diff --git a/records/sd1/sweep/782e0d34306dd8c45f2a/output.webp b/records/sd1/sweep/782e0d34306dd8c45f2a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..95197caf4c7520192605d8f2e0255f825e822701
--- /dev/null
+++ b/records/sd1/sweep/782e0d34306dd8c45f2a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b0a7584c350d44f9a7cd44c9f3a88b5292dee56d65ad39b9668ba03f9a2e6fc
+size 33172
diff --git a/records/sd1/sweep/7835219479fa3eb99912/output.webp b/records/sd1/sweep/7835219479fa3eb99912/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7a77b6726318f01ea09ccc31fd9906989e1f9a5d
--- /dev/null
+++ b/records/sd1/sweep/7835219479fa3eb99912/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:730ac06d0693c54c24cb87aab340b9131c9f684eb6931c543e15bbab4276c37d
+size 52008
diff --git a/records/sd1/sweep/783f07e402f3192ce7e1/output.webp b/records/sd1/sweep/783f07e402f3192ce7e1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..74e9d022197a4ce9d490c876df385dbf82056a62
--- /dev/null
+++ b/records/sd1/sweep/783f07e402f3192ce7e1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54af1ac1b1839b81a15c954d18d3ebf79bab7e076d7e0b7c53010ccb7b99724b
+size 36588
diff --git a/records/sd1/sweep/78402d2816e50c0ec106/output.webp b/records/sd1/sweep/78402d2816e50c0ec106/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9fc19742d2512258eda95ac65f4d8bbe38c97b99
--- /dev/null
+++ b/records/sd1/sweep/78402d2816e50c0ec106/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe917a580d55fbb55c19aa86d77dddce80815c241e2a03d2225f437f514e1dda
+size 48086
diff --git a/records/sd1/sweep/784a55c6fdb74082f404/output.webp b/records/sd1/sweep/784a55c6fdb74082f404/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7ba5d423c015ecf3828b799e35243df50d3f0c3c
--- /dev/null
+++ b/records/sd1/sweep/784a55c6fdb74082f404/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3938767b295cd6e93acc6cf88723fa08b39d48c0e2e1161c90941c10fe9c3b9
+size 34540
diff --git a/records/sd1/sweep/785f1310d9709172abec/output.webp b/records/sd1/sweep/785f1310d9709172abec/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..69a3ec10d6014f9813515839094ae79bd04bb7d1
--- /dev/null
+++ b/records/sd1/sweep/785f1310d9709172abec/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4920d2e1f7a04d66f4723702cace51bc413192616243df7bb49766dae19ba997
+size 45458
diff --git a/records/sd1/sweep/786030205ecfe7b6b827/output.webp b/records/sd1/sweep/786030205ecfe7b6b827/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8ad2606a7cae28d91ae31c41a6580c92cc48afd9
--- /dev/null
+++ b/records/sd1/sweep/786030205ecfe7b6b827/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1de0eb70a029a297e982ea16159321d683ded4335cbc09c6946b29dd880bc05
+size 28880
diff --git a/records/sd1/sweep/7864e87287f500f7e560/output.webp b/records/sd1/sweep/7864e87287f500f7e560/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..eabff8b20553257d05b5a1effcddf88ed42f9a5a
--- /dev/null
+++ b/records/sd1/sweep/7864e87287f500f7e560/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7548fa8b5751339e19afaeab927a4b49a47264afd49365963f734bf3481d68db
+size 19932
diff --git a/records/sd1/sweep/786b72edfc5b8969003c/output.webp b/records/sd1/sweep/786b72edfc5b8969003c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..442b45e9517a078271ae15905da895797555d03e
--- /dev/null
+++ b/records/sd1/sweep/786b72edfc5b8969003c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c1c508a6bb7a85ae4571251af91bfc5353f51bda36aecec7734d80664629724
+size 22876
diff --git a/records/sd1/sweep/7870ef8ff992c2660db4/output.webp b/records/sd1/sweep/7870ef8ff992c2660db4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c4723682c8045ecde28e7b16050858e617f449f7
--- /dev/null
+++ b/records/sd1/sweep/7870ef8ff992c2660db4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bb3dd1bf33f063bcb0f8be5bd8c873ea3eb4fa084e61726a95e31b0255b5384
+size 21378
diff --git a/records/sd1/sweep/78801c1607303074208a/output.webp b/records/sd1/sweep/78801c1607303074208a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e4707c620e40d5a54eabac4d8ea5da821e532ec6
--- /dev/null
+++ b/records/sd1/sweep/78801c1607303074208a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6374cf0692412fb26652835bdcf016307a2df3537043f6a8f409f09447076bc3
+size 30046
diff --git a/records/sd1/sweep/78830710a66ae69b2737/output.webp b/records/sd1/sweep/78830710a66ae69b2737/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d1453f62eafc3d676e91fa6e0efadf38fc2cc0f5
--- /dev/null
+++ b/records/sd1/sweep/78830710a66ae69b2737/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:debdbf2fc34784a3e0554dd6c7bae48958ba3c25584ac8786979e2c35d4c7c2d
+size 36976
diff --git a/records/sd1/sweep/788482b0f55f3bb277c7/output.webp b/records/sd1/sweep/788482b0f55f3bb277c7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b061362218f3e7c133a73fc56d780dfe69c74548
--- /dev/null
+++ b/records/sd1/sweep/788482b0f55f3bb277c7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:704722dd37d09475098ffd0f96a304af7050f850fd51da34558741f55961fbbb
+size 47568
diff --git a/records/sd1/sweep/7887166e6beadf0e5eb4/output.webp b/records/sd1/sweep/7887166e6beadf0e5eb4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7224d0486d87adc9958e13bc006ff2e86ba5c015
--- /dev/null
+++ b/records/sd1/sweep/7887166e6beadf0e5eb4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87d82a71d9c0133a725f005d64a975f7a665279dbf19e0ca4933c94630169587
+size 36452
diff --git a/records/sd1/sweep/788770fac356ffd8faf3/output.webp b/records/sd1/sweep/788770fac356ffd8faf3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..82f36c2c710820fdebf6290ef27e167c252188a6
--- /dev/null
+++ b/records/sd1/sweep/788770fac356ffd8faf3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d02c2e644a3aa5bb36d41b29d2c2732ae082614c6f6b35be1755691142b1fbd6
+size 23814
diff --git a/records/sd1/sweep/7889d5419c692f224d0e/output.webp b/records/sd1/sweep/7889d5419c692f224d0e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fa3bf513ffc8858dcb8a8373c365384ba14c5f44
--- /dev/null
+++ b/records/sd1/sweep/7889d5419c692f224d0e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63a9c1e0cb0671a83c1a263e9441c204f7a6676754327ab8548167d72fa537ba
+size 32922
diff --git a/records/sd1/sweep/78960332676d22746b35/output.webp b/records/sd1/sweep/78960332676d22746b35/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ce88ebb133c62f9ec90de64aea37adb21d14f04c
--- /dev/null
+++ b/records/sd1/sweep/78960332676d22746b35/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6495ca6f09e5ea355bd6ab3db356a572d0da858c873e6d140f4ba0e51f42e52a
+size 49354
diff --git a/records/sd1/sweep/789a0735e21e271389b0/output.webp b/records/sd1/sweep/789a0735e21e271389b0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a2821e8b3debe6fe5616e67443359462cd9be9f7
--- /dev/null
+++ b/records/sd1/sweep/789a0735e21e271389b0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4774f07fcc369f71711fb20ff2366adbee0d98d69f2c6bf42b8a38b74dfb6557
+size 26914
diff --git a/records/sd1/sweep/789ce78491ce0e8e0191/output.webp b/records/sd1/sweep/789ce78491ce0e8e0191/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7d8b0c5a1035181b51f4146caa5110c93063c038
--- /dev/null
+++ b/records/sd1/sweep/789ce78491ce0e8e0191/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da387086384adf2cbb1783f4f8cbe32797d109cf59499ce8178d4ffa20fe7cc6
+size 35434
diff --git a/records/sd1/sweep/78a7d8d4424de4df48bc/output.webp b/records/sd1/sweep/78a7d8d4424de4df48bc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..784bfdacc12933448ed43679e9daacf33bbb3fe4
--- /dev/null
+++ b/records/sd1/sweep/78a7d8d4424de4df48bc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:971de642faf6af83280e33faaa999fc5b1940ebeae9ec8f1cd896cbe7776af40
+size 30258
diff --git a/records/sd1/sweep/78b49303d094456cabb8/output.webp b/records/sd1/sweep/78b49303d094456cabb8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..59db91fbef7ae1452d0115f08d26f96f25a81418
--- /dev/null
+++ b/records/sd1/sweep/78b49303d094456cabb8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8dccef04cf8dfeb2359d26ecfe8fba11f4ac2e5f6cc5214a1c521426863e7b8
+size 36502
diff --git a/records/sd1/sweep/78bcd46501b2131e6b5e/output.webp b/records/sd1/sweep/78bcd46501b2131e6b5e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..058750105935993e2bfa59b5be523e484468a42c
--- /dev/null
+++ b/records/sd1/sweep/78bcd46501b2131e6b5e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eea9a7a544d31d869e55cdc9af82b41bc218f3d1bc6bf68bb3694a9a3869fb90
+size 20572
diff --git a/records/sd1/sweep/78bfbad9dfc13a37465e/output.webp b/records/sd1/sweep/78bfbad9dfc13a37465e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2d695f9229d5e78cf53a185c46313f2a44a37bc6
--- /dev/null
+++ b/records/sd1/sweep/78bfbad9dfc13a37465e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76c7d1c86a9a7718e849360539f9166cda6238743fa84229730329ff97c520bc
+size 46040
diff --git a/records/sd1/sweep/78d074cab027ede997d0/output.webp b/records/sd1/sweep/78d074cab027ede997d0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f8d0e5c379c9b533fce583ee7cff92fbdfbf79ea
--- /dev/null
+++ b/records/sd1/sweep/78d074cab027ede997d0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db533e68df2553426413437c159ab353f6fbc22945d230f2a65ed091bdd884b9
+size 66812
diff --git a/records/sd1/sweep/78d56c2aa3c081c35986/output.webp b/records/sd1/sweep/78d56c2aa3c081c35986/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..79e6f055a7c098122bd2908e21272d0e1561e475
--- /dev/null
+++ b/records/sd1/sweep/78d56c2aa3c081c35986/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d81274c0d6adc5a661ed8a85d00efd08a303351774c63bc5aa8ad150fdce517d
+size 21032
diff --git a/records/sd1/sweep/78d91444b50ff67be290/output.webp b/records/sd1/sweep/78d91444b50ff67be290/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..441e4d018219fe49134f20e5612f277406373f4a
--- /dev/null
+++ b/records/sd1/sweep/78d91444b50ff67be290/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e34900093edada383d898d8e6fecdf094eeecebf5edce8e43b794727b04434a
+size 16016
diff --git a/records/sd1/sweep/78d9c13115cb957e2ea5/output.webp b/records/sd1/sweep/78d9c13115cb957e2ea5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3e10ed11eb690586457da7173732fcb09bf1c80a
--- /dev/null
+++ b/records/sd1/sweep/78d9c13115cb957e2ea5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bd20e5bf8db50f42a044d142c0a597ec6089bb92173b415a6520638405264d4
+size 40542
diff --git a/records/sd1/sweep/78dff70e5e2c1f5c4dac/output.webp b/records/sd1/sweep/78dff70e5e2c1f5c4dac/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0faf354227d5ceafdbec8018f674d8792a438b78
--- /dev/null
+++ b/records/sd1/sweep/78dff70e5e2c1f5c4dac/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b00f5c988933a828abfc447a3f275cbf6e9135627e20541faf495d376323fccb
+size 38570
diff --git a/records/sd1/sweep/78e2776e005642e03506/output.webp b/records/sd1/sweep/78e2776e005642e03506/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5e55bccf6e32be1262b8012facf273fafbd85adb
--- /dev/null
+++ b/records/sd1/sweep/78e2776e005642e03506/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da0ecdd889c382c534cc8be97da349ef65cc3ef098dcefccd9fe281a33158e50
+size 15726
diff --git a/records/sd1/sweep/78e8c951d4ca1c663294/output.webp b/records/sd1/sweep/78e8c951d4ca1c663294/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..10d688dad2467dbf571da033cb173e441971bc11
--- /dev/null
+++ b/records/sd1/sweep/78e8c951d4ca1c663294/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14e05c6a2030e651d865f3c449339c3ba5d866be190a8e790827a4e57d22afda
+size 27294
diff --git a/records/sd1/sweep/78f977af5496d8dbf7a1/output.webp b/records/sd1/sweep/78f977af5496d8dbf7a1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b8dd37e0894062ad40ef92e9d3a8ee2d833c36cd
--- /dev/null
+++ b/records/sd1/sweep/78f977af5496d8dbf7a1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9b92075dbbcdaf95e7f0f1f2670c11cb633aced114a5a86ba0e0b0c227e404b
+size 22002
diff --git a/records/sd1/sweep/78fe49d39543b538edc5/output.webp b/records/sd1/sweep/78fe49d39543b538edc5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f5dd57dd1f058896b8339eb3da6bf59d3e8a4deb
--- /dev/null
+++ b/records/sd1/sweep/78fe49d39543b538edc5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a98436d5fa15c02483c1013097194caaf6d92339d0fe2999a66e5f29f03dc66
+size 18354
diff --git a/records/sd1/sweep/7903370ff477d98c38c6/output.webp b/records/sd1/sweep/7903370ff477d98c38c6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2693ce40924257d29acf58a27b79c157d2f1134d
--- /dev/null
+++ b/records/sd1/sweep/7903370ff477d98c38c6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1158a417bdc00f466c386b27d7201116fcd8bc53f9e55e55a71693263ea6cb9a
+size 32974
diff --git a/records/sd1/sweep/7905f6af300dae548f55/output.webp b/records/sd1/sweep/7905f6af300dae548f55/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b3b862ed3a94d6516b79a7f8e4bd84112cfcfc40
--- /dev/null
+++ b/records/sd1/sweep/7905f6af300dae548f55/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ad0d98dc34d8b23160aad1d892a6ce320bcdb4317d886e7242f80293706ff907
+size 32616
