweep/78801c1607303074208a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd940aadc44ac331df79f39c16ba0af3156819a7
--- /dev/null
+++ b/records/sd1/sweep/78801c1607303074208a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "78801c1607303074208a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/a80f5e96267fc9c5f64b194a88a23a7d360cd084b2343cc355240ff5399ba61e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a80f5e96267fc9c5f64b194a88a23a7d360cd084b2343cc355240ff5399ba61e",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "74c8ef9037f71b5e34c8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78801c1607303074208a/workflow.json b/records/sd1/sweep/78801c1607303074208a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..28cb3f16d8ae690a621a0f735de908567ae93517
--- /dev/null
+++ b/records/sd1/sweep/78801c1607303074208a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78801c1607303074208a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78830710a66ae69b2737/measurements.json b/records/sd1/sweep/78830710a66ae69b2737/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fdb809a6d261fab485e585a8ce4d352851cbfc07
--- /dev/null
+++ b/records/sd1/sweep/78830710a66ae69b2737/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78830710a66ae69b2737",
+ "values": {
+  "std": {
+   "value": 42.925,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02204,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.535,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.258978,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.33638,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.13573,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.047248,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9336,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0882,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78830710a66ae69b2737/record.json b/records/sd1/sweep/78830710a66ae69b2737/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..348f971ccedfeef0a15bec7ec4ee61f8b35f6897
--- /dev/null
+++ b/records/sd1/sweep/78830710a66ae69b2737/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "78830710a66ae69b2737",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:17:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "damp",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   0.5,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4b76057a95930fcf7744"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78830710a66ae69b2737/workflow.json b/records/sd1/sweep/78830710a66ae69b2737/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed8fdec9770e1f837d0a53b98278c6f48b3e8be0
--- /dev/null
+++ b/records/sd1/sweep/78830710a66ae69b2737/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78830710a66ae69b2737)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/788482b0f55f3bb277c7/interpretations.jsonl b/records/sd1/sweep/788482b0f55f3bb277c7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..16e521df47c87acdf208c3281ee32e1e78edfc84
--- /dev/null
+++ b/records/sd1/sweep/788482b0f55f3bb277c7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "788482b0f55f3bb277c7"}, "kind": "caption", "value": "Painted portrait of a woman with dark bob, teal collar, on cream speckled background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:14:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "07291186e4a7f3d2"}
+{"target": {"record": "788482b0f55f3bb277c7"}, "kind": "change", "value": "The rendering shifts to a grainier painted style with darker hair, a teal undershirt showing under a black jacket, and a mottled cream-and-ochre background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:14:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "643a8079cfbfa65f"}
+{"target": {"record": "788482b0f55f3bb277c7"}, "kind": "keywords", "value": ["teal collar", "speckled background", "painted portrait", "dark bob", "grainy"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:14:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ebe8a916111bd6fe"}
+{"target": {"record": "788482b0f55f3bb277c7"}, "kind": "effect_tags", "value": ["painterly", "texture", "colour_shift", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:14:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5ec64b1e9af8b7dc"}
diff --git a/records/sd1/sweep/788482b0f55f3bb277c7/measurements.json b/records/sd1/sweep/788482b0f55f3bb277c7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7ffdb545dc3335991333c161c3132adf7133a10
--- /dev/null
+++ b/records/sd1/sweep/788482b0f55f3bb277c7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "788482b0f55f3bb277c7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.294214,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.342731,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.213018,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.052502,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.557,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00441,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.927,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9704,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0729,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/788482b0f55f3bb277c7/record.json b/records/sd1/sweep/788482b0f55f3bb277c7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45eee45f8a83d968f7d701040ecbd02724e0b2cc
--- /dev/null
+++ b/records/sd1/sweep/788482b0f55f3bb277c7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "788482b0f55f3bb277c7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/14261b62c69618091373b5894c8cf29b3a6f8154ca85e9020dac7242e89ce15d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "14261b62c69618091373b5894c8cf29b3a6f8154ca85e9020dac7242e89ce15d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "3ecc389757d81f075c83"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/788482b0f55f3bb277c7/workflow.json b/records/sd1/sweep/788482b0f55f3bb277c7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d7933d8183bdf801a7fd320073b0698860d9cfb
--- /dev/null
+++ b/records/sd1/sweep/788482b0f55f3bb277c7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 788482b0f55f3bb277c7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7887166e6beadf0e5eb4/interpretations.jsonl b/records/sd1/sweep/7887166e6beadf0e5eb4/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..4d8d6ad4100dabcc0dc521fcbe5cab39021de53d
--- /dev/null
+++ b/records/sd1/sweep/7887166e6beadf0e5eb4/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7887166e6beadf0e5eb4"}, "kind": "caption", "value": "Frontal young woman with shoulder-length light brown hair, blue eyes, dark sweater, cream backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:05:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a8ae5e8fb3f49d65"}
+{"target": {"record": "7887166e6beadf0e5eb4"}, "kind": "change", "value": "The pose becomes fully frontal and the subject is a different woman with longer, lighter hair and blue eyes, set against a warmer cream background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:05:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4bc346f92f8d0941"}
+{"target": {"record": "7887166e6beadf0e5eb4"}, "kind": "keywords", "value": ["frontal portrait", "new identity", "blue eyes", "light brown hair", "warm backdrop"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:05:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f68ed3b63a2fe6c7"}
+{"target": {"record": "7887166e6beadf0e5eb4"}, "kind": "effect_tags", "value": ["subject_kept", "colour_shift", "symmetry"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:05:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "829356fcc88ae657"}
diff --git a/records/sd1/sweep/7887166e6beadf0e5eb4/measurements.json b/records/sd1/sweep/7887166e6beadf0e5eb4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..079a4e26dc5983937c4eff6c495e342932d996d0
--- /dev/null
+++ b/records/sd1/sweep/7887166e6beadf0e5eb4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7887166e6beadf0e5eb4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.411235,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.469739,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.407863,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.114654,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.659,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00648,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.678,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0825,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2168,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7887166e6beadf0e5eb4/record.json b/records/sd1/sweep/7887166e6beadf0e5eb4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7694722d8a43c69badc1a8d8959a1d068da600d4
--- /dev/null
+++ b/records/sd1/sweep/7887166e6beadf0e5eb4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7887166e6beadf0e5eb4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:12Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/3087af74070d8ff49050408c98f34de947cbdbf7ac4c4349e6995aa50873a7bf.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3087af74070d8ff49050408c98f34de947cbdbf7ac4c4349e6995aa50873a7bf",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c581fbe4e0e0b940e5cd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7887166e6beadf0e5eb4/workflow.json b/records/sd1/sweep/7887166e6beadf0e5eb4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bece628d83fad733ba224ca4029788d9237cef8
--- /dev/null
+++ b/records/sd1/sweep/7887166e6beadf0e5eb4/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7887166e6beadf0e5eb4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/788770fac356ffd8faf3/measurements.json b/records/sd1/sweep/788770fac356ffd8faf3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..50da5e93b02a614075126fd903c8232e73d25c4b
--- /dev/null
+++ b/records/sd1/sweep/788770fac356ffd8faf3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "788770fac356ffd8faf3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.322413,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.35759,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.147189,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.057084,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.678,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00329,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.975,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9109,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2661,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/788770fac356ffd8faf3/record.json b/records/sd1/sweep/788770fac356ffd8faf3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f8f3682bfabe8f6a1dd68112d89641598434fea
--- /dev/null
+++ b/records/sd1/sweep/788770fac356ffd8faf3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "788770fac356ffd8faf3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:43Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/544cb18066a5230f54dea2145215454f8b44caa55a1992f93d3d40d4d7232811.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "544cb18066a5230f54dea2145215454f8b44caa55a1992f93d3d40d4d7232811",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e4d58035924cc0c136f2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/788770fac356ffd8faf3/workflow.json b/records/sd1/sweep/788770fac356ffd8faf3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..82ccb46872351717f392cc0de4b9a15625f1cb10
--- /dev/null
+++ b/records/sd1/sweep/788770fac356ffd8faf3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 788770fac356ffd8faf3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7889d5419c692f224d0e/measurements.json b/records/sd1/sweep/7889d5419c692f224d0e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ed27b8561306f579947160b7b4785c274883fec
--- /dev/null
+++ b/records/sd1/sweep/7889d5419c692f224d0e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7889d5419c692f224d0e",
+ "values": {
+  "std": {
+   "value": 31.82,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01954,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.604,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.479166,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.60399,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.68927,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.159627,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0407,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0272,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7889d5419c692f224d0e/record.json b/records/sd1/sweep/7889d5419c692f224d0e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..60f352ee1bc60b42f41e610e56ba5db2bc0dfce6
--- /dev/null
+++ b/records/sd1/sweep/7889d5419c692f224d0e/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7889d5419c692f224d0e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:07:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "half",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "rotate",
+   180.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "875db98bb7c397aeab16"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7889d5419c692f224d0e/workflow.json b/records/sd1/sweep/7889d5419c692f224d0e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a86f1f772530a854362aefc9ea6f5a5cfc8d7ccd
--- /dev/null
+++ b/records/sd1/sweep/7889d5419c692f224d0e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7889d5419c692f224d0e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78960332676d22746b35/measurements.json b/records/sd1/sweep/78960332676d22746b35/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6027ebd7f8563a62cf8c77363cb5d63eb0fd424e
--- /dev/null
+++ b/records/sd1/sweep/78960332676d22746b35/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78960332676d22746b35",
+ "values": {
+  "std": {
+   "value": 43.796,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0168,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 3.049,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.005011,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.016587,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.010396,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00701,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9785,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0641,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78960332676d22746b35/record.json b/records/sd1/sweep/78960332676d22746b35/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cc18f6f89fecc7b2ccff3fc29d0b665c1951d64
--- /dev/null
+++ b/records/sd1/sweep/78960332676d22746b35/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "78960332676d22746b35",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:54:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_scalar",
+   -0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5ca18162d40033345020"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78960332676d22746b35/workflow.json b/records/sd1/sweep/78960332676d22746b35/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..69735ca1b619880d4f41a4d3768b048fac1ab3ea
--- /dev/null
+++ b/records/sd1/sweep/78960332676d22746b35/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78960332676d22746b35)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/789a0735e21e271389b0/measurements.json b/records/sd1/sweep/789a0735e21e271389b0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f096f1a9262ff83e1d0679fd71ba437529a0491
--- /dev/null
+++ b/records/sd1/sweep/789a0735e21e271389b0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "789a0735e21e271389b0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.113926,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.122841,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.118399,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.032474,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.378,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00401,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.217,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0022,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1942,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/789a0735e21e271389b0/record.json b/records/sd1/sweep/789a0735e21e271389b0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4b4315cadf9b8ae4680f51bd7a64a3fd5ece1998
--- /dev/null
+++ b/records/sd1/sweep/789a0735e21e271389b0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "789a0735e21e271389b0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/50e6fc0467fda306a8b39718aa8f0a3551aa89e0ceee75028e7549bd3f1cbca7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "50e6fc0467fda306a8b39718aa8f0a3551aa89e0ceee75028e7549bd3f1cbca7",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8949887c2495c8caf19c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/789a0735e21e271389b0/workflow.json b/records/sd1/sweep/789a0735e21e271389b0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..89845ae5bb6bf31769c613a562ee546f647cebd1
--- /dev/null
+++ b/records/sd1/sweep/789a0735e21e271389b0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 789a0735e21e271389b0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/789ce78491ce0e8e0191/measurements.json b/records/sd1/sweep/789ce78491ce0e8e0191/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9e4fec3a0287b71b3cce99da5995f9b45fe86b9a
--- /dev/null
+++ b/records/sd1/sweep/789ce78491ce0e8e0191/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "789ce78491ce0e8e0191",
+ "values": {
+  "cosine_distance": {
+   "value": 0.010896,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.024028,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.009838,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003577,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.109,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00604,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.734,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0189,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1096,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/789ce78491ce0e8e0191/record.json b/records/sd1/sweep/789ce78491ce0e8e0191/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5b95a711eba12311428480b727eab6a26062d78
--- /dev/null
+++ b/records/sd1/sweep/789ce78491ce0e8e0191/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "789ce78491ce0e8e0191",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:00Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/f1264d0ead4a16ef03ac2208a42d749331d8b19e657574b3ef25b042e7634c0b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "f1264d0ead4a16ef03ac2208a42d749331d8b19e657574b3ef25b042e7634c0b",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2c617c94455a7e1dbce1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/789ce78491ce0e8e0191/workflow.json b/records/sd1/sweep/789ce78491ce0e8e0191/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..394db7ba254966be176bf7247f64220b50e364ba
--- /dev/null
+++ b/records/sd1/sweep/789ce78491ce0e8e0191/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 789ce78491ce0e8e0191)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.in_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78a7d8d4424de4df48bc/measurements.json b/records/sd1/sweep/78a7d8d4424de4df48bc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8eacadc7488c75a1f7845c4aa32d1eeed7f908a5
--- /dev/null
+++ b/records/sd1/sweep/78a7d8d4424de4df48bc/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78a7d8d4424de4df48bc",
+ "values": {
+  "std": {
+   "value": 95.191,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.003,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 91.012,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.867003,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.806738,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.675342,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.344806,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7636,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8399,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78a7d8d4424de4df48bc/record.json b/records/sd1/sweep/78a7d8d4424de4df48bc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7552ffe5c026d6a04c80398ebb8b0a1325121c5a
--- /dev/null
+++ b/records/sd1/sweep/78a7d8d4424de4df48bc/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "78a7d8d4424de4df48bc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:56:12Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "quarter",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   90.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [0.0, 5.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "703ad17c4e43ef74941c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78a7d8d4424de4df48bc/workflow.json b/records/sd1/sweep/78a7d8d4424de4df48bc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e57cad1b99a69301976cd324a08cb302ed0dfbc
--- /dev/null
+++ b/records/sd1/sweep/78a7d8d4424de4df48bc/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78a7d8d4424de4df48bc)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78b49303d094456cabb8/interpretations.jsonl b/records/sd1/sweep/78b49303d094456cabb8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..38a31f2b488c2469ec02db782fe6e32164edce58
--- /dev/null
+++ b/records/sd1/sweep/78b49303d094456cabb8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "78b49303d094456cabb8"}, "kind": "caption", "value": "Illustrated portrait of a woman with wavy hair and black turtleneck, unchanged", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:21:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "173c31b045f28260"}
+{"target": {"record": "78b49303d094456cabb8"}, "kind": "change", "value": "Almost no visible change.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:21:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a944681b68043ad4"}
+{"target": {"record": "78b49303d094456cabb8"}, "kind": "keywords", "value": ["no change", "portrait", "identical"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:21:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8bb64b7f2012a7bc"}
+{"target": {"record": "78b49303d094456cabb8"}, "kind": "effect_tags", "value": ["subject_kept", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:21:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "74003a23ac5846bb"}
diff --git a/records/sd1/sweep/78b49303d094456cabb8/measurements.json b/records/sd1/sweep/78b49303d094456cabb8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f967f613fe33dbd0ad2ba242d83bf7652e116ea4
--- /dev/null
+++ b/records/sd1/sweep/78b49303d094456cabb8/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78b49303d094456cabb8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.618,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00534,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1553,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78b49303d094456cabb8/record.json b/records/sd1/sweep/78b49303d094456cabb8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..83610eea4762e8f16fd3da0094c9594394dc6c14
--- /dev/null
+++ b/records/sd1/sweep/78b49303d094456cabb8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "78b49303d094456cabb8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5cb450ddece4dd4ae2ad944731d0c8b3c19f828f6250b5a8acf810ec295766eb.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5cb450ddece4dd4ae2ad944731d0c8b3c19f828f6250b5a8acf810ec295766eb",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2acb28c9876e9c73138c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78b49303d094456cabb8/workflow.json b/records/sd1/sweep/78b49303d094456cabb8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d03d1b500b32079480af11317d7e42897e0b3f1e
--- /dev/null
+++ b/records/sd1/sweep/78b49303d094456cabb8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78b49303d094456cabb8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78bcd46501b2131e6b5e/measurements.json b/records/sd1/sweep/78bcd46501b2131e6b5e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee541bfc085dffd0adfcbe2ac4117ee377f7b655
--- /dev/null
+++ b/records/sd1/sweep/78bcd46501b2131e6b5e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78bcd46501b2131e6b5e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.457417,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.533796,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.349633,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.142158,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.713,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0049,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.309,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9578,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3257,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78bcd46501b2131e6b5e/record.json b/records/sd1/sweep/78bcd46501b2131e6b5e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ceb9404d9f58c362c0c55d6f587200b396997eb
--- /dev/null
+++ b/records/sd1/sweep/78bcd46501b2131e6b5e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78bcd46501b2131e6b5e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.out_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ad4c68b600cc567f5009f12e4d0cc559d937e2063664a67809a863c5bdcc4be5.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ad4c68b600cc567f5009f12e4d0cc559d937e2063664a67809a863c5bdcc4be5",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "9eef5811a2e7a86651df"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78bcd46501b2131e6b5e/workflow.json b/records/sd1/sweep/78bcd46501b2131e6b5e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..920d09d8b43ad9919148a6d99f4ddfa382247c2e
--- /dev/null
+++ b/records/sd1/sweep/78bcd46501b2131e6b5e/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78bcd46501b2131e6b5e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78bfbad9dfc13a37465e/measurements.json b/records/sd1/sweep/78bfbad9dfc13a37465e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4884e4aa654ecf92e75ca6ab36ba0904f7b54b40
--- /dev/null
+++ b/records/sd1/sweep/78bfbad9dfc13a37465e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78bfbad9dfc13a37465e",
+ "values": {
+  "std": {
+   "value": 29.84,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02893,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 1.757,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001158,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.005,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004648,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002174,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0062,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1896,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78bfbad9dfc13a37465e/record.json b/records/sd1/sweep/78bfbad9dfc13a37465e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5e89d6aa949ae86e84e54462162531cb944766a
--- /dev/null
+++ b/records/sd1/sweep/78bfbad9dfc13a37465e/record.json
@@ -0,0 +1,86 @@
+{
+ "id": "78bfbad9dfc13a37465e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:02:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "medium",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_noise",
+   0.2,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "96886869233ee93390bd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78bfbad9dfc13a37465e/workflow.json b/records/sd1/sweep/78bfbad9dfc13a37465e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2db1452dd7c16fe6959400c79d1ee4fe850e049
--- /dev/null
+++ b/records/sd1/sweep/78bfbad9dfc13a37465e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78bfbad9dfc13a37465e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d074cab027ede997d0/measurements.json b/records/sd1/sweep/78d074cab027ede997d0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..775248cdc6898fa8bf7cdcbf1e40e2aaa7160926
--- /dev/null
+++ b/records/sd1/sweep/78d074cab027ede997d0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78d074cab027ede997d0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.322376,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.384922,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.399119,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.246393,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.93,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01276,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.956,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9697,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4821,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d074cab027ede997d0/record.json b/records/sd1/sweep/78d074cab027ede997d0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0231b2956c8e78476e9633667dccd1a51ac5b3f5
--- /dev/null
+++ b/records/sd1/sweep/78d074cab027ede997d0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78d074cab027ede997d0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/7018ebc2c1d994b15546649acf4f6b20a60eea9e5c3a430f0a1f3603cdb56064.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7018ebc2c1d994b15546649acf4f6b20a60eea9e5c3a430f0a1f3603cdb56064",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "4d3526df312e734e810d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d074cab027ede997d0/workflow.json b/records/sd1/sweep/78d074cab027ede997d0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d42776082a8668599364f782ccb2f0e6300b8ee1
--- /dev/null
+++ b/records/sd1/sweep/78d074cab027ede997d0/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78d074cab027ede997d0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d56c2aa3c081c35986/measurements.json b/records/sd1/sweep/78d56c2aa3c081c35986/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..55c26df07ea16602beda809c37105e13aabe0e19
--- /dev/null
+++ b/records/sd1/sweep/78d56c2aa3c081c35986/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78d56c2aa3c081c35986",
+ "values": {
+  "cosine_distance": {
+   "value": 0.243511,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.257468,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.127829,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.092122,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.674,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00355,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 27.892,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0546,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0857,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d56c2aa3c081c35986/record.json b/records/sd1/sweep/78d56c2aa3c081c35986/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3cc4b4786c5a5f47bd00f0c5119c84546a2d116c
--- /dev/null
+++ b/records/sd1/sweep/78d56c2aa3c081c35986/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78d56c2aa3c081c35986",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:15Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/9cce99bd14d0d20010321cb82f803fc37ce79dccdb860b3d41ef4689d0c02e0b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9cce99bd14d0d20010321cb82f803fc37ce79dccdb860b3d41ef4689d0c02e0b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "41ae32ac4f3a98721271"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d56c2aa3c081c35986/workflow.json b/records/sd1/sweep/78d56c2aa3c081c35986/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bfe30a61c40812baf0768d0e500a22b327fd4f57