diff --git a/records/sd1/sweep/790ff2e8b792c1d96715/output.webp b/records/sd1/sweep/790ff2e8b792c1d96715/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c24b45a6d82e95394b09f364c56f532bf06a12dc
--- /dev/null
+++ b/records/sd1/sweep/790ff2e8b792c1d96715/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1774b89c59e8ac6e94757c474bf87320476896165c0deb710e7f487a8f7cd263
+size 26808
diff --git a/records/sd1/sweep/79121598e1083a5625bf/output.webp b/records/sd1/sweep/79121598e1083a5625bf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0ae84d25b11aa221904efb827455ec9bb912d57b
--- /dev/null
+++ b/records/sd1/sweep/79121598e1083a5625bf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55bcd3a9c7e89ea825f3918f0d58cdf21f6cd5852b3320b84f5218491ec6db45
+size 31468
diff --git a/records/sd1/sweep/79123f11e03b81473def/output.webp b/records/sd1/sweep/79123f11e03b81473def/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4aaf0c1c266f15c965b6dc90c92bc870331d5107
--- /dev/null
+++ b/records/sd1/sweep/79123f11e03b81473def/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8eb7ef265caef8ca6f2565cdbdbac2d0a6deedd42311e2b5ff6affe2001617d
+size 31934
diff --git a/records/sd1/sweep/79140fa4073bc441db75/output.webp b/records/sd1/sweep/79140fa4073bc441db75/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9930b2261d99e915dca534151da3cd3afc23ff1e
--- /dev/null
+++ b/records/sd1/sweep/79140fa4073bc441db75/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1e38175611d6ea410c7e434e8b87cfb85d9cf2ef46d607bcf9fa4cc5285fc69
+size 82838
diff --git a/records/sd1/sweep/7917a11b9eb89fb469d1/output.webp b/records/sd1/sweep/7917a11b9eb89fb469d1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b8f6971fc6b33ca6d3a0cffb0144a3dad555b22d
--- /dev/null
+++ b/records/sd1/sweep/7917a11b9eb89fb469d1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f0f7489e827bb28d83c7390246914747d361d7d35a29b79d46eac09baef92e62
+size 30266
diff --git a/records/sd1/sweep/791b3e746a86cf821831/output.webp b/records/sd1/sweep/791b3e746a86cf821831/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..73052bb6f778e6d4f9693e832ac29b28b9738a33
--- /dev/null
+++ b/records/sd1/sweep/791b3e746a86cf821831/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0229abc508fb26d937bf9af0b42657178e51889be3da44884a23a369cc1d8d12
+size 15754
diff --git a/records/sd1/sweep/791c6079fcb221282249/output.webp b/records/sd1/sweep/791c6079fcb221282249/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6e8886a01cd6cfba36ead8dbe388cf79191627f1
--- /dev/null
+++ b/records/sd1/sweep/791c6079fcb221282249/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32aa51755ccb2942e5052d69cd11f4962d7d6da351a6c72168c5b8c6f12a0778
+size 26968
diff --git a/records/sd1/sweep/79212ca28c8035d5e037/output.webp b/records/sd1/sweep/79212ca28c8035d5e037/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..557b198d6e1a2bcc58bdd59171a6c9450e19cbc1
--- /dev/null
+++ b/records/sd1/sweep/79212ca28c8035d5e037/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b84e077a273162a629f3995873dad1f7576eacb2792020913bcd822b433e1816
+size 51020
diff --git a/records/sd1/sweep/792253136179707dc0f9/output.webp b/records/sd1/sweep/792253136179707dc0f9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3617bd4b3e4ef012473d434d24a9d12f40de4437
--- /dev/null
+++ b/records/sd1/sweep/792253136179707dc0f9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b4c2027cfbea4b5dae38d3dc0117a3a2919c41809c71c0ba1702ac21d415c06
+size 45864
diff --git a/records/sd1/sweep/792ee44d7466102d08ae/output.webp b/records/sd1/sweep/792ee44d7466102d08ae/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6e0d299548f1a5a0f23ff42d5411bb615399c47f
--- /dev/null
+++ b/records/sd1/sweep/792ee44d7466102d08ae/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e1b167d63b60a09b0fd58d82f4517467569f8c4b17c4fc6d1309070913d1f93
+size 19936
diff --git a/records/sd1/sweep/792f21bfea9c439b8117/output.webp b/records/sd1/sweep/792f21bfea9c439b8117/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9397b8dc04957d912b4f0fdf8e4b9cc2068c83a0
--- /dev/null
+++ b/records/sd1/sweep/792f21bfea9c439b8117/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40a9a7c36411199feb45ed3b39dca7df4073d5bb1d2b0c3d24ff07953b8a1383
+size 35122
diff --git a/records/sd1/sweep/79327da7278c59e2af1a/output.webp b/records/sd1/sweep/79327da7278c59e2af1a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ae71d4ebec1f1cf192ffe126b2ac5b9effea8460
--- /dev/null
+++ b/records/sd1/sweep/79327da7278c59e2af1a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:611b611e1e58d116bbe78dd88559e48bf82e25603ef58e7cff5b82612263a3ae
+size 48756
diff --git a/records/sd1/sweep/7933510e61c1f029941d/output.webp b/records/sd1/sweep/7933510e61c1f029941d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..862b0662a3577df9f6ba37970ad91f3f0fcaa72c
--- /dev/null
+++ b/records/sd1/sweep/7933510e61c1f029941d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cac98532cd897d97d9a514d9a492241c7ddc753a1a153e70132a24fceab9ba6
+size 37244
diff --git a/records/sd1/sweep/7937b504b7c1c27b7309/output.webp b/records/sd1/sweep/7937b504b7c1c27b7309/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b0f12b6e206ebf787c0190174c2e20ed54549856
--- /dev/null
+++ b/records/sd1/sweep/7937b504b7c1c27b7309/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92e2b84eccf92e5fe557c1b989581f991d970adf59de4f3114b00bd8e9306796
+size 19934
diff --git a/records/sd1/sweep/794366aeed40db7cf1be/output.webp b/records/sd1/sweep/794366aeed40db7cf1be/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..27498fc513b2d7cb24e43eb354b564a72f570444
--- /dev/null
+++ b/records/sd1/sweep/794366aeed40db7cf1be/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d761ce07530f8b18e195cc08c5fa661fa4397a438b7ce56bf4c8e02802831d0
+size 39506
diff --git a/records/sd1/sweep/7945f28346e9bbb41306/output.webp b/records/sd1/sweep/7945f28346e9bbb41306/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5e04e96a1d25391a519c447c0cb80e36ba0f6a7e
--- /dev/null
+++ b/records/sd1/sweep/7945f28346e9bbb41306/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:859ff3092e771d54ebaf1337dce98a3434cd93672b5f2873c7fe3d3035ac88e6
+size 14506
diff --git a/records/sd1/sweep/7949ba99ac37d2a22167/output.webp b/records/sd1/sweep/7949ba99ac37d2a22167/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..eff1100ae086a1161ef6b55e244a0cac65d1abba
--- /dev/null
+++ b/records/sd1/sweep/7949ba99ac37d2a22167/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4de1d9f2a37eea53b052291586c3089d5c2e25892e46b17216904c7d28b6065b
+size 15312
diff --git a/records/sd1/sweep/795d7f0bc068386a5602/output.webp b/records/sd1/sweep/795d7f0bc068386a5602/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..641f373c0f73962715beb34afb48485f892a98a9
--- /dev/null
+++ b/records/sd1/sweep/795d7f0bc068386a5602/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75ddaf3fb82ea620326a3c976980b575eff23b5e0afe54922fe1db0ddbabf094
+size 17968
diff --git a/records/sd1/sweep/79678ed0022c048247eb/output.webp b/records/sd1/sweep/79678ed0022c048247eb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ca047c67f9b31b2916f37e404bfb681009e0a7f0
--- /dev/null
+++ b/records/sd1/sweep/79678ed0022c048247eb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90ad61856bf2bc66abe3abc636f72846a655531d683c381b358f1cf3e3b1d9bf
+size 51276
diff --git a/records/sd1/sweep/79696775b05b2d3fdfe0/output.webp b/records/sd1/sweep/79696775b05b2d3fdfe0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a2d22a58831c364444cfec5bbd964b8632f81535
--- /dev/null
+++ b/records/sd1/sweep/79696775b05b2d3fdfe0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2b0d9ee28ff9b28d3f0bfd7318b97b6aff7164f113c5276d1e59cbc6539dfd8
+size 99726
diff --git a/records/sd1/sweep/796e09d9ac95f79bd847/output.webp b/records/sd1/sweep/796e09d9ac95f79bd847/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a353657b9b70364286582fb27b7b903a111ef0ac
--- /dev/null
+++ b/records/sd1/sweep/796e09d9ac95f79bd847/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:562e767034e8531934b2e58382b63e9fb1861f407b49cc9a35c9532115ebce41
+size 16030
diff --git a/records/sd1/sweep/797b0798e07ff53fded6/output.webp b/records/sd1/sweep/797b0798e07ff53fded6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f729953e6925de32653c16bfc2acc4e876ca5be4
--- /dev/null
+++ b/records/sd1/sweep/797b0798e07ff53fded6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12e83547a61749054c6528c50ff77635e8b9fc73fe3d8911229ed18399055bb7
+size 33802
diff --git a/records/sd1/sweep/798540e931882f350abd/output.webp b/records/sd1/sweep/798540e931882f350abd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a65b761c861d8d219a7892478c466b5458af2b7c
--- /dev/null
+++ b/records/sd1/sweep/798540e931882f350abd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:282f31e2e1f59f3116d7529802a25553073f2181bfbe1cac3b418d03fa0dc7db
+size 17624
diff --git a/records/sd1/sweep/79890e0ff0f43345147d/output.webp b/records/sd1/sweep/79890e0ff0f43345147d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6cbf5c195e1585d195c4bd21c2e4eeeae51903d5
--- /dev/null
+++ b/records/sd1/sweep/79890e0ff0f43345147d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3633e16f9e7ff11d2dde51f992c8937c24547134bca4579fce10711a06b6d256
+size 41792
diff --git a/records/sd1/sweep/7995ad792d257ad0b4bc/output.webp b/records/sd1/sweep/7995ad792d257ad0b4bc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a1d87295ba63a99754bb9d5f0e3433cf6cd108fd
--- /dev/null
+++ b/records/sd1/sweep/7995ad792d257ad0b4bc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:daee1c998012286520380cd3117a0cf8a6a3b14260db575d91232c3c978764a9
+size 11762
diff --git a/records/sd1/sweep/79b7fe7cccacf74eae90/output.webp b/records/sd1/sweep/79b7fe7cccacf74eae90/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..41200bf0f98106fb3044032a077e955a1689fc10
--- /dev/null
+++ b/records/sd1/sweep/79b7fe7cccacf74eae90/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a75161a06c6c999b169f2e9f0f0555822a20fe48f10e5aa43548dc9a0497cad2
+size 18034
diff --git a/records/sd1/sweep/79b81e6f8275ffba469d/output.webp b/records/sd1/sweep/79b81e6f8275ffba469d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cb6dd0699c4c7f286129bad70c5fa1e794096859
--- /dev/null
+++ b/records/sd1/sweep/79b81e6f8275ffba469d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:22fcbdaa553250866e52cab43853a4bab27867ffc4055d60659c0dff4ad2065e
+size 36514
diff --git a/records/sd1/sweep/79bb5e9467c193a6ff2b/output.webp b/records/sd1/sweep/79bb5e9467c193a6ff2b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e82ddf841f8107c91062e5c8a6a498f0baf3c60c
--- /dev/null
+++ b/records/sd1/sweep/79bb5e9467c193a6ff2b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e42be6d1b8b5c395e2e6d93e96c852ccd68a5bc3ea8d17873cbccffb7d9aa90c
+size 41284
diff --git a/records/sd1/sweep/79bca60931d854f8267e/output.webp b/records/sd1/sweep/79bca60931d854f8267e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..22156ee9138905a8d25ef70e7c55c0068c155dd7
--- /dev/null
+++ b/records/sd1/sweep/79bca60931d854f8267e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8249dbf17ad559225c488b439e7d5da94c22d4d01937f5186b7bd1d00a9b1c4e
+size 29920
diff --git a/records/sd1/sweep/79bddbc7bb3f67b0cfd5/output.webp b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d91f11667d04ab4b992d7fd871682af00c18b570
--- /dev/null
+++ b/records/sd1/sweep/79bddbc7bb3f67b0cfd5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15106258ccb24034fca81aad59f78d62700c97d9eb7b348dcc5869b9e784cc4d
+size 16002
diff --git a/records/sd1/sweep/79c42a428ad89b29f968/output.webp b/records/sd1/sweep/79c42a428ad89b29f968/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5e0f96aeda4b09e8cea74ecf47bf2b6681bb9bbf
--- /dev/null
+++ b/records/sd1/sweep/79c42a428ad89b29f968/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e58f90a1cb1c8f72419c34d3298ab7beef180d637946d265c427166f6ba49abb
+size 16396
diff --git a/records/sd1/sweep/79e83513845e84b8d66e/output.webp b/records/sd1/sweep/79e83513845e84b8d66e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cf6f09e0c56a6c13c6a7c85151baa009bf514274
--- /dev/null
+++ b/records/sd1/sweep/79e83513845e84b8d66e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77fcb23b1f528d263bf4ecb480de312aac3412cf84d48907b20ad228b5454cbc
+size 52966
diff --git a/records/sd1/sweep/79e988557b8a9032deff/output.webp b/records/sd1/sweep/79e988557b8a9032deff/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e496260593f85e8fbe7d779c9438111e4dac4bfa
--- /dev/null
+++ b/records/sd1/sweep/79e988557b8a9032deff/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6156440c67ec8ab0b095f8345b9f04b33f10ebacd657ab9b8bc7a87b1ce05d3
+size 64798
diff --git a/records/sd1/sweep/79efbb80d87066740d03/output.webp b/records/sd1/sweep/79efbb80d87066740d03/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8aa8c33ab4110c0b01991356dbb115b253ee7439
--- /dev/null
+++ b/records/sd1/sweep/79efbb80d87066740d03/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:760baa8c064dbca2ec2440b4f4898b6333b035144537d3202d1fede47399d8a8
+size 25678
diff --git a/records/sd1/sweep/79f668289f23f61ade87/output.webp b/records/sd1/sweep/79f668289f23f61ade87/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..33d0c8238c539b79d74e2342d35e21a162cbcaa6