--- /dev/null
+++ b/records/sd1/sweep/78d56c2aa3c081c35986/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78d56c2aa3c081c35986)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d91444b50ff67be290/measurements.json b/records/sd1/sweep/78d91444b50ff67be290/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c63194280be85eaf88b6c4eaa163082fbe004a16
--- /dev/null
+++ b/records/sd1/sweep/78d91444b50ff67be290/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78d91444b50ff67be290",
+ "values": {
+  "cosine_distance": {
+   "value": 0.335352,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.460475,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.510145,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.207791,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.291,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00308,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 34.275,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0479,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1392,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d91444b50ff67be290/record.json b/records/sd1/sweep/78d91444b50ff67be290/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b2d36ef951b17556d03a41a06a868e31966bbd7
--- /dev/null
+++ b/records/sd1/sweep/78d91444b50ff67be290/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78d91444b50ff67be290",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:44Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ba59ee9b2c115fa9a8349bdb21d71337dda9697cbdc4810f7872bc214c2d0be4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ba59ee9b2c115fa9a8349bdb21d71337dda9697cbdc4810f7872bc214c2d0be4",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "71d25172c6265cdc2479"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d91444b50ff67be290/workflow.json b/records/sd1/sweep/78d91444b50ff67be290/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8d4a6cd53f36ac0f040119f913768caeaa5d21c
--- /dev/null
+++ b/records/sd1/sweep/78d91444b50ff67be290/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78d91444b50ff67be290)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d9c13115cb957e2ea5/measurements.json b/records/sd1/sweep/78d9c13115cb957e2ea5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..303c09f9cd58b9fad18d7f4d816d0b659fdca407
--- /dev/null
+++ b/records/sd1/sweep/78d9c13115cb957e2ea5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78d9c13115cb957e2ea5",
+ "values": {
+  "std": {
+   "value": 37.708,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01464,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.624,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000141,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.000947,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000275,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 9.2e-05,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0017,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0139,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d9c13115cb957e2ea5/record.json b/records/sd1/sweep/78d9c13115cb957e2ea5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3922362c9ebd69809e8b7c004a99cb5cef1d3f7c
--- /dev/null
+++ b/records/sd1/sweep/78d9c13115cb957e2ea5/record.json
@@ -0,0 +1,86 @@
+{
+ "id": "78d9c13115cb957e2ea5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:14:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "medium",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_noise",
+   0.2,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for out.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0ee3cf18255a0dcaaabd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78d9c13115cb957e2ea5/workflow.json b/records/sd1/sweep/78d9c13115cb957e2ea5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..66f6ee17c113d8c9ec55390f6755dc289563f13a
--- /dev/null
+++ b/records/sd1/sweep/78d9c13115cb957e2ea5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78d9c13115cb957e2ea5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 1234\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78dff70e5e2c1f5c4dac/interpretations.jsonl b/records/sd1/sweep/78dff70e5e2c1f5c4dac/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6d62f0aaee6f81854b7fd24ba137e4dd512928ed
--- /dev/null
+++ b/records/sd1/sweep/78dff70e5e2c1f5c4dac/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "78dff70e5e2c1f5c4dac"}, "kind": "caption", "value": "Painted woman with wavy dark hair and black turtleneck on a grey textured panel", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "905852de8aa2a883"}
+{"target": {"record": "78dff70e5e2c1f5c4dac"}, "kind": "change", "value": "Almost no visible change.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e69a9a7d19298844"}
+{"target": {"record": "78dff70e5e2c1f5c4dac"}, "kind": "keywords", "value": ["portrait", "near identical", "turtleneck"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2843d1ec8f773796"}
+{"target": {"record": "78dff70e5e2c1f5c4dac"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "eddc029cca919014"}
diff --git a/records/sd1/sweep/78dff70e5e2c1f5c4dac/measurements.json b/records/sd1/sweep/78dff70e5e2c1f5c4dac/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ffebc265c452ac84378d681dd8296c0bb7627cc
--- /dev/null
+++ b/records/sd1/sweep/78dff70e5e2c1f5c4dac/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78dff70e5e2c1f5c4dac",
+ "values": {
+  "cosine_distance": {
+   "value": 0.021319,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.042615,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.009283,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002003,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.785,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0068,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.137,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9951,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1054,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78dff70e5e2c1f5c4dac/record.json b/records/sd1/sweep/78dff70e5e2c1f5c4dac/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c00028a18a44a43413f5c68931d427c3c499d303
--- /dev/null
+++ b/records/sd1/sweep/78dff70e5e2c1f5c4dac/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "78dff70e5e2c1f5c4dac",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/f71eb11b5dd674720ca72e2248eb125496a83ba085a1b7c42f5c031d126a29fc.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "f71eb11b5dd674720ca72e2248eb125496a83ba085a1b7c42f5c031d126a29fc",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "eb7296745491ea51a487"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78dff70e5e2c1f5c4dac/workflow.json b/records/sd1/sweep/78dff70e5e2c1f5c4dac/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ea1dc6c30cd792f5788443b45911b914577d407
--- /dev/null
+++ b/records/sd1/sweep/78dff70e5e2c1f5c4dac/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78dff70e5e2c1f5c4dac)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78e2776e005642e03506/measurements.json b/records/sd1/sweep/78e2776e005642e03506/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b4f5a33c91443b1975d213267ab20c24d26479f
--- /dev/null
+++ b/records/sd1/sweep/78e2776e005642e03506/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78e2776e005642e03506",
+ "values": {
+  "cosine_distance": {
+   "value": 0.178642,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.215571,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.11788,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.076937,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.806,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00424,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.711,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9295,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.509,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78e2776e005642e03506/record.json b/records/sd1/sweep/78e2776e005642e03506/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..97701ede5c0c3f0408c459c37f71cfa8eb1f5ae5
--- /dev/null
+++ b/records/sd1/sweep/78e2776e005642e03506/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78e2776e005642e03506",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.in_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/34beb07314cb96d986eefac6e60a7511dd11742cf9411e7e5c4db6edcee7d162.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "34beb07314cb96d986eefac6e60a7511dd11742cf9411e7e5c4db6edcee7d162",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "e11405ca75552dd2c931"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78e2776e005642e03506/workflow.json b/records/sd1/sweep/78e2776e005642e03506/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e5cd21f9c37f55837ee4fb4c011903d5349ac28
--- /dev/null
+++ b/records/sd1/sweep/78e2776e005642e03506/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78e2776e005642e03506)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.in_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78e8c951d4ca1c663294/measurements.json b/records/sd1/sweep/78e8c951d4ca1c663294/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..750f36c8a8998002d06ca979aaa7937e6f21fd91
--- /dev/null
+++ b/records/sd1/sweep/78e8c951d4ca1c663294/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78e8c951d4ca1c663294",
+ "values": {
+  "std": {
+   "value": 32.555,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00894,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 55.99,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.645521,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.658568,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.564265,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.127581,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9951,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0528,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78e8c951d4ca1c663294/record.json b/records/sd1/sweep/78e8c951d4ca1c663294/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..01091f4c1817c0c4c4f4023a2a6ec7c198aad099
--- /dev/null
+++ b/records/sd1/sweep/78e8c951d4ca1c663294/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "78e8c951d4ca1c663294",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:18:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "quarter",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   90.0,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6acce7b6dc5647fa3c00"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78e8c951d4ca1c663294/workflow.json b/records/sd1/sweep/78e8c951d4ca1c663294/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce5a0cda92d1c2eb0eb462589f04eb5f37f8e205
--- /dev/null
+++ b/records/sd1/sweep/78e8c951d4ca1c663294/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78e8c951d4ca1c663294)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78f977af5496d8dbf7a1/measurements.json b/records/sd1/sweep/78f977af5496d8dbf7a1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac15970d152fc7645bb524f42bf1dfcd87fe9c45
--- /dev/null
+++ b/records/sd1/sweep/78f977af5496d8dbf7a1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "78f977af5496d8dbf7a1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.367604,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.402554,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.3098,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.127529,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.007,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00625,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.58,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.052,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0663,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78f977af5496d8dbf7a1/record.json b/records/sd1/sweep/78f977af5496d8dbf7a1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..95c4f89049b0ae595022f4e77c39cad2abd15f06
--- /dev/null
+++ b/records/sd1/sweep/78f977af5496d8dbf7a1/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78f977af5496d8dbf7a1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.proj_in",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "in.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/facf30a3339d9f945942229fe907f545a3d682b0236acfb70671b305decf44cb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "facf30a3339d9f945942229fe907f545a3d682b0236acfb70671b305decf44cb",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "bc9174c7b3705f8534cd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78f977af5496d8dbf7a1/workflow.json b/records/sd1/sweep/78f977af5496d8dbf7a1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..15b8381ef8454540682ed039da46f1450aec8d8e
--- /dev/null
+++ b/records/sd1/sweep/78f977af5496d8dbf7a1/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78f977af5496d8dbf7a1)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.proj_in\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78fe49d39543b538edc5/measurements.json b/records/sd1/sweep/78fe49d39543b538edc5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..57642d81f97599b804369b7efeb195fbfcadad68
--- /dev/null
+++ b/records/sd1/sweep/78fe49d39543b538edc5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "78fe49d39543b538edc5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.096952,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.159212,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.115986,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.094288,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.852,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00336,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.274,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0126,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4409,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78fe49d39543b538edc5/record.json b/records/sd1/sweep/78fe49d39543b538edc5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..49b48b5c3436be33ef45462e0fbc2f73e4797e7b
--- /dev/null
+++ b/records/sd1/sweep/78fe49d39543b538edc5/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "78fe49d39543b538edc5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ed850a08b13a19445a6cc8882242f3c4b3ff9ac6cbf50571cd831eacaecf00ad.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ed850a08b13a19445a6cc8882242f3c4b3ff9ac6cbf50571cd831eacaecf00ad",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "174c1235c73638d3ff8e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/78fe49d39543b538edc5/workflow.json b/records/sd1/sweep/78fe49d39543b538edc5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f728f962878fe7b22e06d87b83774cb94e0cdefc
--- /dev/null
+++ b/records/sd1/sweep/78fe49d39543b538edc5/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 78fe49d39543b538edc5)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7903370ff477d98c38c6/measurements.json b/records/sd1/sweep/7903370ff477d98c38c6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..75ca2d7a2659bd95ffd396b11fcbf2f88b4a4489
--- /dev/null