--- /dev/null
+++ b/records/sd1/sweep/79f668289f23f61ade87/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02e482b5d0034cc37303543326ae37bb637ec39949d48ecfe2f19c71c94abb8e
+size 17420
diff --git a/records/sd1/sweep/79f99460c6f7d0efdc23/output.webp b/records/sd1/sweep/79f99460c6f7d0efdc23/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60afe5116b503cbab8e939d1c6a4b0f66a7c10a1
--- /dev/null
+++ b/records/sd1/sweep/79f99460c6f7d0efdc23/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b53f8a40e75b7ce5e25d99d93eb9d21799afacf7a064718a7c290ea869763c1
+size 52078
diff --git a/records/sd1/sweep/7a0964a0e5c2790131cc/output.webp b/records/sd1/sweep/7a0964a0e5c2790131cc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0aeac091f6606b2ea400a6c1a25fb86458b0cbc4
--- /dev/null
+++ b/records/sd1/sweep/7a0964a0e5c2790131cc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d64bb91770a9cb9f1cb8a9b556034524fd205290d80c3097cbca04053856207b
+size 36504
diff --git a/records/sd1/sweep/7a1315d467d4fa77ab6e/output.webp b/records/sd1/sweep/7a1315d467d4fa77ab6e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b6c24ce725bbd9f1270eae39f551c08962e88731
--- /dev/null
+++ b/records/sd1/sweep/7a1315d467d4fa77ab6e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c485b5e1a2b47cd734637ece2f1f4260b0268fcdea25748af623a811d3bd66a
+size 26106
diff --git a/records/sd1/sweep/7a13bfd6d929dee236d4/output.webp b/records/sd1/sweep/7a13bfd6d929dee236d4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4d42b91a0365351bf5786cb3ea6e3ac9c5b3fccc
--- /dev/null
+++ b/records/sd1/sweep/7a13bfd6d929dee236d4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9c5cef13596a84af6f3e8cd927a33e22f629e0afc15d581f48514e61f8a64a59
+size 25508
diff --git a/records/sd1/sweep/7a1581a30489d19b3043/output.webp b/records/sd1/sweep/7a1581a30489d19b3043/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6a49d5ce6efb606050d38b3bc496746c4a2ca403
--- /dev/null
+++ b/records/sd1/sweep/7a1581a30489d19b3043/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae8ce1170eea0300779009b70b089d0301b01e82f7b56ac0c27d3c7bb30b38db
+size 51766
diff --git a/records/sd1/sweep/7a1894c40c46a48e3ee1/output.webp b/records/sd1/sweep/7a1894c40c46a48e3ee1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b57927df1b5577482610f47aaee9c1e3b5b0ab6c
--- /dev/null
+++ b/records/sd1/sweep/7a1894c40c46a48e3ee1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9c14d0e22ca6643e112c70cc382a8a0a7846b5f844624bae46fb34681164522
+size 35942
diff --git a/records/sd1/sweep/7a18b4cfaff8cd6f78de/output.webp b/records/sd1/sweep/7a18b4cfaff8cd6f78de/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..28a0f0f12d168a10f0334bef60ff5db086f65b60
--- /dev/null
+++ b/records/sd1/sweep/7a18b4cfaff8cd6f78de/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a19ba44bbfe3a7cf33f73f69b5882e7a7da2a35565c2dd490866c70d8abeef60
+size 51434
diff --git a/records/sd1/sweep/7a19989508da994d1ef7/output.webp b/records/sd1/sweep/7a19989508da994d1ef7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bf7072c11568255c63ef5ae59da554bc73e7bb28
--- /dev/null
+++ b/records/sd1/sweep/7a19989508da994d1ef7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14e9807c2df343c3fade7ed17d9528d788ad413200001f95325f5b035b83976d
+size 45758
diff --git a/records/sd1/sweep/7a228879c70ba2d3e7b8/output.webp b/records/sd1/sweep/7a228879c70ba2d3e7b8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d155e9faf6a0b41a53528944cd04fc8584258cca
--- /dev/null
+++ b/records/sd1/sweep/7a228879c70ba2d3e7b8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f08e7556959ef6f8ea440882e73a3a4468ccbf7b5e554e9451108b7b5c1c055
+size 49176
diff --git a/records/sd1/sweep/7a27345027179e157a04/output.webp b/records/sd1/sweep/7a27345027179e157a04/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..13056087fdcc58aaf482cdf37740852c110e4adc
--- /dev/null
+++ b/records/sd1/sweep/7a27345027179e157a04/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d2089fefa51458bed98fbea181c82281db7795df0f9cea0b4d020bca21e61f4
+size 48592
diff --git a/records/sd1/sweep/7a2d9872e2a26b26c958/output.webp b/records/sd1/sweep/7a2d9872e2a26b26c958/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..967a985d4867c024a8ec2a96d8078b23d49bcc22
--- /dev/null
+++ b/records/sd1/sweep/7a2d9872e2a26b26c958/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce20c66edf3c369d5641afd9e4206819b096e20376c92f490b47efe139b86137
+size 29070
diff --git a/records/sd1/sweep/7a33270e8ef14b63f992/output.webp b/records/sd1/sweep/7a33270e8ef14b63f992/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..64543b960d10c5ee0d9d0a02c3aca2c1784e9122
--- /dev/null
+++ b/records/sd1/sweep/7a33270e8ef14b63f992/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79a521e12e267c1095d03461c3d04dee30f4b2ea2439e57bf29f6909fa1f26d7
+size 50920
diff --git a/records/sd1/sweep/7a3416cb61dd63c88ffb/output.webp b/records/sd1/sweep/7a3416cb61dd63c88ffb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..02a70145b82b4eb96c7b02c4a88fae0dc878dbcc
--- /dev/null
+++ b/records/sd1/sweep/7a3416cb61dd63c88ffb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:216f4d35f38a4136d6c14f8db414631db7e95d71400527fe7fc1da6df0e064d4
+size 36932
diff --git a/records/sd1/sweep/7a4148921238700896f0/output.webp b/records/sd1/sweep/7a4148921238700896f0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..19aa18eea3474aff89cdc214e506d75438cb0ee1
--- /dev/null
+++ b/records/sd1/sweep/7a4148921238700896f0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3477c68dadc68d271923bbce9a155765c91bd06a857851955f2c46201099edd9
+size 36502
diff --git a/records/sd1/sweep/7a4a7d080e3903d440e3/output.webp b/records/sd1/sweep/7a4a7d080e3903d440e3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2cd47c021f13df594dba0f62db9b97c0f35fc9f8
--- /dev/null
+++ b/records/sd1/sweep/7a4a7d080e3903d440e3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07100a67c86d67c9f17708ce27af9463237e1f0e840506749d397d06ad776d0c
+size 28918
diff --git a/records/sd1/sweep/7a5a403ce0ae1a54f662/output.webp b/records/sd1/sweep/7a5a403ce0ae1a54f662/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b4825569e31fc1887976a29dfb83868855d611b5
--- /dev/null
+++ b/records/sd1/sweep/7a5a403ce0ae1a54f662/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a8bf91570aba079184bc779d504fd87629ec20697506cac0f0355529155dc36
+size 36036
diff --git a/records/sd1/sweep/7a69a067a28b307cee47/output.webp b/records/sd1/sweep/7a69a067a28b307cee47/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ae1076266e977c1643725ce78a3ab39e086ce03f
--- /dev/null
+++ b/records/sd1/sweep/7a69a067a28b307cee47/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57d0da762232b77349b12ae632608d537763e526f1c83f970fe36e951ae1ca75
+size 37170
diff --git a/records/sd1/sweep/7a73fb2c5f58e07dfed5/output.webp b/records/sd1/sweep/7a73fb2c5f58e07dfed5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..abdaf784bbe78c81492a5eb73a10e14677d17c4a
--- /dev/null
+++ b/records/sd1/sweep/7a73fb2c5f58e07dfed5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bcce74df831d8408eb4f73d5f14e0c79a4409dbdd5d2ae62e8964618ffc6e75
+size 18088
diff --git a/records/sd1/sweep/7a78f3ef32605f72d134/output.webp b/records/sd1/sweep/7a78f3ef32605f72d134/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cff258ad5a9d2fc6e2b668cb7e36e4616f379581
--- /dev/null
+++ b/records/sd1/sweep/7a78f3ef32605f72d134/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f370f8ef0f9baef14ccf33e0d1f0a6e755cb79b6bab9bdd46a708891ced84e83
+size 13676
diff --git a/records/sd1/sweep/7a82c6fba5c9b3f87f98/output.webp b/records/sd1/sweep/7a82c6fba5c9b3f87f98/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a3a018c0e4af370b4f4565371ec9b2d4dc0f7a9c
--- /dev/null
+++ b/records/sd1/sweep/7a82c6fba5c9b3f87f98/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e29deeeb37f234ce953b75cd9e1f35ce1d8595ea324d1cd980d3284492504839
+size 45764
diff --git a/records/sd1/sweep/7a8664ad6fae32d43920/output.webp b/records/sd1/sweep/7a8664ad6fae32d43920/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3275a6c95ff5dcadeb9c2af7b7ddd1e1abf5c6b0
--- /dev/null
+++ b/records/sd1/sweep/7a8664ad6fae32d43920/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e82b731e8421c802a6f2265a60ed62e594542ba91b173875ccb4eaaec459de0
+size 29958
diff --git a/records/sd1/sweep/7a886d339ea6b3dbe607/output.webp b/records/sd1/sweep/7a886d339ea6b3dbe607/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e2559bb342dfed052140ff114665766b00fed5a5
--- /dev/null
+++ b/records/sd1/sweep/7a886d339ea6b3dbe607/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:75356a62068a4f6078b106f2ef1a5c645fd531dc9d523b47da69ab1f55b989de
+size 37348
diff --git a/records/sd1/sweep/7a889a9a355da6d2cdd7/output.webp b/records/sd1/sweep/7a889a9a355da6d2cdd7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d6226961d0f96c251a717789e695240540950123
--- /dev/null
+++ b/records/sd1/sweep/7a889a9a355da6d2cdd7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d839c5a7c40010101bf97abef71981a8eca9b6e61ce9e3083be6c2a849a03c04
+size 50042
diff --git a/records/sd1/sweep/7a955b96149cee8e642a/output.webp b/records/sd1/sweep/7a955b96149cee8e642a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d5353ccda517426b3742395052febca2d21b6142
--- /dev/null
+++ b/records/sd1/sweep/7a955b96149cee8e642a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9f99e7129c89bf48cf0815c920805f25bb26582d68f9a6f00bd78c70326ce0e
+size 39112
diff --git a/records/sd1/sweep/7a9ddb5ceee8b8f1b726/output.webp b/records/sd1/sweep/7a9ddb5ceee8b8f1b726/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..71d86e21a56411e5241790e7c1a3fa5e1b24f27f
--- /dev/null
+++ b/records/sd1/sweep/7a9ddb5ceee8b8f1b726/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33365fa0dd5b8c50388c1440c1f4cd344186d0ebd5e1fed9d6def8a4e3cfaeb9
+size 44952
diff --git a/records/sd1/sweep/7a9ebc4082a055fa0e84/output.webp b/records/sd1/sweep/7a9ebc4082a055fa0e84/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..31d164b86c8ff620fadc77f54582e91018d78393
--- /dev/null
+++ b/records/sd1/sweep/7a9ebc4082a055fa0e84/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6533941f26a3dc6750a9466e21127562efe7c3b6869e15204c4cace326a92c77
+size 48640
diff --git a/records/sd1/sweep/7aa4c40f33be07bd327e/output.webp b/records/sd1/sweep/7aa4c40f33be07bd327e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4efd04b962c2813bf46c2975b25531247a41e90b
--- /dev/null
+++ b/records/sd1/sweep/7aa4c40f33be07bd327e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4b77e897f8a65589e95c3c4e55c55ea602f19a76e29a880cf51b46c6ac08745
+size 58110
diff --git a/records/sd1/sweep/7ab096f8f6ce071602ba/output.webp b/records/sd1/sweep/7ab096f8f6ce071602ba/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3707c394a3b8ce17dfb06bbff37cb84b386474e7
--- /dev/null
+++ b/records/sd1/sweep/7ab096f8f6ce071602ba/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce03e137619793659b8710e5b5aaca792b41d3e59c23366d8813310af9ea6eed
+size 23502
diff --git a/records/sd1/sweep/7ab118ee77855780e8ed/output.webp b/records/sd1/sweep/7ab118ee77855780e8ed/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..25d7d82794f0eb383086db3b96d61260da893629
--- /dev/null
+++ b/records/sd1/sweep/7ab118ee77855780e8ed/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a4817578d5e201bc437d47d467137f4c0e645dac36e718d4105c11e38d5963e
+size 33002
diff --git a/records/sd1/sweep/7ab505d067ad9c9d3a84/output.webp b/records/sd1/sweep/7ab505d067ad9c9d3a84/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..18e35359d0ccd192a17ba8460728ab99ece08d1b
--- /dev/null
+++ b/records/sd1/sweep/7ab505d067ad9c9d3a84/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:100a822b92ed0e8a7dda0b1cb34dfb365d50ea983cdd52fbc8aaf363001eee5b
+size 33254
diff --git a/records/sd1/sweep/7ab9d87951152af10c91/output.webp b/records/sd1/sweep/7ab9d87951152af10c91/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bb2c1d3e717eaec3b638b5275d8a82dcf9370f42
--- /dev/null
+++ b/records/sd1/sweep/7ab9d87951152af10c91/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e066edc95d31bbe03813505965deeb72b29e0351ce315c996ceb6c7dad1bf22
+size 36040
diff --git a/records/sd1/sweep/7ac240b2d8a8773553c6/output.webp b/records/sd1/sweep/7ac240b2d8a8773553c6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5e172267758df24656907dd99234246be31c7d6e
--- /dev/null
+++ b/records/sd1/sweep/7ac240b2d8a8773553c6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1f4caa7682d31a10a13d990da3861e1c63afd1ddfbd5dc5ae12f7d34c3fe3efb
+size 46564
diff --git a/records/sd1/sweep/7ac302b27162e6fca655/output.webp b/records/sd1/sweep/7ac302b27162e6fca655/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8b698884ec5bc1698b2b0b58aac8089212f369e0
--- /dev/null
+++ b/records/sd1/sweep/7ac302b27162e6fca655/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dce638928d39c981f1139d46ab0b062a5f022509f78c61045eaeda4316215bbf
+size 40962
diff --git a/records/sd1/sweep/7ad3ba58abba210fb243/output.webp b/records/sd1/sweep/7ad3ba58abba210fb243/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d372749e615f0636eb0b4d6c3746e0aa2c863141
--- /dev/null
+++ b/records/sd1/sweep/7ad3ba58abba210fb243/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f73756f5e169e9d13ce607fa31da6d43299d77d3bc84cfc21e1115f5965b9a38
+size 37416
diff --git a/records/sd1/sweep/7adab8de040defd09569/output.webp b/records/sd1/sweep/7adab8de040defd09569/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..04b6a2eb92e1bd4b7238738689079b6772414ee2
--- /dev/null
+++ b/records/sd1/sweep/7adab8de040defd09569/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46df1d4e983c3bd0f7b18ede894735b30d35c50b49a0480d51ac05fce0927b34
+size 36922
diff --git a/records/sd1/sweep/7adaf90e0f93d6f2e258/output.webp b/records/sd1/sweep/7adaf90e0f93d6f2e258/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5c00342f13e673ab6090fb2f6dd0403074dac544
--- /dev/null
+++ b/records/sd1/sweep/7adaf90e0f93d6f2e258/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09bd7228fa10acba9e59cf4de69a967f5c1f0d732ab30dd5146373ae7cc3e8f5
+size 16698
diff --git a/records/sd1/sweep/7aea13f3d8b01cef3298/output.webp b/records/sd1/sweep/7aea13f3d8b01cef3298/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b766f480000ca03f3230e835fdd056db187d1cef
--- /dev/null
+++ b/records/sd1/sweep/7aea13f3d8b01cef3298/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9ccd05635e7127d4249f6791e6febb2446c39ddbc300f5d2afd1606559fc7e0
+size 32574
diff --git a/records/sd1/sweep/7af602dbe8f59abb215a/output.webp b/records/sd1/sweep/7af602dbe8f59abb215a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cc92918b516650392f1fa93c0c62241df9c0ca77
--- /dev/null
+++ b/records/sd1/sweep/7af602dbe8f59abb215a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d64fcec21bd643ca944a1f81ed425fc22f6d61324b0f06bf38d0d232043c21a9
+size 57420
diff --git a/records/sd1/sweep/7af90756ecbee8785544/output.webp b/records/sd1/sweep/7af90756ecbee8785544/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4f1428816f09cc36331d405d40ddee125878a9a8
--- /dev/null
+++ b/records/sd1/sweep/7af90756ecbee8785544/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c10f025f92fce1b7b03dab48f5ef4ee8fe366f888a49ccb8e503a6493af40e90
+size 53800
diff --git a/records/sd1/sweep/7afb7a97b3bec1b8e27f/output.webp b/records/sd1/sweep/7afb7a97b3bec1b8e27f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2f06a43df4d7ec8a03272dc4bc0faaa23b0a1b6e
--- /dev/null
+++ b/records/sd1/sweep/7afb7a97b3bec1b8e27f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0b8e01d2ab30a02e5022db76ecf8a770c68522b2c1a66d1d2273e25b18259fa
+size 46972
diff --git a/records/sd1/sweep/7afbcdca9f9562b35826/output.webp b/records/sd1/sweep/7afbcdca9f9562b35826/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..02c298779f5eb29513706edd3108da3bb396f375
--- /dev/null
+++ b/records/sd1/sweep/7afbcdca9f9562b35826/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4849e36dcfa39708a0600a5eb774a613b18f9ba41741a75e8b4c4963feec5fb
+size 44908
diff --git a/records/sd1/sweep/7b0362251dfb31174e6c/output.webp b/records/sd1/sweep/7b0362251dfb31174e6c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..09bc4a1dc8f684532e43add46a25c9a1c54a61b2
--- /dev/null
+++ b/records/sd1/sweep/7b0362251dfb31174e6c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6e025a85b4d982452f5a57025b1dc9b428b3155789ab07396a140d3bd7888c4
+size 19344
diff --git a/records/sd1/sweep/7b062a4a4cbb8ce68594/output.webp b/records/sd1/sweep/7b062a4a4cbb8ce68594/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..409003238b08c28af6ab3d2495aa6c557d1ff112
--- /dev/null
+++ b/records/sd1/sweep/7b062a4a4cbb8ce68594/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1baa8a2bd6032217dff040a466645775baa6ad3f689b87e4b461e722f67531e4
+size 38834
diff --git a/records/sd1/sweep/7b12eae7369a8b32e801/output.webp b/records/sd1/sweep/7b12eae7369a8b32e801/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d092eb35d960ad8eaa8c6e095f348899bf39dcb0
--- /dev/null
+++ b/records/sd1/sweep/7b12eae7369a8b32e801/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9d250e869cc1e285611348eaa48247b3df842ebad9682ae4a86b97995110c0b
+size 23294
diff --git a/records/sd1/sweep/7b1790f2e576f0b6f76c/output.webp b/records/sd1/sweep/7b1790f2e576f0b6f76c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9be12ca8eb87d877fc195dfed4a1668317688ab9
--- /dev/null
+++ b/records/sd1/sweep/7b1790f2e576f0b6f76c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3542bbc5cf8f26aff2cb66626a75f7c3471e3c2a503b1507b8367964e93a4b68
+size 23516
diff --git a/records/sd1/sweep/7b213ed31f073a758ac8/output.webp b/records/sd1/sweep/7b213ed31f073a758ac8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..154b7015889629cc9d3b4a96307c5e979b28a75b
--- /dev/null
+++ b/records/sd1/sweep/7b213ed31f073a758ac8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c910077ac5fceeef08f0d78b115220badb15e86c0fad9ce22d758fe274ad0155
+size 18212
diff --git a/records/sd1/sweep/7b258c306392cfeb06f3/output.webp b/records/sd1/sweep/7b258c306392cfeb06f3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..db568ff38c9e1d053103d63e6163eeb670e46f31
--- /dev/null
+++ b/records/sd1/sweep/7b258c306392cfeb06f3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e40bb9b2df457d66d1e973be7084a7190bff60056f145fd4d4e14e868dc34ddf
+size 41304
diff --git a/records/sd1/sweep/7b2781e0560763e67277/output.webp b/records/sd1/sweep/7b2781e0560763e67277/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e13a85c2888b46899badf63e43be3a962aa45469
--- /dev/null
+++ b/records/sd1/sweep/7b2781e0560763e67277/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec4b9909ca78c6270aaad511660138f84fa21aef0a96df70614930874c42256f
+size 21430
diff --git a/records/sd1/sweep/7b317283a63f45429603/output.webp b/records/sd1/sweep/7b317283a63f45429603/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e6be11fcab2ac6de27fa0bfc08b0a3c617b36fdd
--- /dev/null
+++ b/records/sd1/sweep/7b317283a63f45429603/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1afc9be15ec18173ea4554e577c797b769c283d385564a031d81c823863aae6
+size 24972
diff --git a/records/sd1/sweep/7b321e9db4258622f239/output.webp b/records/sd1/sweep/7b321e9db4258622f239/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..37a57442c45bd123a7f784410bb06451567c1fe4
--- /dev/null
+++ b/records/sd1/sweep/7b321e9db4258622f239/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:631f01901e0248322022e69f78e7481edf2443a362ce699828d573e95d8e84e9
+size 11666
diff --git a/records/sd1/sweep/7b354de3897e344e4bdf/output.webp b/records/sd1/sweep/7b354de3897e344e4bdf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..272ed1e15022d8358d1af7d1f594a496b36eebfd
--- /dev/null
+++ b/records/sd1/sweep/7b354de3897e344e4bdf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:01c3fc3a95aa340f61d45cb360f173d11ffaf3049fb22eacd16a1379018f22fb
+size 19406
diff --git a/records/sd1/sweep/7b390377065b5d096483/output.webp b/records/sd1/sweep/7b390377065b5d096483/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8df9b0b5d2baf1dd1d94745c635083ccbd427b9b
--- /dev/null
+++ b/records/sd1/sweep/7b390377065b5d096483/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1da5e77bf719d93ff025d76753b944e9f5580ad812eb8735ccc97501890e8e3c
+size 30382
diff --git a/records/sd1/sweep/7b3f4c2c5bd2bc0f66e2/output.webp b/records/sd1/sweep/7b3f4c2c5bd2bc0f66e2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..77ab07567fe40a627f74691dc4170fb39e0230c1
--- /dev/null
+++ b/records/sd1/sweep/7b3f4c2c5bd2bc0f66e2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c54215ba9b5dd94bf5e0faaed20dbae56e3ba691688c91657233f6d1b189b70
+size 39618
diff --git a/records/sd1/sweep/7b419609f6488f7143c8/output.webp b/records/sd1/sweep/7b419609f6488f7143c8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c4c624ff447590851aac2fef8b0a28318167fc76
--- /dev/null
+++ b/records/sd1/sweep/7b419609f6488f7143c8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2b2e724b7960e81f8f57eb9b85ae0588e79526bb73c42fb2f1edb3a7ff49a5c
+size 51464
diff --git a/records/sd1/sweep/7b5e6a90be7a7a4be989/output.webp b/records/sd1/sweep/7b5e6a90be7a7a4be989/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..38363438fcec7eef53355fb9d8da83eb4f989720
--- /dev/null
+++ b/records/sd1/sweep/7b5e6a90be7a7a4be989/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:089facd215f035a49771e6f327f23e4258d9cadcef61d7821090af477fc34f3e
+size 49880
diff --git a/records/sd1/sweep/7b60a499af9058474275/output.webp b/records/sd1/sweep/7b60a499af9058474275/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2d79ae2e9c3152b68ccf328ca92deb5fb0fed208
--- /dev/null
+++ b/records/sd1/sweep/7b60a499af9058474275/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac20dfe39bdc4b2606ed68f260bc58538d92c9c4be0487db80e4b36f44a731f3
+size 19190
diff --git a/records/sd1/sweep/7b615cb42669d1c1dbdd/output.webp b/records/sd1/sweep/7b615cb42669d1c1dbdd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..217364f7834f2ba13ad17a55bdd514140c234956
--- /dev/null
+++ b/records/sd1/sweep/7b615cb42669d1c1dbdd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49f113c7ab9b8aa36cf22b0801f5fd65c29bf7aa1eb12ef7638e87d20cbfe68d
+size 49240
diff --git a/records/sd1/sweep/7b6f06988175ca514b8c/output.webp b/records/sd1/sweep/7b6f06988175ca514b8c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8d5bd80b3335056f505bf1f66014e9713d6d1f99
--- /dev/null
+++ b/records/sd1/sweep/7b6f06988175ca514b8c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46e391aa39c5628085e6695085c0e831f49853dc7fce35027aa17ec3de88d1ad
+size 32568
diff --git a/records/sd1/sweep/7b75c7f679a722028d50/output.webp b/records/sd1/sweep/7b75c7f679a722028d50/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2b41c60101860a02b6b5eeb0524a643fd8c0c420
--- /dev/null
+++ b/records/sd1/sweep/7b75c7f679a722028d50/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3321f6a17348f41a9cb2ad1f59a529e1694a2e7cf8460971389076c08fa67a1e
+size 21260
diff --git a/records/sd1/sweep/7b8506ac28b8f6037cd6/output.webp b/records/sd1/sweep/7b8506ac28b8f6037cd6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..22ee34ac9c138a4f6641aae0292016c15b537022
--- /dev/null
+++ b/records/sd1/sweep/7b8506ac28b8f6037cd6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bea8e6b47c535afb73249d2be3e251c0390ae565766c5b03a12126fd59f05ac1
+size 11404
diff --git a/records/sd1/sweep/7ba67ebb6f8cfc020dc7/output.webp b/records/sd1/sweep/7ba67ebb6f8cfc020dc7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..db2c27e338c4d1217ac6ffa22e64c792580451df
--- /dev/null
+++ b/records/sd1/sweep/7ba67ebb6f8cfc020dc7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a7106315a10c5086de0a09a0ec229a08d8ecec04e80c16b223ffb99addae787a
+size 51330
diff --git a/records/sd1/sweep/7baa2110d6a3dc342f8b/output.webp b/records/sd1/sweep/7baa2110d6a3dc342f8b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..93c7f994e86f4d90ebaf5b12c35ad309f2d72fb5
--- /dev/null
+++ b/records/sd1/sweep/7baa2110d6a3dc342f8b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e96567571ae1436b5272cdad62e5d035ddad7a29d1131a204df479759971d784
+size 19514
diff --git a/records/sd1/sweep/7bb15545d904231e38cc/output.webp b/records/sd1/sweep/7bb15545d904231e38cc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5491c83af9ae433bfa8308906be47ff137afafda
--- /dev/null
+++ b/records/sd1/sweep/7bb15545d904231e38cc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb3af51a5f5b13e78fe2a293ec6195cd1a4604790cc2818c317c09d9e323fcb1
+size 35284
diff --git a/records/sd1/sweep/7bb425fbee630bef7540/output.webp b/records/sd1/sweep/7bb425fbee630bef7540/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..93455acaf01b7a89f875a8a40632a73dca3e1ba0
--- /dev/null
+++ b/records/sd1/sweep/7bb425fbee630bef7540/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15b9f202ea9b89e3e7bf3c965802d16680df6e3e2609fd1f783be777c0841220
+size 58870
diff --git a/records/sd1/sweep/7bb73d3ef2ee91d490c1/output.webp b/records/sd1/sweep/7bb73d3ef2ee91d490c1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..54ae47388813aa3cd0dd91717c3644d644f91db5
--- /dev/null
+++ b/records/sd1/sweep/7bb73d3ef2ee91d490c1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cfc6f70e2db482dc2514032a3123c90de4d18f24744808ab3c165019510a791
+size 23776
diff --git a/records/sd1/sweep/7bc556fc05356be00fdb/output.webp b/records/sd1/sweep/7bc556fc05356be00fdb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f916248de9d16d8f60fd2e157c0c358b8b3ebf44
--- /dev/null
+++ b/records/sd1/sweep/7bc556fc05356be00fdb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6acb453c0746501a15a8b60c4d6fd0b6f9f25c8589681b2f8a1c5a5088841828
+size 26492
diff --git a/records/sd1/sweep/7bcf9b624b46079a00ba/output.webp b/records/sd1/sweep/7bcf9b624b46079a00ba/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..094ba0905781868091df3e316caa22b04fa24022
--- /dev/null
+++ b/records/sd1/sweep/7bcf9b624b46079a00ba/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db5ec3e9444bf933018e6d637c975c1ab50912af6e3ad15537f8e062ba2d4b04
+size 32738
diff --git a/records/sd1/sweep/7bd2063124d3d63c0a18/output.webp b/records/sd1/sweep/7bd2063124d3d63c0a18/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e5e9340fc4914c3ff9fea894342201e3025fb9c8