+++ b/records/sd1/sweep/7903370ff477d98c38c6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7903370ff477d98c38c6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.090666,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.096569,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.030491,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02998,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.843,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00571,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.921,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9886,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0745,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7903370ff477d98c38c6/record.json b/records/sd1/sweep/7903370ff477d98c38c6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbe4ae720aca98858d626bba22c6cd952c841db4
--- /dev/null
+++ b/records/sd1/sweep/7903370ff477d98c38c6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7903370ff477d98c38c6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/75dcfa0d69f6b26bcbca940426ecae79e84cb63404f32528cd2d01aa7c14453e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "75dcfa0d69f6b26bcbca940426ecae79e84cb63404f32528cd2d01aa7c14453e",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "11a3b5e5bbb1e5023c23"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7903370ff477d98c38c6/workflow.json b/records/sd1/sweep/7903370ff477d98c38c6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7807271ef8b93f4326461ba451192664c82cf610
--- /dev/null
+++ b/records/sd1/sweep/7903370ff477d98c38c6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7903370ff477d98c38c6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7905f6af300dae548f55/measurements.json b/records/sd1/sweep/7905f6af300dae548f55/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bad6aad0026245e163f1c876789d85580d68a39
--- /dev/null
+++ b/records/sd1/sweep/7905f6af300dae548f55/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7905f6af300dae548f55",
+ "values": {
+  "cosine_distance": {
+   "value": 0.175632,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.276245,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.061941,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.030223,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.953,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00333,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.318,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.999,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1479,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7905f6af300dae548f55/record.json b/records/sd1/sweep/7905f6af300dae548f55/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9db294a7c3239b88eea4bf96c9f292ebe6b5fc5a
--- /dev/null
+++ b/records/sd1/sweep/7905f6af300dae548f55/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7905f6af300dae548f55",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.mid",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/03752f3f254a89dca3e47651f20ef0b05d94e559ed50f3f9e6dd67b83de7e7d7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "03752f3f254a89dca3e47651f20ef0b05d94e559ed50f3f9e6dd67b83de7e7d7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6ad6ae1c36116b19e7bf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7905f6af300dae548f55/workflow.json b/records/sd1/sweep/7905f6af300dae548f55/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5b5d1a4be285cda192cbd6d58873b68d045a5d0
--- /dev/null
+++ b/records/sd1/sweep/7905f6af300dae548f55/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7905f6af300dae548f55)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/790ff2e8b792c1d96715/measurements.json b/records/sd1/sweep/790ff2e8b792c1d96715/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f25e9580081c943f61430b7717af7818a708ee0
--- /dev/null
+++ b/records/sd1/sweep/790ff2e8b792c1d96715/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "790ff2e8b792c1d96715",
+ "values": {
+  "std": {
+   "value": 50.32,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01026,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.295,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.017206,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.060556,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038586,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00694,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9739,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0129,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/790ff2e8b792c1d96715/record.json b/records/sd1/sweep/790ff2e8b792c1d96715/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..28c334d3c4f0b909acfcec27a4a0399e424daa26
--- /dev/null
+++ b/records/sd1/sweep/790ff2e8b792c1d96715/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "790ff2e8b792c1d96715",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:47:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "mid",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_scalar",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/790ff2e8b792c1d96715/workflow.json b/records/sd1/sweep/790ff2e8b792c1d96715/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b72ba15685e3859122211769c5d6a34c370db94
--- /dev/null
+++ b/records/sd1/sweep/790ff2e8b792c1d96715/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 790ff2e8b792c1d96715)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79121598e1083a5625bf/measurements.json b/records/sd1/sweep/79121598e1083a5625bf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ccc3f1018baec1c96ec11c2f5b5fbc01a8eafb0
--- /dev/null
+++ b/records/sd1/sweep/79121598e1083a5625bf/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "79121598e1083a5625bf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.071117,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.101397,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.069783,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.024867,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.908,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00583,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.693,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0389,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.119,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79121598e1083a5625bf/record.json b/records/sd1/sweep/79121598e1083a5625bf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8742c5881de0598ea97f86f3ddf76774d446a62
--- /dev/null
+++ b/records/sd1/sweep/79121598e1083a5625bf/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79121598e1083a5625bf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/41d9942ed6e400d080f97a5b90580a81c9e919a4d23ace8f1167c720c0c4d54a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "41d9942ed6e400d080f97a5b90580a81c9e919a4d23ace8f1167c720c0c4d54a",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ba478eaa172747eaf213"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79121598e1083a5625bf/workflow.json b/records/sd1/sweep/79121598e1083a5625bf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..701bcba01ab07809e4bd269c8976508aa9728841
--- /dev/null
+++ b/records/sd1/sweep/79121598e1083a5625bf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79121598e1083a5625bf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79123f11e03b81473def/measurements.json b/records/sd1/sweep/79123f11e03b81473def/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b22e5122e2c84aeb924ee14261f0d3a025504f2
--- /dev/null
+++ b/records/sd1/sweep/79123f11e03b81473def/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79123f11e03b81473def",
+ "values": {
+  "cosine_distance": {
+   "value": 0.468942,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.687927,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.755593,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.315168,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 37.259,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00277,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.996,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.946,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.329,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79123f11e03b81473def/record.json b/records/sd1/sweep/79123f11e03b81473def/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5e39f4347d445e2437fcfb23f88350b5f953290
--- /dev/null
+++ b/records/sd1/sweep/79123f11e03b81473def/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79123f11e03b81473def",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.2.conv",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "upsample",
+   "module_type": "Conv2d",
+   "container": "Upsample",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/67c113ef2ad2f5c41610e44373f806accb643e61bafc37f11a323f4ae0966a44.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "67c113ef2ad2f5c41610e44373f806accb643e61bafc37f11a323f4ae0966a44",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.Upsample.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "9f9b7310051441b660dd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79123f11e03b81473def/workflow.json b/records/sd1/sweep/79123f11e03b81473def/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ddbd91f5aafeb31717d264e0abb9d25c97b13645
--- /dev/null
+++ b/records/sd1/sweep/79123f11e03b81473def/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79123f11e03b81473def)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.2.conv\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79140fa4073bc441db75/measurements.json b/records/sd1/sweep/79140fa4073bc441db75/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..065cd7fb85ae3d5bc84c6b1add1c5da697decff4
--- /dev/null
+++ b/records/sd1/sweep/79140fa4073bc441db75/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79140fa4073bc441db75",
+ "values": {
+  "cosine_distance": {
+   "value": 0.46082,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.669127,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.505996,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.359244,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.584,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.06019,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 56.17,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9271,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1262,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79140fa4073bc441db75/record.json b/records/sd1/sweep/79140fa4073bc441db75/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6e02454b23b4bc4906dab8a4a5da0d11f1a62bf
--- /dev/null
+++ b/records/sd1/sweep/79140fa4073bc441db75/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79140fa4073bc441db75",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:00Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.norm",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/a730bceff256015b3c297564fbd003be3c59ab8ac840f09c1643d65e3c476d5e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a730bceff256015b3c297564fbd003be3c59ab8ac840f09c1643d65e3c476d5e",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "448a162b7763e37936c0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79140fa4073bc441db75/workflow.json b/records/sd1/sweep/79140fa4073bc441db75/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3feac6db737f947964baca3674b1c8b49b5c9156
--- /dev/null
+++ b/records/sd1/sweep/79140fa4073bc441db75/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79140fa4073bc441db75)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.norm\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7917a11b9eb89fb469d1/measurements.json b/records/sd1/sweep/7917a11b9eb89fb469d1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce91a1b262f0ebe203564ecbf349ac29baad51f2
--- /dev/null
+++ b/records/sd1/sweep/7917a11b9eb89fb469d1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7917a11b9eb89fb469d1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.369534,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.359835,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.127774,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.037598,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.176,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00219,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.787,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9902,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.083,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7917a11b9eb89fb469d1/record.json b/records/sd1/sweep/7917a11b9eb89fb469d1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..28d4180c0e5d2bd8ae13ad24aff869000b2cd44b
--- /dev/null
+++ b/records/sd1/sweep/7917a11b9eb89fb469d1/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7917a11b9eb89fb469d1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d5310637f6a8b42797d64ff70584e729b50771028539875a096622eae8bfda37.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d5310637f6a8b42797d64ff70584e729b50771028539875a096622eae8bfda37",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "08547de14c259ed557a6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7917a11b9eb89fb469d1/workflow.json b/records/sd1/sweep/7917a11b9eb89fb469d1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e87a0cd719e447440f56f749b7d8ad28ea438717
--- /dev/null
+++ b/records/sd1/sweep/7917a11b9eb89fb469d1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7917a11b9eb89fb469d1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/791b3e746a86cf821831/measurements.json b/records/sd1/sweep/791b3e746a86cf821831/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a13d0b5657110aefd9bbafdfbba3fc694ef05d1
--- /dev/null
+++ b/records/sd1/sweep/791b3e746a86cf821831/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "791b3e746a86cf821831",
+ "values": {
+  "cosine_distance": {
+   "value": 0.17176,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.184586,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.07671,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.078875,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.956,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00465,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.056,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9476,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4654,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/791b3e746a86cf821831/record.json b/records/sd1/sweep/791b3e746a86cf821831/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a2300b02d4119d13c0e61de6b0aa25db5d8fc58c
--- /dev/null
+++ b/records/sd1/sweep/791b3e746a86cf821831/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "791b3e746a86cf821831",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8c96c58745caaee52ea0839583af05e48fa7abee9214da3901c1160adc61b1aa.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8c96c58745caaee52ea0839583af05e48fa7abee9214da3901c1160adc61b1aa",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "d79230d823d3f0c2b270"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/791b3e746a86cf821831/workflow.json b/records/sd1/sweep/791b3e746a86cf821831/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ed652a2f4c1a9591da523b9451b74807a946f53
--- /dev/null
+++ b/records/sd1/sweep/791b3e746a86cf821831/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 791b3e746a86cf821831)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.norm3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/791c6079fcb221282249/measurements.json b/records/sd1/sweep/791c6079fcb221282249/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a17bcc51901581c0f56bb729e6de86a14f543d6
--- /dev/null
+++ b/records/sd1/sweep/791c6079fcb221282249/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "791c6079fcb221282249",
+ "values": {
+  "cosine_distance": {