--- /dev/null
+++ b/records/sd1/sweep/7bd2063124d3d63c0a18/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3adaa2c779138d8dfbe945edf5a0ec11abbf1045468d09181297bd90e55280c8
+size 55792
diff --git a/records/sd1/sweep/7bdede0ed1695005bba9/output.webp b/records/sd1/sweep/7bdede0ed1695005bba9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e5030c236e9f7d72a1e272a6a4686ddbd4338706
--- /dev/null
+++ b/records/sd1/sweep/7bdede0ed1695005bba9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5699b9050ac108baff78429b12d51baaad637d1e6de680b753d7014362995a17
+size 33966
diff --git a/records/sd1/sweep/7be5153507d604ad2a2f/output.webp b/records/sd1/sweep/7be5153507d604ad2a2f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..30c59c6e0f9c1ebcfe8eaea34ae6af5f0e5bf313
--- /dev/null
+++ b/records/sd1/sweep/7be5153507d604ad2a2f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58d572ab8ed80d5e750bf0e4360d9aa7bbb1e988f7c99ff7db7c4a516237513e
+size 22608
diff --git a/records/sd1/sweep/7bf30c4d4a3ddf2ddd2b/output.webp b/records/sd1/sweep/7bf30c4d4a3ddf2ddd2b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5d30ad1b6820020c14fe2dab8ccd0325e76a6cf5
--- /dev/null
+++ b/records/sd1/sweep/7bf30c4d4a3ddf2ddd2b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12f2b5b3a2d14ee148cfd0be085ce082ba1ee3b735ac7804b67816ac467dd01f
+size 59120
diff --git a/records/sd1/sweep/7bfa1d205bba6125f8e6/output.webp b/records/sd1/sweep/7bfa1d205bba6125f8e6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..843255757d97037ffd240a59dd5bdc28541dacb8
--- /dev/null
+++ b/records/sd1/sweep/7bfa1d205bba6125f8e6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63d424737148bc6be54b58e1ff84b4436cfa06536c3f7b669cb3ce6c5c192512
+size 52834
diff --git a/records/sd1/sweep/7bfabf55072afc32bf3a/output.webp b/records/sd1/sweep/7bfabf55072afc32bf3a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6568f4ae380319a493b6b824506901eecc07332f
--- /dev/null
+++ b/records/sd1/sweep/7bfabf55072afc32bf3a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a139007d690231f15a4f01ece4c05d318e4f126d08d96d3dbd8539155d462f9b
+size 25060
diff --git a/records/sd1/sweep/7c0168830c91bc396766/output.webp b/records/sd1/sweep/7c0168830c91bc396766/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5c9fb6b37349ee97b68d8d5b4a6770f68a9a93d5
--- /dev/null
+++ b/records/sd1/sweep/7c0168830c91bc396766/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d21b19f87c228043f8c70ef45fac77bf210460056a8c4de15e9c97adb78c2ed4
+size 33134
diff --git a/records/sd1/sweep/7c03243e4a93ebc41ee4/output.webp b/records/sd1/sweep/7c03243e4a93ebc41ee4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..34cf2ba382d4f3fe91a95f9d355becc70de0754c
--- /dev/null
+++ b/records/sd1/sweep/7c03243e4a93ebc41ee4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8373243daed4b33b6e89fabf51fc022bda924c9492b5a2dc6f0311a17cb7af89
+size 58052
diff --git a/records/sd1/sweep/7c0ccf39fe5c1818063c/output.webp b/records/sd1/sweep/7c0ccf39fe5c1818063c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1c211c2cf2c5bb7759a333be73ceb2e5a68db1af
--- /dev/null
+++ b/records/sd1/sweep/7c0ccf39fe5c1818063c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ba6dc5900dc936d8fdb726ccdce368f674f6b1b342080c8edf717454ab136b1
+size 32454
diff --git a/records/sd1/sweep/7c1021781fb16ead9ca0/output.webp b/records/sd1/sweep/7c1021781fb16ead9ca0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2e79f4e4c9fdf109977e3fc33bdbdc6c0887553a
--- /dev/null
+++ b/records/sd1/sweep/7c1021781fb16ead9ca0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b7aa48d8acade9e83b2fbffe40532e9576d42e6af0cb9c8c7e42ae5d6f9f17a
+size 38198
diff --git a/records/sd1/sweep/7c19f4b89c1f78d05542/output.webp b/records/sd1/sweep/7c19f4b89c1f78d05542/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f5baa22c722ecbc94406d11c5bee0d800883e22e
--- /dev/null
+++ b/records/sd1/sweep/7c19f4b89c1f78d05542/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a17f53cf6b46aad5df0596bf0be73c371a5d498fc3a8a1072098bb80bdf4f1dd
+size 50582
diff --git a/records/sd1/sweep/7c28b7a84d37d858eb32/output.webp b/records/sd1/sweep/7c28b7a84d37d858eb32/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4d8c5e4c1fb91120fd0eb58768d41aaa30ab5f4b
--- /dev/null
+++ b/records/sd1/sweep/7c28b7a84d37d858eb32/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:336d89a53549934cee8dcc12e23ecaebd0e71a26b2a6d4456c25add423f95092
+size 19932
diff --git a/records/sd1/sweep/7c29f7cf50a498072339/output.webp b/records/sd1/sweep/7c29f7cf50a498072339/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..93fdb490415f0f57c26f3f264ccab71c9e2a6875
--- /dev/null
+++ b/records/sd1/sweep/7c29f7cf50a498072339/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95368d1654a4774489d26e8a0289262bed21f500f4ed8bc448e3c81c970da2f3
+size 38724
diff --git a/records/sd1/sweep/7c2c7a5618f0f3c6c29b/output.webp b/records/sd1/sweep/7c2c7a5618f0f3c6c29b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..20c1e54304564e26328c3c9a3a429c27fe3a9e96
--- /dev/null
+++ b/records/sd1/sweep/7c2c7a5618f0f3c6c29b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:594a50cf86ad8cff84300a749fe84f60402147b7806775c73bbc75db60556d31
+size 46192
diff --git a/records/sd1/sweep/7c2e2cbf18f26a58fa1c/output.webp b/records/sd1/sweep/7c2e2cbf18f26a58fa1c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bfa28873fa3f0f6bb758a2807c8e21d2db8672dd
--- /dev/null
+++ b/records/sd1/sweep/7c2e2cbf18f26a58fa1c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:537a618143f1046f47d6dc1857f6fa082cd9607c461590dad9270c803b23ca7d
+size 38222
diff --git a/records/sd1/sweep/7c360b1e58219a3dc2aa/output.webp b/records/sd1/sweep/7c360b1e58219a3dc2aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e32d4c09133525f87bdd91edc04c1cc31d00e57b
--- /dev/null
+++ b/records/sd1/sweep/7c360b1e58219a3dc2aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d0fa38d0de82ae06485ead7a32031f75041dca85839f6600bfdf5ea1850f4c4
+size 46606
diff --git a/records/sd1/sweep/7c4034772b5c35e06294/output.webp b/records/sd1/sweep/7c4034772b5c35e06294/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..96fb5c628cda5653dd533c2c0f49636de30183ed
--- /dev/null
+++ b/records/sd1/sweep/7c4034772b5c35e06294/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ace4d35c2ed6f1e037f51052eafbff8f131903d535e7a407b79f98ecd8215504
+size 54758
diff --git a/records/sd1/sweep/7c42f83753a58534f197/output.webp b/records/sd1/sweep/7c42f83753a58534f197/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..28fb282d87aab61d121f9b78d132215c1a3d94ff
--- /dev/null
+++ b/records/sd1/sweep/7c42f83753a58534f197/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6a5bb5e4caee320e98762eb5961a34a714940d932820bc56bf15d2c89a3996b1
+size 16906
diff --git a/records/sd1/sweep/7c45316515329487e00a/output.webp b/records/sd1/sweep/7c45316515329487e00a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..39d2f5548a121d2819c41d137de279968acd9724
--- /dev/null
+++ b/records/sd1/sweep/7c45316515329487e00a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de64bf9f687720ee5d0122668bc54f474797e2b916d72858a5080e36c6c470f2
+size 35538
diff --git a/records/sd1/sweep/7c526e182c1ce49ac8c6/output.webp b/records/sd1/sweep/7c526e182c1ce49ac8c6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b2af6bd1e76f0da76483e67ae8576c1dda90ef5c
--- /dev/null
+++ b/records/sd1/sweep/7c526e182c1ce49ac8c6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:add9b7e67fa6a3d461b5d118dbc44bcce094b2f1b609f572483e8abc98370fef
+size 46884
diff --git a/records/sd1/sweep/7c5f04bbb196fae9ddc1/output.webp b/records/sd1/sweep/7c5f04bbb196fae9ddc1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..906c87bd1ecea129674199c54ea9bd256540bb66
--- /dev/null
+++ b/records/sd1/sweep/7c5f04bbb196fae9ddc1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d71c7a11fa9980d6f20ba5f36db354c552b8cad978357ddb0071def7835cf195
+size 23324
diff --git a/records/sd1/sweep/7c63e4b27966a4e9e3c4/output.webp b/records/sd1/sweep/7c63e4b27966a4e9e3c4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..13af5a334c1ad6063355bd67646635605b4f8b77
--- /dev/null
+++ b/records/sd1/sweep/7c63e4b27966a4e9e3c4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e2d87bcbbd79d20f4afeadcd40c52fd88ea9779e7bd0e06cd79f7b4dda7a77f
+size 46930
diff --git a/records/sd1/sweep/7c65bb86bf26148b534b/output.webp b/records/sd1/sweep/7c65bb86bf26148b534b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ce672760caff8d197d72aa38d88aa5da5afda82f
--- /dev/null
+++ b/records/sd1/sweep/7c65bb86bf26148b534b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9c5eaddfd4dda32d5098bca6f134c4ce2aff006e900025a6ee6d31895d77b0f
+size 36664
diff --git a/records/sd1/sweep/7c7e6ba30bd8c69ee3ee/output.webp b/records/sd1/sweep/7c7e6ba30bd8c69ee3ee/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..02f157d2128b99d76476581e4a936e105c1dfa47
--- /dev/null
+++ b/records/sd1/sweep/7c7e6ba30bd8c69ee3ee/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42925af177685f9d96b8dd3c3fe7b0620fbed0af7ff62fbaffc558aa7e32b35d
+size 53616
diff --git a/records/sd1/sweep/7c8f6c40e212b2accfec/output.webp b/records/sd1/sweep/7c8f6c40e212b2accfec/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..df1373efeb42c7ed51d0a0618985b65277604991
--- /dev/null
+++ b/records/sd1/sweep/7c8f6c40e212b2accfec/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e97dd26a9782e9459911c4551aa86cd271e51a868167fec8b1b4e3f2c1a7563
+size 14814
diff --git a/records/sd1/sweep/7c9ebb19082ae47ef539/output.webp b/records/sd1/sweep/7c9ebb19082ae47ef539/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7a7d54acbb8c4d58660a106de71243bd1eeb5a0f
--- /dev/null
+++ b/records/sd1/sweep/7c9ebb19082ae47ef539/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51af65a5bd980f44ac08dc6fcae3516f65e4ef8cc5736086c7cd5cf24fca3941
+size 23562
diff --git a/records/sd1/sweep/7ca4570099a6c23d8573/output.webp b/records/sd1/sweep/7ca4570099a6c23d8573/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..25e77a43fff18f888d3e1f268827d26a0e0a2615
--- /dev/null
+++ b/records/sd1/sweep/7ca4570099a6c23d8573/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d9f18b553c9f8afc893f28121fa55b09fb3986e135923331177e8a00c9b152f
+size 26592
diff --git a/records/sd1/sweep/7cad196330025554fe2d/output.webp b/records/sd1/sweep/7cad196330025554fe2d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4704cb274ff51bafa5a12e93a64a36517568ce61
--- /dev/null
+++ b/records/sd1/sweep/7cad196330025554fe2d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93f912a16b64135df634d972369d62f36b8c5eae8fe47cc18e9f818e167d69da
+size 50808
diff --git a/records/sd1/sweep/7cb25f75b35afd202707/output.webp b/records/sd1/sweep/7cb25f75b35afd202707/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..61ac46dec832417fb39be0b04a6e8529ae4a3371
--- /dev/null
+++ b/records/sd1/sweep/7cb25f75b35afd202707/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:946395737364e17ea535f37186d0341d892db706ec19a84d0e5c990a1bc1bba4
+size 26558
diff --git a/records/sd1/sweep/7cb44a51584ede65d285/output.webp b/records/sd1/sweep/7cb44a51584ede65d285/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fc6963a47fda1f4f6b713e69afc009eacd63a4da
--- /dev/null
+++ b/records/sd1/sweep/7cb44a51584ede65d285/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73403e4939ba0a83721997554d56ec1e3501bd9a8241d19a303101e0cb6f2523
+size 43392
diff --git a/records/sd1/sweep/7cce9a6180b4a2b22fb7/output.webp b/records/sd1/sweep/7cce9a6180b4a2b22fb7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0a1f0c641e1c8db9737a3de8303e5aaf0ce127ba
--- /dev/null
+++ b/records/sd1/sweep/7cce9a6180b4a2b22fb7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a75cfa5a554257c5325ed326aa3ae6aaff43cfb08843ea7b15df563dd4877e44
+size 44526
diff --git a/records/sd1/sweep/7ccf76865963a7bcfa55/output.webp b/records/sd1/sweep/7ccf76865963a7bcfa55/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2dff5a5eccae7321210dfdd98fa927cc91313dc3
--- /dev/null
+++ b/records/sd1/sweep/7ccf76865963a7bcfa55/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dfe08889aed47d35cb64a24bb2709ae28907b74ed71e63e256c16bc6ef2a363
+size 18672
diff --git a/records/sd1/sweep/7cd7520cf52de3fc993b/output.webp b/records/sd1/sweep/7cd7520cf52de3fc993b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c761efb93a4bdc6864f3e0a0d202732820441c01
--- /dev/null
+++ b/records/sd1/sweep/7cd7520cf52de3fc993b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e945cbf5886292b311fda2e8d971039adb242036191d8c4e06b290cf2a3c9ec
+size 24552
diff --git a/records/sd1/sweep/7ce8b6065dadef284759/output.webp b/records/sd1/sweep/7ce8b6065dadef284759/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7dc69e4a35553743139dabb79bd8fc02ceae39d0
--- /dev/null
+++ b/records/sd1/sweep/7ce8b6065dadef284759/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82b802ff2bb784bd9ac45ec6e82a367875859f453c77f9938041f3e38d36fe35
+size 46276
diff --git a/records/sd1/sweep/7ceb2a4f489093708ec2/output.webp b/records/sd1/sweep/7ceb2a4f489093708ec2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bbb20dbe7af2842199e28f42465c78e739c24f70
--- /dev/null
+++ b/records/sd1/sweep/7ceb2a4f489093708ec2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8839e8b604efb273770ab54e313f43104cdb658d712b1c7b5e3ae1d2357d9af5
+size 18900
diff --git a/records/sd1/sweep/7cf3b9c076b297497883/output.webp b/records/sd1/sweep/7cf3b9c076b297497883/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..75852ec849b304ded33bc5b9262f61893aedc0d6
--- /dev/null
+++ b/records/sd1/sweep/7cf3b9c076b297497883/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e25c737b75352aa49762e6822816c5b700039d057a158429899453a33c6feff0
+size 17798
diff --git a/records/sd1/sweep/7cfd4f87f11d9e37aa0a/output.webp b/records/sd1/sweep/7cfd4f87f11d9e37aa0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c3703311d476079ed969ce92ed6c75d8cc899b29
--- /dev/null
+++ b/records/sd1/sweep/7cfd4f87f11d9e37aa0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cf6ece0308c806b167a96ec20bfe330e6423046bd5972deea7f86b1618f6dc6
+size 46410
diff --git a/records/sd1/sweep/7cfe24f508a7a9bbb936/output.webp b/records/sd1/sweep/7cfe24f508a7a9bbb936/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4f0be673d3bc8344ae11024dac86939abd90189d
--- /dev/null
+++ b/records/sd1/sweep/7cfe24f508a7a9bbb936/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf2faa0a990e0ef1eec0a6d1fcc2b398939096d815537af6ff48068a84037f4d
+size 19932
diff --git a/records/sd1/sweep/7d00c99f7ef150a7de0c/output.webp b/records/sd1/sweep/7d00c99f7ef150a7de0c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e1e4cc0f8475cefa4b84550e4d688dc144b76c45
--- /dev/null
+++ b/records/sd1/sweep/7d00c99f7ef150a7de0c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a68f69452040ef13af39120a6e6b53efc42d710ceeeaef1811816b578305ef9a
+size 14770
diff --git a/records/sd1/sweep/7d0dfa26b659bfd0443c/output.webp b/records/sd1/sweep/7d0dfa26b659bfd0443c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..819a15c6b0beac2e497eb3065d6c538c4ffc2eb3
--- /dev/null
+++ b/records/sd1/sweep/7d0dfa26b659bfd0443c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f68885230baed367bbdf19451e0d0242427a4bcd1530de86ffbcd646e6ad1db9
+size 46000
diff --git a/records/sd1/sweep/7d12cb89585e81577bd5/output.webp b/records/sd1/sweep/7d12cb89585e81577bd5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6d5cb7a5fd2e2348f77010ffb113d60f7e1e9488
--- /dev/null
+++ b/records/sd1/sweep/7d12cb89585e81577bd5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f61e4f2cdcc9417d603616da0ebbc2c0aff3981ddb9e332e10fced3ab529729
+size 24870
diff --git a/records/sd1/sweep/7d1d90ebc8988706e386/output.webp b/records/sd1/sweep/7d1d90ebc8988706e386/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..74161116c221ed1bce4f2c7a4fc05611f2d37ac4
--- /dev/null
+++ b/records/sd1/sweep/7d1d90ebc8988706e386/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a811f1f9e964422f2d29eea84d6d3a74a3e9e195771fd8dcc4eff74d35e67f57
+size 33032
diff --git a/records/sd1/sweep/7d1dbec87f3b5590f4db/output.webp b/records/sd1/sweep/7d1dbec87f3b5590f4db/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5b25380d333bbcad894e2dac428f0e855c74a287
--- /dev/null
+++ b/records/sd1/sweep/7d1dbec87f3b5590f4db/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa0344205f1ab518649e09edf5f6bb67d89e6a0c88c5ade36ad5ca731e5b7b3f
+size 17794
diff --git a/records/sd1/sweep/7d2c5afd7129eda5ff5b/output.webp b/records/sd1/sweep/7d2c5afd7129eda5ff5b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..48b46b1e586e9c7ad4c09e859d4d5ee94ad321f3
--- /dev/null
+++ b/records/sd1/sweep/7d2c5afd7129eda5ff5b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db1cb9a981c933303f91c0a5e1980452a6d57abe7a3ccfcdd72c1b96bb06e55e
+size 17764
diff --git a/records/sd1/sweep/7d338af20e707aa9a435/output.webp b/records/sd1/sweep/7d338af20e707aa9a435/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1da0cc6d6323be34f010ccc57d933bfccb1c4356
--- /dev/null
+++ b/records/sd1/sweep/7d338af20e707aa9a435/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a41718c51a6af8f46a081c3772bf05e371d4fee73d2809bd0c9f304de6238bc
+size 40358
diff --git a/records/sd1/sweep/7d34fbac9d24c0726b2e/output.webp b/records/sd1/sweep/7d34fbac9d24c0726b2e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dc17df092d7893f7f8ad0b7645291688f29e4f52
--- /dev/null
+++ b/records/sd1/sweep/7d34fbac9d24c0726b2e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a92ebfa4ad8381b90e2e4402d4816a35b946091871967e807f43dbf062bec395
+size 50084
diff --git a/records/sd1/sweep/7d39182d6c40e8a53c20/output.webp b/records/sd1/sweep/7d39182d6c40e8a53c20/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f6e5ad0c88b83523840ccd03aa3285f8e7d6b076
--- /dev/null
+++ b/records/sd1/sweep/7d39182d6c40e8a53c20/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:352e4c4b51fc3688b66d722b6589861541bbef6e885500fdab4b1151f16a967e
+size 39516
diff --git a/records/sd1/sweep/7d42b36ca31da2fbfa96/output.webp b/records/sd1/sweep/7d42b36ca31da2fbfa96/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..495eb8be3e38b2f0dc2d1b63f60522966f9f0a9f
--- /dev/null
+++ b/records/sd1/sweep/7d42b36ca31da2fbfa96/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:df7db1e4b7972d53c6cd7733805b023ae98b33927e71dc648c1f1f16aef049c1
+size 28050
diff --git a/records/sd1/sweep/7d42c82c41b8f6b05866/output.webp b/records/sd1/sweep/7d42c82c41b8f6b05866/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dc2c9da0e7d0f19ad63c1d1b0116eb08f5c568b8
--- /dev/null
+++ b/records/sd1/sweep/7d42c82c41b8f6b05866/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4648f5a59bfddaacd37fe79ba777effa7b66ccdd66df316d94bd2957fc02a8ca
+size 23214
diff --git a/records/sd1/sweep/7d528f08de08471a7a44/output.webp b/records/sd1/sweep/7d528f08de08471a7a44/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fdde578f182827cefa9f642f38d2e5f7441359e8
--- /dev/null
+++ b/records/sd1/sweep/7d528f08de08471a7a44/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90dbb67f704c0bcc63c201a1ec17616ae798aa50bcafd26d53c822a847cfd4d6
+size 22552
diff --git a/records/sd1/sweep/7d5685689a992bcdf9dd/output.webp b/records/sd1/sweep/7d5685689a992bcdf9dd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..de1f8380b862c0660c0e6cb50387d97256d693e7
--- /dev/null
+++ b/records/sd1/sweep/7d5685689a992bcdf9dd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48ecf7f1957fa3df266a72c012300c7d644fb9413d2af79d4058392c91df2492
+size 16494
diff --git a/records/sd1/sweep/7d636c68e9decd7270b1/output.webp b/records/sd1/sweep/7d636c68e9decd7270b1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..13031377bd058950f940200b801eb953aed6eff8
--- /dev/null
+++ b/records/sd1/sweep/7d636c68e9decd7270b1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c98b4264a74ce019c83c6d72eeaaaaf6b8855610c3d8cf312c1f89a51af22f1
+size 31214
diff --git a/records/sd1/sweep/7d74ff1d2ba0e7deba9e/output.webp b/records/sd1/sweep/7d74ff1d2ba0e7deba9e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7e08d29655b68b09b89528535a8a8453afb05066
--- /dev/null
+++ b/records/sd1/sweep/7d74ff1d2ba0e7deba9e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1feeda0c79d760c8149bf4ec9e57f869c291a6d01cb4e7ba9fb9f31d9d95e2ac
+size 44450
diff --git a/records/sd1/sweep/7d7637096f3402a52231/output.webp b/records/sd1/sweep/7d7637096f3402a52231/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1b6eb010c10889ad19a89e2da0f6a74814fb890d
--- /dev/null
+++ b/records/sd1/sweep/7d7637096f3402a52231/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:876d16d798264ada37505e1d967c6834db2e5acafc404cc07b2d4d3e450b3a56
+size 19932
diff --git a/records/sd1/sweep/7d84713adec3504e40c8/output.webp b/records/sd1/sweep/7d84713adec3504e40c8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3a4d842d9874cc9c7ccb2c90d494aedf31509bce
--- /dev/null
+++ b/records/sd1/sweep/7d84713adec3504e40c8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7597f4df7ecd923f41d127294b6cf3c698bdc1841eb1a4fe3c28daff320331b7
+size 17906
diff --git a/records/sd1/sweep/7d84b059fce9a9f1b272/output.webp b/records/sd1/sweep/7d84b059fce9a9f1b272/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..57fe1c0951e09152b3b03c53c7a4e6cb13a591f4
--- /dev/null
+++ b/records/sd1/sweep/7d84b059fce9a9f1b272/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dd2f8b4e3c22ef9b19741b1c394dfbbaf8cda742b56296fb04b28bb95497017
+size 27362
diff --git a/records/sd1/sweep/7d864a581b2602b4e58b/output.webp b/records/sd1/sweep/7d864a581b2602b4e58b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cc48e9ecf38fc00341124daceb8977b53790ed51
--- /dev/null
+++ b/records/sd1/sweep/7d864a581b2602b4e58b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41ae71a521cdd3dc7a6806234112050ebc517bac40004c507a4f5882eebbd62f
+size 35682
diff --git a/records/sd1/sweep/7d9e8e38966db9cd0ca2/output.webp b/records/sd1/sweep/7d9e8e38966db9cd0ca2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1c0eab4dd4068a29de255ae89d964c9582640b5a
--- /dev/null
+++ b/records/sd1/sweep/7d9e8e38966db9cd0ca2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b8a98ab690df82b7a6df55e9981d7ae52207e45d1654b0a5f68f5b7ae29c2c8
+size 44086
diff --git a/records/sd1/sweep/7da64d85e5a390a0793a/output.webp b/records/sd1/sweep/7da64d85e5a390a0793a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..407885ac4b12d5e01dfb3e86fc0b54f1b92ab92a
--- /dev/null
+++ b/records/sd1/sweep/7da64d85e5a390a0793a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bf46db0ebf6d9ed0788763f1896a7ea49a51ae6a36b6981341c402a1a838cda
+size 52828
diff --git a/records/sd1/sweep/7dadd607b2e1f7d89333/output.webp b/records/sd1/sweep/7dadd607b2e1f7d89333/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..70d8033ccc4cbaddbba25af1351b26c04730bbc4
--- /dev/null
+++ b/records/sd1/sweep/7dadd607b2e1f7d89333/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af3074b19764816a83e5961551380c677edcf696d599485014faf1065ffcdf59
+size 42072
diff --git a/records/sd1/sweep/7db57587e704fc0f326f/output.webp b/records/sd1/sweep/7db57587e704fc0f326f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0caa83ea38b63f910caf87ca2dd521451429d8ce
--- /dev/null
+++ b/records/sd1/sweep/7db57587e704fc0f326f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ddf1f5e38993f2eeb68141bded6879afa8250dfd37b56dd12d9d4f89a1ec7e6
+size 56462
diff --git a/records/sd1/sweep/7dbdf7d10c165a390ee4/output.webp b/records/sd1/sweep/7dbdf7d10c165a390ee4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d9b0f028fb27c3cb1713b8a936a18503b40a81c2
--- /dev/null
+++ b/records/sd1/sweep/7dbdf7d10c165a390ee4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcef179eba2527bad18dc78786e1f8ff3ab7fff1768dbe2bb7e5c43b1b57a0a0
+size 27146
diff --git a/records/sd1/sweep/7ddccd2d9cf9669a3504/output.webp b/records/sd1/sweep/7ddccd2d9cf9669a3504/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..48f44c579e68c6c05a5679abf7e6771752554040
--- /dev/null
+++ b/records/sd1/sweep/7ddccd2d9cf9669a3504/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ab47447f2ebea2434b995e8f7a99490ef8614a3b802007cb1642d0338c8d8664
+size 56204
diff --git a/records/sd1/sweep/7de3965b7e7209cc6db7/output.webp b/records/sd1/sweep/7de3965b7e7209cc6db7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3ac0073d7a7edab354cea5ade3fdaa031b58e7b3
--- /dev/null
+++ b/records/sd1/sweep/7de3965b7e7209cc6db7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccbf83b843e76734a2eecb275cc0aa383676a828b05ea47c5120da39982adb26
+size 30104
diff --git a/records/sd1/sweep/7e00fbcb152cd8f83f03/output.webp b/records/sd1/sweep/7e00fbcb152cd8f83f03/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..637e2a3ea70d4a113b3764d26773b05c7fc83d66
--- /dev/null
+++ b/records/sd1/sweep/7e00fbcb152cd8f83f03/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55698c13aec4667f8de34eaf51d1713635d30434cabdc5b32a172e08f7588d65
+size 15528
diff --git a/records/sd1/sweep/7e01ee265c64d279cc10/output.webp b/records/sd1/sweep/7e01ee265c64d279cc10/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c0ddc8dab5f9f02ff6d74c9a46090f29b25245bd
--- /dev/null
+++ b/records/sd1/sweep/7e01ee265c64d279cc10/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:451c3af7649c442d78d35630f8c36b39fd919dd23594ef7122f1f2743a9f163d
+size 14832
diff --git a/records/sd1/sweep/7e0a091b7f62bd12fa79/output.webp b/records/sd1/sweep/7e0a091b7f62bd12fa79/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..602522f3dc7379c1f9479e7dcf778037155053d5
--- /dev/null
+++ b/records/sd1/sweep/7e0a091b7f62bd12fa79/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5462ada0b124ac393c08748b0d4e7241c9c13d93bd2ea4e67391f62ef45272c6
+size 53638
diff --git a/records/sd1/sweep/7e0db55c072c70aab728/output.webp b/records/sd1/sweep/7e0db55c072c70aab728/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..da79245b31b92ef61afd3050811332fc19406e75
--- /dev/null
+++ b/records/sd1/sweep/7e0db55c072c70aab728/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ac8a00c1e4f6160c8a57531e0ace77b06f93bf425f6a249add1f247d8259b3a
+size 41454
diff --git a/records/sd1/sweep/7e18231d02e259fa898d/output.webp b/records/sd1/sweep/7e18231d02e259fa898d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b1d685c3f7a75f7a70188321f252aa0b264f46a9
--- /dev/null