+   "value": 0.312868,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.34667,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.235282,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.193256,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.763,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00423,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 34.273,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9402,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1943,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/791c6079fcb221282249/record.json b/records/sd1/sweep/791c6079fcb221282249/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0883af0506b3007e96750050abc6b583f426f279
--- /dev/null
+++ b/records/sd1/sweep/791c6079fcb221282249/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "791c6079fcb221282249",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/1641e69323523f425ba618385a7095075fdb1a6c7ac31de896a3ac1569191af7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1641e69323523f425ba618385a7095075fdb1a6c7ac31de896a3ac1569191af7",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "453fc7df952d729a77d4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/791c6079fcb221282249/workflow.json b/records/sd1/sweep/791c6079fcb221282249/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b91666dc2680a1fbfd84f63f69ec3800801998b
--- /dev/null
+++ b/records/sd1/sweep/791c6079fcb221282249/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 791c6079fcb221282249)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.proj_out\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79212ca28c8035d5e037/measurements.json b/records/sd1/sweep/79212ca28c8035d5e037/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..695d76128a2fbf62eff69467d9a4edea261054cb
--- /dev/null
+++ b/records/sd1/sweep/79212ca28c8035d5e037/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79212ca28c8035d5e037",
+ "values": {
+  "cosine_distance": {
+   "value": 0.64394,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.721191,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.894718,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.484864,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.397,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00835,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 64.506,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8678,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9523,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79212ca28c8035d5e037/record.json b/records/sd1/sweep/79212ca28c8035d5e037/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f2a3999296ac9232fd4d736c8daa5918d6317a5
--- /dev/null
+++ b/records/sd1/sweep/79212ca28c8035d5e037/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79212ca28c8035d5e037",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/b83583433e7dcfababfea3a6c1e01a0fd73538fcc9870fe6ec4045f66a493598.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b83583433e7dcfababfea3a6c1e01a0fd73538fcc9870fe6ec4045f66a493598",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "9bc3b5818510ee521ce6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79212ca28c8035d5e037/workflow.json b/records/sd1/sweep/79212ca28c8035d5e037/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..282ee713c8ef19a72d3d7935a61ae014f52dc562
--- /dev/null
+++ b/records/sd1/sweep/79212ca28c8035d5e037/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79212ca28c8035d5e037)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.skip_connection\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792253136179707dc0f9/measurements.json b/records/sd1/sweep/792253136179707dc0f9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..82c59b42d00ad98526ece2ff5f7cc947d4eebfbe
--- /dev/null
+++ b/records/sd1/sweep/792253136179707dc0f9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "792253136179707dc0f9",
+ "values": {
+  "std": {
+   "value": 66.256,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00672,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "clip_degenerate"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 63.212,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.505692,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.693978,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.962071,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.319333,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.662,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9766,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792253136179707dc0f9/record.json b/records/sd1/sweep/792253136179707dc0f9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1890f1dff4b549995750159153e66b5f320c967
--- /dev/null
+++ b/records/sd1/sweep/792253136179707dc0f9/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "792253136179707dc0f9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:10:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "quarter",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   90.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [0.0, 5.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7002dcafbf80e62d8d96"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792253136179707dc0f9/workflow.json b/records/sd1/sweep/792253136179707dc0f9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c99879fd05e6ad9aac519832469d602a037bf4c
--- /dev/null
+++ b/records/sd1/sweep/792253136179707dc0f9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 792253136179707dc0f9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792ee44d7466102d08ae/interpretations.jsonl b/records/sd1/sweep/792ee44d7466102d08ae/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..07ac772aa2a70758ea2bd1ac6acdd3f19c550d09
--- /dev/null
+++ b/records/sd1/sweep/792ee44d7466102d08ae/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "792ee44d7466102d08ae"}, "kind": "caption", "value": "Short-haired young woman in black shirt, three-quarter portrait against grey-green vintage photo backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b84e5cd3135c15bb"}
+{"target": {"record": "792ee44d7466102d08ae"}, "kind": "change", "value": "Almost no visible change; the lips look a touch redder and facial shading is marginally deeper.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "65ee31283065919d"}
+{"target": {"record": "792ee44d7466102d08ae"}, "kind": "keywords", "value": ["portrait", "near identical", "vintage photo", "subtle lip colour"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7e4970fd5ca3b483"}
+{"target": {"record": "792ee44d7466102d08ae"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5f5b9f4551d983df"}
diff --git a/records/sd1/sweep/792ee44d7466102d08ae/measurements.json b/records/sd1/sweep/792ee44d7466102d08ae/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c131477262aa2d65e21f067318c59c390f4176d0
--- /dev/null
+++ b/records/sd1/sweep/792ee44d7466102d08ae/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "792ee44d7466102d08ae",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792ee44d7466102d08ae/record.json b/records/sd1/sweep/792ee44d7466102d08ae/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5696a0f4da24457a5af4f78c467eef371c878ff4
--- /dev/null
+++ b/records/sd1/sweep/792ee44d7466102d08ae/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "792ee44d7466102d08ae",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b8d762fa75acfa2e0a8122a4cf5fb8b1999f698c48e35c55cf14169c5af9a8c6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b8d762fa75acfa2e0a8122a4cf5fb8b1999f698c48e35c55cf14169c5af9a8c6",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "0e6aacb8ce02171ec7df"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792ee44d7466102d08ae/workflow.json b/records/sd1/sweep/792ee44d7466102d08ae/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8bceecf5dc89d2bc1518152cefa00ef1f8c640ec
--- /dev/null
+++ b/records/sd1/sweep/792ee44d7466102d08ae/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 792ee44d7466102d08ae)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792f21bfea9c439b8117/interpretations.jsonl b/records/sd1/sweep/792f21bfea9c439b8117/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..9102fec4c65550e9c36cee934bac6d264d90dd9a
--- /dev/null
+++ b/records/sd1/sweep/792f21bfea9c439b8117/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "792f21bfea9c439b8117"}, "kind": "caption", "value": "Small vivid lighthouse on a green islet framed by a dark smoky brown border.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3ed892557ab962d1"}
+{"target": {"record": "792f21bfea9c439b8117"}, "kind": "change", "value": "The lighthouse scene shrinks into a saturated, painterly vignette in the centre while a dark smudged border frames it.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "64e771ff8407ef91"}
+{"target": {"record": "792f21bfea9c439b8117"}, "kind": "keywords", "value": ["smoky frame", "vignette lighthouse", "painterly island", "shrunken scene", "dark border"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a9ae5f183e309aec"}
+{"target": {"record": "792f21bfea9c439b8117"}, "kind": "effect_tags", "value": ["framing", "painterly", "saturated", "zoom", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cd249fa7ad5ec20a"}
diff --git a/records/sd1/sweep/792f21bfea9c439b8117/measurements.json b/records/sd1/sweep/792f21bfea9c439b8117/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d949425a4c707baaa56f7c4628800cf00a07e48
--- /dev/null
+++ b/records/sd1/sweep/792f21bfea9c439b8117/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "792f21bfea9c439b8117",
+ "values": {
+  "std": {
+   "value": 49.644,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00531,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.209,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.467551,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.672873,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.464252,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.258757,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.975,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0545,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792f21bfea9c439b8117/record.json b/records/sd1/sweep/792f21bfea9c439b8117/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..12e25a5f997d6e8c5e13452dd0471828eba51451
--- /dev/null
+++ b/records/sd1/sweep/792f21bfea9c439b8117/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "792f21bfea9c439b8117",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:01:41Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "shrink",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "scale",
+   0.7,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.7,
+    "applied": 0.7,
+    "reason": "outside safe range [0.95, 1.05] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/792f21bfea9c439b8117/workflow.json b/records/sd1/sweep/792f21bfea9c439b8117/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..74d1eb668610a0f4da1a0dc83d11332b3fe3254a
--- /dev/null
+++ b/records/sd1/sweep/792f21bfea9c439b8117/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 792f21bfea9c439b8117)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79327da7278c59e2af1a/measurements.json b/records/sd1/sweep/79327da7278c59e2af1a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a33b62b0d437fce7e35b903d42d4c6c405cbae1
--- /dev/null
+++ b/records/sd1/sweep/79327da7278c59e2af1a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79327da7278c59e2af1a",
+ "values": {
+  "std": {
+   "value": 43.228,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01669,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.925,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.021891,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.056316,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.023677,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010876,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0056,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0252,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79327da7278c59e2af1a/record.json b/records/sd1/sweep/79327da7278c59e2af1a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..75fd88dea2bdddc31e83377c5fb1adc53ba59c22
--- /dev/null
+++ b/records/sd1/sweep/79327da7278c59e2af1a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "79327da7278c59e2af1a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:58:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "mid",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_scalar",
+   0.8,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "46385f45fa5dde929490"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79327da7278c59e2af1a/workflow.json b/records/sd1/sweep/79327da7278c59e2af1a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2686095c8e6140e47e846e2d7fa70f785da7c75e
--- /dev/null
+++ b/records/sd1/sweep/79327da7278c59e2af1a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79327da7278c59e2af1a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7933510e61c1f029941d/measurements.json b/records/sd1/sweep/7933510e61c1f029941d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dec956fccde2e960b9c0ad13c34fea298aee781f
--- /dev/null
+++ b/records/sd1/sweep/7933510e61c1f029941d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7933510e61c1f029941d",
+ "values": {
+  "std": {
+   "value": 55.906,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00696,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.88,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.101393,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.098376,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038151,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.014376,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9994,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2941,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7933510e61c1f029941d/record.json b/records/sd1/sweep/7933510e61c1f029941d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..44530dfa7e3dacde149abcbc7c614e67cd381ca3
--- /dev/null
+++ b/records/sd1/sweep/7933510e61c1f029941d/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "7933510e61c1f029941d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:57:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_scalar",
+   -0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e233ee3cf6ddff5a7903"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7933510e61c1f029941d/workflow.json b/records/sd1/sweep/7933510e61c1f029941d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a276efdaaa1adbe5aac6be85e25f0c3d434a222
--- /dev/null
+++ b/records/sd1/sweep/7933510e61c1f029941d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7933510e61c1f029941d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7937b504b7c1c27b7309/interpretations.jsonl b/records/sd1/sweep/7937b504b7c1c27b7309/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2cbf9716310ddb18593740186af7524120f3a745
--- /dev/null
+++ b/records/sd1/sweep/7937b504b7c1c27b7309/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7937b504b7c1c27b7309"}, "kind": "caption", "value": "Three-quarter portrait of a young woman with a short dark bob on a grey backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b3ee5da4d289d348"}
+{"target": {"record": "7937b504b7c1c27b7309"}, "kind": "change", "value": "Almost no visible change, apart from slightly stronger contrast in the eyes and lips.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0e0a3b4458572d5c"}