+++ b/records/sd1/sweep/7e18231d02e259fa898d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80ca4d72d13d5007189923b3ed16eb9d67610867c45418fb2fa36e3b540dbba3
+size 29340
diff --git a/records/sd1/sweep/7e1b216018fcdfb43bb9/output.webp b/records/sd1/sweep/7e1b216018fcdfb43bb9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..911806d0438371a758673db2d882e43b870163ed
--- /dev/null
+++ b/records/sd1/sweep/7e1b216018fcdfb43bb9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8619b51615b7b1b57c810ba8aa90f4782d0d0f6dd98e22f078de495f17248698
+size 29786
diff --git a/records/sd1/sweep/7e24fc4ca575f8f33a58/output.webp b/records/sd1/sweep/7e24fc4ca575f8f33a58/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0f211cc539f06a1110e1f95b634b44a6de2bdbd9
--- /dev/null
+++ b/records/sd1/sweep/7e24fc4ca575f8f33a58/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6099fbabb11b3cebe3d46404a49198c05b569e4d36ca166aeb8b73093b20559c
+size 19104
diff --git a/records/sd1/sweep/7e3241fe8077c9019994/output.webp b/records/sd1/sweep/7e3241fe8077c9019994/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..01e5e2d32112be1cb5cbc2dec440305519a6e1c8
--- /dev/null
+++ b/records/sd1/sweep/7e3241fe8077c9019994/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:68a4391b74944c92df6d8ff10726616613863ec5a2dc9fd66cdc21ae9feb06f7
+size 22530
diff --git a/records/sd1/sweep/7e36b3d8e416f5034587/output.webp b/records/sd1/sweep/7e36b3d8e416f5034587/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60b7544afe6eb824c2ca5394509b58bc3ec91e92
--- /dev/null
+++ b/records/sd1/sweep/7e36b3d8e416f5034587/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa32ab586464256ad0de55ef4daf65b873ad9c92fe426231085f00d7f8da0ad5
+size 25728
diff --git a/records/sd1/sweep/7e3f97db95a281fc9560/output.webp b/records/sd1/sweep/7e3f97db95a281fc9560/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f541d5d6f99052a2da6989307851bd1c6e455461
--- /dev/null
+++ b/records/sd1/sweep/7e3f97db95a281fc9560/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a656ea951030cdf283011b2d81df0921badd299ba7e70cfc2c19a058d719d86e
+size 43428
diff --git a/records/sd1/sweep/7e4cf224b4e920757668/output.webp b/records/sd1/sweep/7e4cf224b4e920757668/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..978e7171228ec59d7b955e8bc4684b8d995bf5f2
--- /dev/null
+++ b/records/sd1/sweep/7e4cf224b4e920757668/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ad7cbd35796eb4a6309ded220297c12fc4fae8a63bab0ed57aea0954997d537
+size 34054
diff --git a/records/sd1/sweep/7e52d673c863008c65ab/output.webp b/records/sd1/sweep/7e52d673c863008c65ab/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..542e840b86e0ac91854b145e43851391087ef31c
--- /dev/null
+++ b/records/sd1/sweep/7e52d673c863008c65ab/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4973ba26dd267b751a84d52f3b121c869d71e06a5e3b65fc79d64657e8a6916
+size 17226
diff --git a/records/sd1/sweep/7e5d717d588ce6891081/output.webp b/records/sd1/sweep/7e5d717d588ce6891081/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..18f0e2d1b5859a3685cf8116f9794035c7b8ca90
--- /dev/null
+++ b/records/sd1/sweep/7e5d717d588ce6891081/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed9bbe0fbbf0dde780f28710791e4748a3322b54b4832dd35d66a25a434cdb37
+size 48958
diff --git a/records/sd1/sweep/7e74685074a69e7d8af6/output.webp b/records/sd1/sweep/7e74685074a69e7d8af6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..082870d83393425991260ccf5707fc87a3bfadcf
--- /dev/null
+++ b/records/sd1/sweep/7e74685074a69e7d8af6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80d9278e189f9fe688d174d4026d65c02ad880fcf2aee2a18ef55ae89f50b178
+size 20996
diff --git a/records/sd1/sweep/7e7612f5506f79310d85/output.webp b/records/sd1/sweep/7e7612f5506f79310d85/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ed6edb84df742f043ea1fe912d4d58b226b29c28
--- /dev/null
+++ b/records/sd1/sweep/7e7612f5506f79310d85/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00dae02724ef490de5b8ab017e04b97f3f633167880e63dd3ff528f06fa6d5aa
+size 85274
diff --git a/records/sd1/sweep/7e79b717db3295db6e29/output.webp b/records/sd1/sweep/7e79b717db3295db6e29/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1a9e67614067e9bc282d0f404af1fda9caaca2d2
--- /dev/null
+++ b/records/sd1/sweep/7e79b717db3295db6e29/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:475093084f1bce56249d3ecce63248b13862a8f098590365b12994acb7b4154f
+size 31820
diff --git a/records/sd1/sweep/7e8b66f0be89c6db75cb/output.webp b/records/sd1/sweep/7e8b66f0be89c6db75cb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ff39521341783779b796c0b54cdccc4dd29527f1
--- /dev/null
+++ b/records/sd1/sweep/7e8b66f0be89c6db75cb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:87525abc926a2da18001ea48f80aef567c7bf035439beaa892a8e47ec9fde8d2
+size 30862
diff --git a/records/sd1/sweep/7e8d66936084855443bf/output.webp b/records/sd1/sweep/7e8d66936084855443bf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..427ba9d309cf7641484065a0375d10b40e5ffcc9
--- /dev/null
+++ b/records/sd1/sweep/7e8d66936084855443bf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d255ce5590898fba0bade7f0d866505d02bab1cc64494516099deafba4866af1
+size 22990
diff --git a/records/sd1/sweep/7ea1666133f695431767/output.webp b/records/sd1/sweep/7ea1666133f695431767/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8b824289b7bd43bd6b21c6beefdac4f7b1e7145a
--- /dev/null
+++ b/records/sd1/sweep/7ea1666133f695431767/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3617d31455d30935ddd847544d21f3dde1c41bee5d5a348bdc6462be91155fe0
+size 27732
diff --git a/records/sd1/sweep/7ea23f245d6956372b73/output.webp b/records/sd1/sweep/7ea23f245d6956372b73/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..75a9ec0cd83ae3f713d516fda8e45538b9051676
--- /dev/null
+++ b/records/sd1/sweep/7ea23f245d6956372b73/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fbb7c81485e6b3c84a1074cf79ab65a0401fa033513d2143de74fe14a8483c7
+size 20352
diff --git a/records/sd1/sweep/7ea2fd061a6c3a5db796/output.webp b/records/sd1/sweep/7ea2fd061a6c3a5db796/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..39098909c60cb472ceb0b07fd0186ccafbba1220
--- /dev/null
+++ b/records/sd1/sweep/7ea2fd061a6c3a5db796/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6e75a13dc3c2943af959cca88d1655f212e3702300debf2f1ecfb1053f23e2b
+size 39444
diff --git a/records/sd1/sweep/7ea5be3165066b42ed38/output.webp b/records/sd1/sweep/7ea5be3165066b42ed38/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7890e3f0f1dd54f32345064d103b7844bce64089
--- /dev/null
+++ b/records/sd1/sweep/7ea5be3165066b42ed38/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:592b122f6a77673346ee886eddb2b8a1a9827eef4e6f5cfff8cfdadf5ea8813b
+size 39866
diff --git a/records/sd1/sweep/7ea89fbc3addb177c986/output.webp b/records/sd1/sweep/7ea89fbc3addb177c986/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3306fbdd4db45059711633f4229c4ce4620ef760
--- /dev/null
+++ b/records/sd1/sweep/7ea89fbc3addb177c986/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0022aa3535b86d762ca9d5a6cfa9cdb03ec9dbd7eabd2a8a58d8419d7dbb5db7
+size 27876
diff --git a/records/sd1/sweep/7eaed0a16b36be4dd967/output.webp b/records/sd1/sweep/7eaed0a16b36be4dd967/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1763670cf68489bf560cbdab387c74e6fdc3666a
--- /dev/null
+++ b/records/sd1/sweep/7eaed0a16b36be4dd967/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6d6824d6bb7954a8bfe0e147d0cd1dc9143b29394874dbdbd557961612e1a74
+size 45400
diff --git a/records/sd1/sweep/7eb63b60397c7ab83bfb/output.webp b/records/sd1/sweep/7eb63b60397c7ab83bfb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9b3e5fc16b22c3282294b695de75778946f305b8
--- /dev/null
+++ b/records/sd1/sweep/7eb63b60397c7ab83bfb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a876b3c61f38156c8fb25782e75c4714c2bd37ef11328a5a05cb41e5c2e1829
+size 37896
diff --git a/records/sd1/sweep/7ec139e99339fc45b80a/output.webp b/records/sd1/sweep/7ec139e99339fc45b80a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..46b0abfa92f05765dfbe0dd64eb41997f82d4ce9
--- /dev/null
+++ b/records/sd1/sweep/7ec139e99339fc45b80a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bdaa20d3231276f7f71a4b24f3628013faae59d356f83854f9ff49caa9f9b6b8
+size 33630
diff --git a/records/sd1/sweep/7ec3c8b01e4e97ae7021/output.webp b/records/sd1/sweep/7ec3c8b01e4e97ae7021/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60e1b1dbd5efcdb4a4ebd1a0781de295027b5eda
--- /dev/null
+++ b/records/sd1/sweep/7ec3c8b01e4e97ae7021/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8da9889f248c0338162c15d482f5529d49c210f05671b227a57796af3ecaaadc
+size 19716
diff --git a/records/sd1/sweep/7ed2bf26c1f13f69926b/output.webp b/records/sd1/sweep/7ed2bf26c1f13f69926b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fe7d22311b554257fb47b9cca0bbe44c9311f7d8
--- /dev/null
+++ b/records/sd1/sweep/7ed2bf26c1f13f69926b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9f9a94bf6d0f64b4f173080a71553ab1fc29383edf63e02f8af1be374611daf
+size 30416
diff --git a/records/sd1/sweep/7ee02fb0296eb50c5495/output.webp b/records/sd1/sweep/7ee02fb0296eb50c5495/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a2fcba53b5bbfa9f1ec0ab26b94aa983e9f6f904
--- /dev/null
+++ b/records/sd1/sweep/7ee02fb0296eb50c5495/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85d51a1072a73e7f267b846b3838a702db4caf5d73fa65c3a1c7ac57428d8201
+size 19934
diff --git a/records/sd1/sweep/7f02ca430bbec51e0457/output.webp b/records/sd1/sweep/7f02ca430bbec51e0457/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5d1b5e8d427ba3024fcf97217911b5d74939764a
--- /dev/null
+++ b/records/sd1/sweep/7f02ca430bbec51e0457/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:239964af5847f7516260c846731c9abed768ae70dbdd5cdd60b06d4a8edcd8a3
+size 27702
diff --git a/records/sd1/sweep/7f04f91220162b77b987/output.webp b/records/sd1/sweep/7f04f91220162b77b987/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..79532d534ed7c6b0ada71179dbaefac379448b55
--- /dev/null
+++ b/records/sd1/sweep/7f04f91220162b77b987/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:259db0dd2929fca9f06466c3ee905a3852a636ec0eabe717349d460b0610d603
+size 31302
diff --git a/records/sd1/sweep/7f0c6f60d6a460e1d7b4/output.webp b/records/sd1/sweep/7f0c6f60d6a460e1d7b4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7de8e1b607eef64ae701ab0ec392c351f12ad899
--- /dev/null
+++ b/records/sd1/sweep/7f0c6f60d6a460e1d7b4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:730b6692e18890c77428126494c3641a1f5859c8d070c29dfb10add15dd766de
+size 25876
diff --git a/records/sd1/sweep/7f12d9698136bec05749/output.webp b/records/sd1/sweep/7f12d9698136bec05749/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f90b4c311742589008ea90e6b83a8ccc86ba0e50
--- /dev/null
+++ b/records/sd1/sweep/7f12d9698136bec05749/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45e9e12263d7e83e155109fe2eaa487943b373971604985074702a359d9d89fd
+size 32610
diff --git a/records/sd1/sweep/7f13579e7cfe28984419/output.webp b/records/sd1/sweep/7f13579e7cfe28984419/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..daae2d87a983a4cc87123f84d13e2065fbba0b25
--- /dev/null
+++ b/records/sd1/sweep/7f13579e7cfe28984419/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c3e0402c88eac8d400d6e9a96056318173805cf54c07d991c2fad43e0aa5e42
+size 28454
diff --git a/records/sd1/sweep/7f21c98b890c05f771ee/output.webp b/records/sd1/sweep/7f21c98b890c05f771ee/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..344e3ab8ba72d57006cf1e89a51e8bd80e107da7
--- /dev/null
+++ b/records/sd1/sweep/7f21c98b890c05f771ee/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d8e47ad3b889a42f895f6d6718de27d5252d2af239aac5b1f4869dc6509e830
+size 17676
diff --git a/records/sd1/sweep/7f2f5affd2bcdfdf8975/output.webp b/records/sd1/sweep/7f2f5affd2bcdfdf8975/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b1f7401e9b1b9984123c5d11e424a9435e04212a
--- /dev/null
+++ b/records/sd1/sweep/7f2f5affd2bcdfdf8975/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:23cdcd37f091329a3b7dbd7e3be4eda1a42d6f6a4b2bd5726f413ecb88b29da8
+size 36128
diff --git a/records/sd1/sweep/7f394f12d77331dd58af/output.webp b/records/sd1/sweep/7f394f12d77331dd58af/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9fa60ac4d6ead07ae511132e2126fb0b1f19909f
--- /dev/null
+++ b/records/sd1/sweep/7f394f12d77331dd58af/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:97bf3d4d4bb3cea2bfe8e3abf6d8137bdc6be4b60fa3d7eff2a6a1640a77b8c7
+size 51150
diff --git a/records/sd1/sweep/7f5d8480a290c857872b/output.webp b/records/sd1/sweep/7f5d8480a290c857872b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7b5658de3ae10c776ea98d6cd27c4deaa8e13591
--- /dev/null
+++ b/records/sd1/sweep/7f5d8480a290c857872b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:784e3d097be5b7d8d81d1d69b8e8ece0737ac5d88304848056480a7033017d06
+size 19934
diff --git a/records/sd1/sweep/7f650fe77e2aebcd03ca/output.webp b/records/sd1/sweep/7f650fe77e2aebcd03ca/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dc3c4a468f19bfd6d864071d35957b782f9d601f
--- /dev/null