+{"target": {"record": "7937b504b7c1c27b7309"}, "kind": "keywords", "value": ["near identical", "portrait", "dark bob", "minor change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7402b8635e999abd"}
+{"target": {"record": "7937b504b7c1c27b7309"}, "kind": "effect_tags", "value": ["subject_kept", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ca6aff1b6573fef4"}
diff --git a/records/sd1/sweep/7937b504b7c1c27b7309/measurements.json b/records/sd1/sweep/7937b504b7c1c27b7309/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ae26c804b6241fa968c700a363b81f4fe32743f
--- /dev/null
+++ b/records/sd1/sweep/7937b504b7c1c27b7309/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7937b504b7c1c27b7309",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": -0.0,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.477,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00637,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0093,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2957,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7937b504b7c1c27b7309/record.json b/records/sd1/sweep/7937b504b7c1c27b7309/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb80b6f5333922ba7a835fdb0f2163883d6bc7b7
--- /dev/null
+++ b/records/sd1/sweep/7937b504b7c1c27b7309/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7937b504b7c1c27b7309",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:27Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5f77567adecdb0c564cf07bc4d8eca21ebc460c81d41d7c90745f49f86a4e7d5.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5f77567adecdb0c564cf07bc4d8eca21ebc460c81d41d7c90745f49f86a4e7d5",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "18dac5c767715396619d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7937b504b7c1c27b7309/workflow.json b/records/sd1/sweep/7937b504b7c1c27b7309/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf6105d4d81e722d1e43b2874faf3c817fd8ee84
--- /dev/null
+++ b/records/sd1/sweep/7937b504b7c1c27b7309/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7937b504b7c1c27b7309)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/794366aeed40db7cf1be/measurements.json b/records/sd1/sweep/794366aeed40db7cf1be/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6313b2e7bf9e6f8b55a8703446433d4be8abee03
--- /dev/null
+++ b/records/sd1/sweep/794366aeed40db7cf1be/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "794366aeed40db7cf1be",
+ "values": {
+  "cosine_distance": {
+   "value": 0.415737,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.509475,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.356405,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.270015,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.963,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00401,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.948,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9317,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6627,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/794366aeed40db7cf1be/record.json b/records/sd1/sweep/794366aeed40db7cf1be/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..68b82057dab7c25712bf4acaf5339b9a33f2d9d3
--- /dev/null
+++ b/records/sd1/sweep/794366aeed40db7cf1be/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "794366aeed40db7cf1be",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:30Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/67e4ff7dcac037624d6a86ee4fb110057de091e4016020cfd21c39b8eacc98b9.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "67e4ff7dcac037624d6a86ee4fb110057de091e4016020cfd21c39b8eacc98b9",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "a0b75034fa6c2716ed1f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/794366aeed40db7cf1be/workflow.json b/records/sd1/sweep/794366aeed40db7cf1be/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2599e4cfea6c394bdb94e2b6bee1d9e7610d248c
--- /dev/null
+++ b/records/sd1/sweep/794366aeed40db7cf1be/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 794366aeed40db7cf1be)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.transformer_blocks.0.norm3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7945f28346e9bbb41306/measurements.json b/records/sd1/sweep/7945f28346e9bbb41306/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b85aaf1fc818baf405ff2548e71978d3a72cb73d
--- /dev/null
+++ b/records/sd1/sweep/7945f28346e9bbb41306/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7945f28346e9bbb41306",
+ "values": {
+  "cosine_distance": {
+   "value": 0.379426,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.655871,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.665984,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.361895,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.605,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00378,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.933,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0792,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2112,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7945f28346e9bbb41306/record.json b/records/sd1/sweep/7945f28346e9bbb41306/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f516146baaf4211c53199d36b0a227cd962db5cb
--- /dev/null
+++ b/records/sd1/sweep/7945f28346e9bbb41306/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7945f28346e9bbb41306",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/a1acfe49ee4800e08a5f31f05abb03a87e37cc7ae1ed65827ae3ac44d3eb2981.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a1acfe49ee4800e08a5f31f05abb03a87e37cc7ae1ed65827ae3ac44d3eb2981",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c5fafeabcb9f07910ffb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7945f28346e9bbb41306/workflow.json b/records/sd1/sweep/7945f28346e9bbb41306/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5983d36305c8f0725d670599778b23a02fbb30b4
--- /dev/null
+++ b/records/sd1/sweep/7945f28346e9bbb41306/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7945f28346e9bbb41306)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7949ba99ac37d2a22167/measurements.json b/records/sd1/sweep/7949ba99ac37d2a22167/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..afd92dc22f269dce98251184d09244f10ec183d6
--- /dev/null
+++ b/records/sd1/sweep/7949ba99ac37d2a22167/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7949ba99ac37d2a22167",
+ "values": {
+  "cosine_distance": {
+   "value": 0.223455,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.233409,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.141978,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.132683,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.465,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00333,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.382,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9729,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3774,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7949ba99ac37d2a22167/record.json b/records/sd1/sweep/7949ba99ac37d2a22167/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f61c5c8cc93abbd590fbe19975f84e787b68abc
--- /dev/null
+++ b/records/sd1/sweep/7949ba99ac37d2a22167/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7949ba99ac37d2a22167",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:59Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/03068fbed148c217bcf122bdb0d1c61bf89cf1ec7aa3f3f4471d3b9087466a67.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "03068fbed148c217bcf122bdb0d1c61bf89cf1ec7aa3f3f4471d3b9087466a67",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "9b4b8f19647748b820a5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7949ba99ac37d2a22167/workflow.json b/records/sd1/sweep/7949ba99ac37d2a22167/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9596cfa8a390fa294874e24d99ce888de97e092e
--- /dev/null
+++ b/records/sd1/sweep/7949ba99ac37d2a22167/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7949ba99ac37d2a22167)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/795d7f0bc068386a5602/measurements.json b/records/sd1/sweep/795d7f0bc068386a5602/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7754e7919974402e13df387c9372b742e3de3901
--- /dev/null
+++ b/records/sd1/sweep/795d7f0bc068386a5602/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "795d7f0bc068386a5602",
+ "values": {
+  "cosine_distance": {
+   "value": 0.174649,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.285691,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.137015,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.082406,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.356,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00342,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.495,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0826,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.114,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/795d7f0bc068386a5602/record.json b/records/sd1/sweep/795d7f0bc068386a5602/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..df72272b2c8ca417317aea38785df732d61ecca3
--- /dev/null
+++ b/records/sd1/sweep/795d7f0bc068386a5602/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "795d7f0bc068386a5602",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.hi",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/646bf8cbd57d7ef6d77f4ee2a36512a0ec132754e685210d6d36ee2fe97aeee5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "646bf8cbd57d7ef6d77f4ee2a36512a0ec132754e685210d6d36ee2fe97aeee5",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "705f602f97b5f98a47df"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/795d7f0bc068386a5602/workflow.json b/records/sd1/sweep/795d7f0bc068386a5602/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e31ab269aacffc3ed9a4032bfac1e714edb13b4a
--- /dev/null
+++ b/records/sd1/sweep/795d7f0bc068386a5602/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 795d7f0bc068386a5602)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79678ed0022c048247eb/measurements.json b/records/sd1/sweep/79678ed0022c048247eb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2a05d33689f02135aaa39578ce8fa91e0cd0a2a
--- /dev/null
+++ b/records/sd1/sweep/79678ed0022c048247eb/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "79678ed0022c048247eb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.191264,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.40674,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.346288,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.165702,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.114,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00451,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.426,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0753,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1218,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79678ed0022c048247eb/record.json b/records/sd1/sweep/79678ed0022c048247eb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4470768fda6395e428a511b2caad789cc64efb9e
--- /dev/null
+++ b/records/sd1/sweep/79678ed0022c048247eb/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79678ed0022c048247eb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:17Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ac220cf2e20796d9e4dbcb8f43414b95306178ce2d8f444ed9b264799ed48b13.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ac220cf2e20796d9e4dbcb8f43414b95306178ce2d8f444ed9b264799ed48b13",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "043aca68ba6a0d162604"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79678ed0022c048247eb/workflow.json b/records/sd1/sweep/79678ed0022c048247eb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b4f34d8a2d21b3fdd9395b2cad7640cfa94631a9
--- /dev/null
+++ b/records/sd1/sweep/79678ed0022c048247eb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79678ed0022c048247eb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79696775b05b2d3fdfe0/interpretations.jsonl b/records/sd1/sweep/79696775b05b2d3fdfe0/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ebe921e89882ab9e14ea52b38151d3ce8d1b1e31
--- /dev/null
+++ b/records/sd1/sweep/79696775b05b2d3fdfe0/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "79696775b05b2d3fdfe0"}, "kind": "caption", "value": "Fine rainbow speckle noise covering the whole frame, no recognisable subject", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "91630cd9f4e57833"}
+{"target": {"record": "79696775b05b2d3fdfe0"}, "kind": "change", "value": "The portrait collapses entirely into fine saturated colour static.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "acedc96f67336129"}
+{"target": {"record": "79696775b05b2d3fdfe0"}, "kind": "keywords", "value": ["colour noise", "static", "broken image", "rainbow speckle", "collapse"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d114f9ffb87b90e8"}
+{"target": {"record": "79696775b05b2d3fdfe0"}, "kind": "effect_tags", "value": ["degenerate", "subject_lost", "abstract", "saturated", "glitch", "pattern"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fa3c71b319789d27"}
diff --git a/records/sd1/sweep/79696775b05b2d3fdfe0/measurements.json b/records/sd1/sweep/79696775b05b2d3fdfe0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6af39d6c7d9f0fe0dfe98485e951f4d80bde4a3
--- /dev/null
+++ b/records/sd1/sweep/79696775b05b2d3fdfe0/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "79696775b05b2d3fdfe0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.723996,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.834873,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.941464,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.552896,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 91.637,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02459,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "clipped"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "clip_degenerate"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 101.263,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.7761,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9857,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79696775b05b2d3fdfe0/record.json b/records/sd1/sweep/79696775b05b2d3fdfe0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..749a1cf6854010b9c6f97c420077f0e10033260d
--- /dev/null
+++ b/records/sd1/sweep/79696775b05b2d3fdfe0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79696775b05b2d3fdfe0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "out.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.conv",
+   "kind": "out_conv",
+   "module_type": "SiLU",
+   "container": "Sequential",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/82c65108c474205c98076a9999125737f9ccf8c493e55ca9b78d0f6e9269d2bc.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "82c65108c474205c98076a9999125737f9ccf8c493e55ca9b78d0f6e9269d2bc",
+  "type_path": "UNetModel.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "3da19df1da1f7926c4c7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79696775b05b2d3fdfe0/workflow.json b/records/sd1/sweep/79696775b05b2d3fdfe0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..556be7b7970ccf47a2fca3d81bec9baa3273276c
--- /dev/null
+++ b/records/sd1/sweep/79696775b05b2d3fdfe0/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79696775b05b2d3fdfe0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"out.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/796e09d9ac95f79bd847/measurements.json b/records/sd1/sweep/796e09d9ac95f79bd847/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..10c710dbcae1fe1350361fdf9646385bc8fb13cf
--- /dev/null
+++ b/records/sd1/sweep/796e09d9ac95f79bd847/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "796e09d9ac95f79bd847",
+ "values": {
+  "cosine_distance": {
+   "value": 0.25856,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.315417,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.150039,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.15225,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.472,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00541,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.363,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9799,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2396,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/796e09d9ac95f79bd847/record.json b/records/sd1/sweep/796e09d9ac95f79bd847/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0843559e10caf222ba0d8773817f123ac399c661
--- /dev/null
+++ b/records/sd1/sweep/796e09d9ac95f79bd847/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "796e09d9ac95f79bd847",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8c8e305bff8f44937642c8c3247db1f8af28753278208b66dab94ec31ca2508d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8c8e305bff8f44937642c8c3247db1f8af28753278208b66dab94ec31ca2508d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "adecda8d3c140104bf98"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/796e09d9ac95f79bd847/workflow.json b/records/sd1/sweep/796e09d9ac95f79bd847/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a07840d54cb104240e440c85ba1d6e990a7d22f
--- /dev/null
+++ b/records/sd1/sweep/796e09d9ac95f79bd847/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 796e09d9ac95f79bd847)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/797b0798e07ff53fded6/measurements.json b/records/sd1/sweep/797b0798e07ff53fded6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08feb3de110a65ae9c214f11f61db7d05d8b37f0
--- /dev/null
+++ b/records/sd1/sweep/797b0798e07ff53fded6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "797b0798e07ff53fded6",
+ "values": {
+  "std": {
+   "value": 65.861,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00439,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 58.101,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.503872,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.669821,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.96701,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.328291,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.683,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8441,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/797b0798e07ff53fded6/record.json b/records/sd1/sweep/797b0798e07ff53fded6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b87df3c481c3e3a0226cebda90ff3c390bdca2a
--- /dev/null
+++ b/records/sd1/sweep/797b0798e07ff53fded6/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "797b0798e07ff53fded6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:31:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "slight",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   30.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 30.0,
+    "applied": 30.0,
+    "reason": "outside safe range [0.0, 5.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "9d41fcd8578ee9d69ca6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/797b0798e07ff53fded6/workflow.json b/records/sd1/sweep/797b0798e07ff53fded6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..98ce586ba2ad17e207ee5d9127c08d0dece0e4a6
--- /dev/null
+++ b/records/sd1/sweep/797b0798e07ff53fded6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 797b0798e07ff53fded6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/798540e931882f350abd/measurements.json b/records/sd1/sweep/798540e931882f350abd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2466478da3459016739cb0db090a2dd67520fb9
--- /dev/null
+++ b/records/sd1/sweep/798540e931882f350abd/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "798540e931882f350abd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.109569,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.110481,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.036115,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.092847,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.369,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00404,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.935,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9629,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2905,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/798540e931882f350abd/record.json b/records/sd1/sweep/798540e931882f350abd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..40483d0227a97486f052b15abb8e1b7d6ca1f86f
--- /dev/null
+++ b/records/sd1/sweep/798540e931882f350abd/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "798540e931882f350abd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ec4412f8eea98a2d6fd1333b2b3cea35892e99f0a308a8d9650e504ebc156821.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ec4412f8eea98a2d6fd1333b2b3cea35892e99f0a308a8d9650e504ebc156821",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "e1e1b9b70f74b6c1ce8f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/798540e931882f350abd/workflow.json b/records/sd1/sweep/798540e931882f350abd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..17a3dae310776b50936c35dd7a9a9454815f0c0c
--- /dev/null
+++ b/records/sd1/sweep/798540e931882f350abd/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 798540e931882f350abd)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79890e0ff0f43345147d/measurements.json b/records/sd1/sweep/79890e0ff0f43345147d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..39f4ef479d8da47c40bf415035244eca2fd8c908
--- /dev/null
+++ b/records/sd1/sweep/79890e0ff0f43345147d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79890e0ff0f43345147d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.320457,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.418521,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.253993,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.137279,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 35.315,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02575,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.782,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9852,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0099,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79890e0ff0f43345147d/record.json b/records/sd1/sweep/79890e0ff0f43345147d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ee0b69987b17a18cdb844b06981ef6a5a2536a8
--- /dev/null
+++ b/records/sd1/sweep/79890e0ff0f43345147d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79890e0ff0f43345147d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/96a8b6b412313a6b0b2ddb110eb8fd56078b14123d80788ad3c097199b0c1e9c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "96a8b6b412313a6b0b2ddb110eb8fd56078b14123d80788ad3c097199b0c1e9c",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "843e6a2e04d1ae859094"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79890e0ff0f43345147d/workflow.json b/records/sd1/sweep/79890e0ff0f43345147d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f2bda6606b41d8807785f41c5b24ccb475b0cb5
--- /dev/null
+++ b/records/sd1/sweep/79890e0ff0f43345147d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79890e0ff0f43345147d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.norm3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7995ad792d257ad0b4bc/measurements.json b/records/sd1/sweep/7995ad792d257ad0b4bc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9483e366ab96cbedb83dc152d02f300f00578079
--- /dev/null
+++ b/records/sd1/sweep/7995ad792d257ad0b4bc/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7995ad792d257ad0b4bc",
+ "values": {
+  "cosine_distance": {
+   "value": 0.449936,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.716123,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.51714,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.343411,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.887,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0019,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 47.871,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.907,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7131,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7995ad792d257ad0b4bc/record.json b/records/sd1/sweep/7995ad792d257ad0b4bc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..077fe79bdf9da18186c487132d6fca0c07699b04
--- /dev/null
+++ b/records/sd1/sweep/7995ad792d257ad0b4bc/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7995ad792d257ad0b4bc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.out_layers.3",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/db9c2c11c9886ae24dcd6e4d7465e149a4791b05a99c3d1fb1b69352b23c1fc1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "db9c2c11c9886ae24dcd6e4d7465e149a4791b05a99c3d1fb1b69352b23c1fc1",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "4cc05849a2f6234a4752"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7995ad792d257ad0b4bc/workflow.json b/records/sd1/sweep/7995ad792d257ad0b4bc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8fb929eef22b4d268f1004fd233fde7cb65ea6e
--- /dev/null
+++ b/records/sd1/sweep/7995ad792d257ad0b4bc/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7995ad792d257ad0b4bc)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.out_layers.3\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79b7fe7cccacf74eae90/measurements.json b/records/sd1/sweep/79b7fe7cccacf74eae90/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ebb3bbba3523fdd440a318a6c7cb282e75e7e55
--- /dev/null
+++ b/records/sd1/sweep/79b7fe7cccacf74eae90/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79b7fe7cccacf74eae90",
+ "values": {
+  "cosine_distance": {
+   "value": 0.386848,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.419747,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.317479,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.116313,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.269,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00676,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.818,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0527,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2111,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79b7fe7cccacf74eae90/record.json b/records/sd1/sweep/79b7fe7cccacf74eae90/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..76482a5deaa6c051dbf286f4d12a91b8c5ba9690
--- /dev/null
+++ b/records/sd1/sweep/79b7fe7cccacf74eae90/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79b7fe7cccacf74eae90",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.3",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/870d2193bd562f012541888fb0a2e59bbb7caa31c1142906e0a104b49bfd652d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "870d2193bd562f012541888fb0a2e59bbb7caa31c1142906e0a104b49bfd652d",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "d48f96989f133c88ffa6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79b7fe7cccacf74eae90/workflow.json b/records/sd1/sweep/79b7fe7cccacf74eae90/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdc6a142caa5982c5f55f24521f7240e5c1dfe06
--- /dev/null
+++ b/records/sd1/sweep/79b7fe7cccacf74eae90/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79b7fe7cccacf74eae90)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.out_layers.3\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79b81e6f8275ffba469d/interpretations.jsonl b/records/sd1/sweep/79b81e6f8275ffba469d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e5f5b68de88887df8d48df3f24b142b5dfa6c02a
--- /dev/null
+++ b/records/sd1/sweep/79b81e6f8275ffba469d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "79b81e6f8275ffba469d"}, "kind": "caption", "value": "Painted woman with flatter wavy hair, blue eyes and black top on a grey panel", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d5f0241d88f6c404"}
+{"target": {"record": "79b81e6f8275ffba469d"}, "kind": "change", "value": "Almost no visible change; the hair lies a little flatter and the eyes turn a clearer blue.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "016a4592d168901c"}
+{"target": {"record": "79b81e6f8275ffba469d"}, "kind": "keywords", "value": ["portrait", "minor variation", "blue eyes", "painted portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "42bf9cccebd5884c"}
+{"target": {"record": "79b81e6f8275ffba469d"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a7150eee73991670"}
diff --git a/records/sd1/sweep/79b81e6f8275ffba469d/measurements.json b/records/sd1/sweep/79b81e6f8275ffba469d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f4b905f975c53b54b4f3c9da7e7f21921859457
--- /dev/null
+++ b/records/sd1/sweep/79b81e6f8275ffba469d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "79b81e6f8275ffba469d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.050138,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.070725,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.028707,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.012508,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.364,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00572,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.354,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9916,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1479,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79b81e6f8275ffba469d/record.json b/records/sd1/sweep/79b81e6f8275ffba469d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..349983b0e9f5a0224f95aadccaa8fc1e2d3566ac
--- /dev/null
+++ b/records/sd1/sweep/79b81e6f8275ffba469d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79b81e6f8275ffba469d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1a4ee3c14c1bd0a2ebccb9a7723bfef0e6fa1c8e84f10f96ac7af57aa9376cdf.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1a4ee3c14c1bd0a2ebccb9a7723bfef0e6fa1c8e84f10f96ac7af57aa9376cdf",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7a532cfccf2f60938c45"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79b81e6f8275ffba469d/workflow.json b/records/sd1/sweep/79b81e6f8275ffba469d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b76cd34a91c52f0ec6afef43bbd1889365c0a067
--- /dev/null
+++ b/records/sd1/sweep/79b81e6f8275ffba469d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79b81e6f8275ffba469d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bb5e9467c193a6ff2b/measurements.json b/records/sd1/sweep/79bb5e9467c193a6ff2b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..143faefe86867fd4dc3a9bd047cdb5e04eb341c3
--- /dev/null
+++ b/records/sd1/sweep/79bb5e9467c193a6ff2b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "79bb5e9467c193a6ff2b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.060292,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.131769,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.062105,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.04761,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.278,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00401,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.832,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0153,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0669,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bb5e9467c193a6ff2b/record.json b/records/sd1/sweep/79bb5e9467c193a6ff2b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c191f85d8f050677c89256228078ab97490ab377
--- /dev/null
+++ b/records/sd1/sweep/79bb5e9467c193a6ff2b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79bb5e9467c193a6ff2b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:43Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.1.norm",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e5b338686abfff253565ac0e942017baa149fb3b922d9e30aeebd71b78055b6f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e5b338686abfff253565ac0e942017baa149fb3b922d9e30aeebd71b78055b6f",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "28b643172a093ed79422"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bb5e9467c193a6ff2b/workflow.json b/records/sd1/sweep/79bb5e9467c193a6ff2b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a2c9202c80647b9088286399989474f4eccf5f4e
--- /dev/null
+++ b/records/sd1/sweep/79bb5e9467c193a6ff2b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79bb5e9467c193a6ff2b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.norm\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bca60931d854f8267e/measurements.json b/records/sd1/sweep/79bca60931d854f8267e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..21321a223feb8be48cbececa68380996d935de59
--- /dev/null
+++ b/records/sd1/sweep/79bca60931d854f8267e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "79bca60931d854f8267e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.342156,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.392947,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.306037,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.136629,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.669,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00533,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.86,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9597,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1132,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bca60931d854f8267e/record.json b/records/sd1/sweep/79bca60931d854f8267e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5661931120db4cb18cf496e046e468a7cc7e7858
--- /dev/null
+++ b/records/sd1/sweep/79bca60931d854f8267e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79bca60931d854f8267e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:27Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/af75a895a827b98f64a55b3f6ffcdf2cd4f9badc29aba0b6823b70b7e9ffe181.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "af75a895a827b98f64a55b3f6ffcdf2cd4f9badc29aba0b6823b70b7e9ffe181",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "0369872b70e12b76933f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bca60931d854f8267e/workflow.json b/records/sd1/sweep/79bca60931d854f8267e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e25713fc50ea62edf3b5d9173a07f059b9078f36
--- /dev/null
+++ b/records/sd1/sweep/79bca60931d854f8267e/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79bca60931d854f8267e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bddbc7bb3f67b0cfd5/measurements.json b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..21acbfdda7b43429471c75610182defe22fd3bfe
--- /dev/null
+++ b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79bddbc7bb3f67b0cfd5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.391843,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.490752,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.319012,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.191929,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.484,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00282,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.603,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0049,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2784,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bddbc7bb3f67b0cfd5/record.json b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f90d87d281226d3f8bb1824ee611a3609808672
--- /dev/null
+++ b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79bddbc7bb3f67b0cfd5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8c3e64fc54bcf37db338d9255f17dae41f17b7e5106f9f52441782b16db64cb9.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8c3e64fc54bcf37db338d9255f17dae41f17b7e5106f9f52441782b16db64cb9",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "2ef28f3c7a944a4317c0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79bddbc7bb3f67b0cfd5/workflow.json b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e337b75a2a2a659d12db50c85f08ea30e7dd7a05
--- /dev/null
+++ b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79bddbc7bb3f67b0cfd5)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79c42a428ad89b29f968/measurements.json b/records/sd1/sweep/79c42a428ad89b29f968/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e87e526b442c3f11433d35041d78761cac480115
--- /dev/null
+++ b/records/sd1/sweep/79c42a428ad89b29f968/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79c42a428ad89b29f968",
+ "values": {
+  "cosine_distance": {
+   "value": 0.512429,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.551177,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.71088,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.39095,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.121,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0067,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.147,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8835,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5798,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79c42a428ad89b29f968/record.json b/records/sd1/sweep/79c42a428ad89b29f968/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e19c0ccc88bd47bff6a361cdd39bad0d5f7ce41
--- /dev/null
+++ b/records/sd1/sweep/79c42a428ad89b29f968/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79c42a428ad89b29f968",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.in_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f2586eb7191749dfed326083e1563a572d7ffdb4dd5079bb1932bb7ce30b6d78.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f2586eb7191749dfed326083e1563a572d7ffdb4dd5079bb1932bb7ce30b6d78",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "13de5ac1175e0833f01a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79c42a428ad89b29f968/workflow.json b/records/sd1/sweep/79c42a428ad89b29f968/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..485050235facf6abd855b88d57e963b9f4dc42eb
--- /dev/null
+++ b/records/sd1/sweep/79c42a428ad89b29f968/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79c42a428ad89b29f968)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.in_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79e83513845e84b8d66e/measurements.json b/records/sd1/sweep/79e83513845e84b8d66e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..386896fab967a03516710a2ee9024bd383c8c5d6
--- /dev/null
+++ b/records/sd1/sweep/79e83513845e84b8d66e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79e83513845e84b8d66e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.264231,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.298933,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.053623,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.045794,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.772,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01068,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 36.419,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0006,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1055,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79e83513845e84b8d66e/record.json b/records/sd1/sweep/79e83513845e84b8d66e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45ab8ef9c51efe32c3b82c8064f5251e956aca3d
--- /dev/null
+++ b/records/sd1/sweep/79e83513845e84b8d66e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79e83513845e84b8d66e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/8bd9e0bb540595c16dcec3783163e6f78af47da17f15b2a889e22a6e9cb9569f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8bd9e0bb540595c16dcec3783163e6f78af47da17f15b2a889e22a6e9cb9569f",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.CrossAttention.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "b48921a683de1d4590fd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79e83513845e84b8d66e/workflow.json b/records/sd1/sweep/79e83513845e84b8d66e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7e15b14de2ca51ff7d5afcfab7721fd57b6c6fd
--- /dev/null
+++ b/records/sd1/sweep/79e83513845e84b8d66e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79e83513845e84b8d66e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.attn2.to_k\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79e988557b8a9032deff/measurements.json b/records/sd1/sweep/79e988557b8a9032deff/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..39e22b4f6fd18256f237f2f6bc73e690ab908cb4
--- /dev/null
+++ b/records/sd1/sweep/79e988557b8a9032deff/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79e988557b8a9032deff",
+ "values": {
+  "std": {
+   "value": 48.667,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01354,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.267,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.278155,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.284738,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.052894,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027405,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0086,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.327,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79e988557b8a9032deff/record.json b/records/sd1/sweep/79e988557b8a9032deff/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9cbd5e0c7fae2634e89b4893bf77f8685f023a9e
--- /dev/null
+++ b/records/sd1/sweep/79e988557b8a9032deff/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "79e988557b8a9032deff",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:24:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79e988557b8a9032deff/workflow.json b/records/sd1/sweep/79e988557b8a9032deff/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf0d15de6c6eae81a903e1537943ab68399c5d2e
--- /dev/null
+++ b/records/sd1/sweep/79e988557b8a9032deff/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79e988557b8a9032deff)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79efbb80d87066740d03/measurements.json b/records/sd1/sweep/79efbb80d87066740d03/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..224fa6ff3a20fdc322314d63d69e21346abe32e8
--- /dev/null
+++ b/records/sd1/sweep/79efbb80d87066740d03/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79efbb80d87066740d03",
+ "values": {
+  "cosine_distance": {
+   "value": 0.445496,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.590785,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.435084,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.276701,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 31.985,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00248,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 42.492,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9756,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2675,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79efbb80d87066740d03/record.json b/records/sd1/sweep/79efbb80d87066740d03/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca2f9ea55a3d23737b96a427e4581a7be401776f
--- /dev/null
+++ b/records/sd1/sweep/79efbb80d87066740d03/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "79efbb80d87066740d03",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/623daab7168fbe58a6c88569df50f4b4d54312d46c6dcdfb5992f73c6443d7fe.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "623daab7168fbe58a6c88569df50f4b4d54312d46c6dcdfb5992f73c6443d7fe",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "109f431fff9c1d1b0ba8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79efbb80d87066740d03/workflow.json b/records/sd1/sweep/79efbb80d87066740d03/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..544ea013d05a5247c4bfb0b9c0f1ab186843b0bb
--- /dev/null
+++ b/records/sd1/sweep/79efbb80d87066740d03/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79efbb80d87066740d03)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.in_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79f668289f23f61ade87/measurements.json b/records/sd1/sweep/79f668289f23f61ade87/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0c4c7c7ea67e23e95da7f1d36e8882f7561cf09
--- /dev/null
+++ b/records/sd1/sweep/79f668289f23f61ade87/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "79f668289f23f61ade87",
+ "values": {
+  "cosine_distance": {
+   "value": 0.209547,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.218793,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.111664,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.09789,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.4,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00582,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.149,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9957,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2196,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79f668289f23f61ade87/record.json b/records/sd1/sweep/79f668289f23f61ade87/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d1efc7b567ae89164570ac168de14a973ec2d9a
--- /dev/null
+++ b/records/sd1/sweep/79f668289f23f61ade87/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "79f668289f23f61ade87",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.10.0.in_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/80a7c0a030c422ad0c12b853b0212ebdfc6d0083e44a164c46fa86815f2ec043.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "80a7c0a030c422ad0c12b853b0212ebdfc6d0083e44a164c46fa86815f2ec043",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "cd5abdb4ffbbb9b7c931"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79f668289f23f61ade87/workflow.json b/records/sd1/sweep/79f668289f23f61ade87/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce7f009b32e5d1b3dfd907b11e1062f05540ce64
--- /dev/null
+++ b/records/sd1/sweep/79f668289f23f61ade87/workflow.json
@@ -0,0 +1,102 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "realisticVisionV51_v51VAE.safetensors"
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 79f668289f23f61ade87)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.in_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/79f99460c6f7d0efdc23/measurements.json b/records/sd1/sweep/79f99460c6f7d0efdc23/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0343db6dd116da2880c8cd5015a98da47c2b4925
--- /dev/null
+++ b/records/sd1/sweep/79f99460c6f7d0efdc23/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "79f99460c6f7d0efdc23",
+ "values": {
+  "std": {
+   "value": 58.826,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01004,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 8.485,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.032878,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.042419,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.007791,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.005786,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.977,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3865,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file