+++ b/records/sd1/sweep/7f650fe77e2aebcd03ca/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e0148934a33d5fcafe47b979c9cb1ed3c82543860f90b242f023ae245b45df96
+size 49316
diff --git a/records/sd1/sweep/7f8354fe62aa63ff5022/output.webp b/records/sd1/sweep/7f8354fe62aa63ff5022/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e644c49048fee81799b26316452941adf400f909
--- /dev/null
+++ b/records/sd1/sweep/7f8354fe62aa63ff5022/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0aa71e01f782be9581bafae9a96eaebcbf306da25bf29b6f61e28cb385a0c512
+size 23728
diff --git a/records/sd1/sweep/7f924795341cf28738b8/output.webp b/records/sd1/sweep/7f924795341cf28738b8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5930a632c52cfca5b82c0a4f84a12013d6446f6a
--- /dev/null
+++ b/records/sd1/sweep/7f924795341cf28738b8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d7b7dd312c947009152ad3d9ad482c19948acc817453ad688f2cb05aba15e23
+size 19068
diff --git a/records/sd1/sweep/7f98a35f26240ea35f0a/output.webp b/records/sd1/sweep/7f98a35f26240ea35f0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..84a502e746a29a56d7f159d0b01987bdf5fce02b
--- /dev/null
+++ b/records/sd1/sweep/7f98a35f26240ea35f0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15f42415d096e5b0ab998c80003f74750aafeec436aae4b371423c9b3d3f7201
+size 31718
diff --git a/records/sd1/sweep/7fa01df87e9b6dbae93e/output.webp b/records/sd1/sweep/7fa01df87e9b6dbae93e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7a2523996102974c3b0edacf0befea93890e84fa
--- /dev/null
+++ b/records/sd1/sweep/7fa01df87e9b6dbae93e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:764e83c7ae6c950e6a869052149d63213844d578945483eaf8f729582dabe038
+size 11452
diff --git a/records/sd1/sweep/7fa04696b8e70c0b377b/output.webp b/records/sd1/sweep/7fa04696b8e70c0b377b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c4a87265f9ed77dc1b1a61c18d2d478eb86131dc
--- /dev/null
+++ b/records/sd1/sweep/7fa04696b8e70c0b377b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74ac780c7f2b4d0b374de60009fba0a9f440aac9fdd73e282dcc1c87d42669ad
+size 31284
diff --git a/records/sd1/sweep/7fa0e82c173a0b539f86/output.webp b/records/sd1/sweep/7fa0e82c173a0b539f86/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..31279a25fcba70d5499ec7b6bb8cc1677f35cb9a
--- /dev/null
+++ b/records/sd1/sweep/7fa0e82c173a0b539f86/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a39ac63b2d5f707c715121f5386f706cc3e3c9e4dc1c7a5bfc05e53c150f7b9
+size 29586
diff --git a/records/sd1/sweep/7facf50dfcaf1dac024b/output.webp b/records/sd1/sweep/7facf50dfcaf1dac024b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..55d6cef54f91d639b8d9de338908382fef69b660
--- /dev/null
+++ b/records/sd1/sweep/7facf50dfcaf1dac024b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:420d2dfcad005cfb4b2515428fc41807582307a2a1d2cd603c791f8a68da94f7
+size 19934
diff --git a/records/sd1/sweep/7fadacf96269ed6a4de2/output.webp b/records/sd1/sweep/7fadacf96269ed6a4de2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d4335feb1dbfd6b6c2f7cafa96cb0d9568021973
--- /dev/null
+++ b/records/sd1/sweep/7fadacf96269ed6a4de2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a6150285552e6a88fa826d593de5195abea4e01061b7a498d2ce1b8dc73d7506
+size 15748
diff --git a/records/sd1/sweep/7fadcfd8b55055ec0a84/output.webp b/records/sd1/sweep/7fadcfd8b55055ec0a84/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7c3022f6a5e9f693d30299931c488e8b61f75adb
--- /dev/null
+++ b/records/sd1/sweep/7fadcfd8b55055ec0a84/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86fe5c791cd40162de6bef9b232fadaa39b6ff3fc948a51a6c216d1f6041e7dd
+size 45384
diff --git a/records/sd1/sweep/7fae7d297a269b12be6a/output.webp b/records/sd1/sweep/7fae7d297a269b12be6a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..715b75e62c9e979bf70b2ab1c7d476155e3c8668
--- /dev/null
+++ b/records/sd1/sweep/7fae7d297a269b12be6a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2226ec47609032a18c0b9bc6b4f11c1e4757a1476a5648768db08e0fbf06872
+size 34512
diff --git a/records/sd1/sweep/7fc1b04eede7a8e168e6/output.webp b/records/sd1/sweep/7fc1b04eede7a8e168e6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bdc1589025c5b30ca922f23af90fd83e260304b2
--- /dev/null
+++ b/records/sd1/sweep/7fc1b04eede7a8e168e6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:189fdc151b738589af98ff496cf090a414b20cd0fa34e934c8fe25637dbf6707
+size 30620
diff --git a/records/sd1/sweep/7fc42f5644ba3714ca9b/output.webp b/records/sd1/sweep/7fc42f5644ba3714ca9b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..844bb294bf50a98d653b59e6bee81dac126ee2b0
--- /dev/null
+++ b/records/sd1/sweep/7fc42f5644ba3714ca9b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eccfdac6c6ae8291cc21e3dc5021b2ab58f5726b8f897a38f294111fa2432aa
+size 11780
diff --git a/records/sd1/sweep/7fca89ef25df8f52d9f9/output.webp b/records/sd1/sweep/7fca89ef25df8f52d9f9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e7a4d44d5d49f6466bed18b755747bbb94691024
--- /dev/null
+++ b/records/sd1/sweep/7fca89ef25df8f52d9f9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe61f437d520963b0ebfbfe8e90f2a95addb88ece2daeea2ecd60e967dd7d7d4
+size 35620
diff --git a/records/sd1/sweep/7fced1b89ecc2b4d94a9/output.webp b/records/sd1/sweep/7fced1b89ecc2b4d94a9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fceec2eba0de4b23026da0bcf4f11c0e19dccb4f
--- /dev/null
+++ b/records/sd1/sweep/7fced1b89ecc2b4d94a9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94660013058ac0b4327302a13596cb63e26148d5b651f271332d2497bf18f30f
+size 62542
diff --git a/records/sd1/sweep/7fd62a662e8389b54ef8/output.webp b/records/sd1/sweep/7fd62a662e8389b54ef8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e1b0885eeb3b3d4e6757d35887424bf9d9d88fb0
--- /dev/null
+++ b/records/sd1/sweep/7fd62a662e8389b54ef8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f306242bed8c25046618a20530ef5da2dd55399a40e4072ac339fd71c2cb761
+size 17892
diff --git a/records/sd1/sweep/7fda350c7b004a4f9c2d/output.webp b/records/sd1/sweep/7fda350c7b004a4f9c2d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..eee5189382ecf519fbc75272b9fa8735a875bd03
--- /dev/null
+++ b/records/sd1/sweep/7fda350c7b004a4f9c2d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfaff17c1289479009682dbdc6246e3cf5cacf63cde0e22eaa9faada6cf6272f
+size 57970
diff --git a/records/sd1/sweep/7fea695f1df505386907/output.webp b/records/sd1/sweep/7fea695f1df505386907/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..86e1800567d20f449e06bf51d0813396ea5e0de8
--- /dev/null
+++ b/records/sd1/sweep/7fea695f1df505386907/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f46fd35d9099c5a2d709da88767890b14911b2287b24d595c7a53e785b40f2eb
+size 21212
diff --git a/records/sd1/sweep/87aafd7ffc689a9c88ee/measurements.json b/records/sd1/sweep/87aafd7ffc689a9c88ee/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b2b3666fb4d5a747e287ac99e19de3250d78266a
--- /dev/null
+++ b/records/sd1/sweep/87aafd7ffc689a9c88ee/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87aafd7ffc689a9c88ee",
+ "values": {
+  "std": {
+   "value": 49.116,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0149,
+   "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": 9.744,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.075772,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.140501,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.078067,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.031098,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9897,
+   "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": 0.0279,
+   "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/87aafd7ffc689a9c88ee/workflow.json b/records/sd1/sweep/87aafd7ffc689a9c88ee/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c50a36976fa6358578cae2fa044b39fd2f9113ce
--- /dev/null
+++ b/records/sd1/sweep/87aafd7ffc689a9c88ee/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 87aafd7ffc689a9c88ee)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\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 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/87ba95c1dbc3cd4068d2/interpretations.jsonl b/records/sd1/sweep/87ba95c1dbc3cd4068d2/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..40598cbf1983ae1d94a1afc755e235f490e7f877
--- /dev/null
+++ b/records/sd1/sweep/87ba95c1dbc3cd4068d2/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "87ba95c1dbc3cd4068d2"}, "kind": "caption", "value": "Studio portrait of bob-haired woman in black top against grey card, nearly unchanged", "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": "c8cb1e471add1183"}
+{"target": {"record": "87ba95c1dbc3cd4068d2"}, "kind": "change", "value": "Almost no visible change, just slightly redder lips and minor tonal warmth.", "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": "471974928bf22e00"}
+{"target": {"record": "87ba95c1dbc3cd4068d2"}, "kind": "keywords", "value": ["near identical", "studio portrait", "minor tone shift", "red lips"], "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": "d6ad78213f06d9e5"}
+{"target": {"record": "87ba95c1dbc3cd4068d2"}, "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": "2026-10-05T08:17:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "aae562f1c2d12dcf"}
diff --git a/records/sd1/sweep/87ba95c1dbc3cd4068d2/measurements.json b/records/sd1/sweep/87ba95c1dbc3cd4068d2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..68cbf0836c8eb77ae574d0588c4a921f9b099e0c
--- /dev/null
+++ b/records/sd1/sweep/87ba95c1dbc3cd4068d2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87ba95c1dbc3cd4068d2",
+ "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/87ba95c1dbc3cd4068d2/record.json b/records/sd1/sweep/87ba95c1dbc3cd4068d2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..89022eac1aa9519d783e2d370220c6cb63c17f6e
--- /dev/null
+++ b/records/sd1/sweep/87ba95c1dbc3cd4068d2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "87ba95c1dbc3cd4068d2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:17Z",
+ "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.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.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/c130cc12cf4d5fa2a98ccdada3984989b69a6deb19f5aacc693f37be925bbc79.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c130cc12cf4d5fa2a98ccdada3984989b69a6deb19f5aacc693f37be925bbc79",
+  "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": "03c75882e0718c4a0161"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87ba95c1dbc3cd4068d2/workflow.json b/records/sd1/sweep/87ba95c1dbc3cd4068d2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d2f0ffced094e2c24ff01d74a513eb55a0cb5cc
--- /dev/null
+++ b/records/sd1/sweep/87ba95c1dbc3cd4068d2/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 87ba95c1dbc3cd4068d2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.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/87c476c4ad3771f15093/measurements.json b/records/sd1/sweep/87c476c4ad3771f15093/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff2abc5e27deca313bd8a2c5d06efccaf4bbe758
--- /dev/null
+++ b/records/sd1/sweep/87c476c4ad3771f15093/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87c476c4ad3771f15093",
+ "values": {
+  "std": {
+   "value": 44.272,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01684,
+   "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": 0.592,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000103,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.000792,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000507,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000141,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.974,
+   "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": 0.069,
+   "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/87c476c4ad3771f15093/record.json b/records/sd1/sweep/87c476c4ad3771f15093/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f57f340edbf256810f104dc19d5b5ad203a2dcd8
--- /dev/null
+++ b/records/sd1/sweep/87c476c4ad3771f15093/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "87c476c4ad3771f15093",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:54:20Z",
+ "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": 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_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "heavy",
+   "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": [
+   "mid",
+   "add_noise",
+   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": "899a7c5b377d673804fb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87c476c4ad3771f15093/workflow.json b/records/sd1/sweep/87c476c4ad3771f15093/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..74272aec6c1d3726d193fab241feb755dc4b7597
--- /dev/null
+++ b/records/sd1/sweep/87c476c4ad3771f15093/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 87c476c4ad3771f15093)"
+  },
+  "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\": 0.8,\n    \"seed\": 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/87c8d12c8af2690e85e6/measurements.json b/records/sd1/sweep/87c8d12c8af2690e85e6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4c686dc6cef7c539f163fc68e6c90cddb290236
--- /dev/null
+++ b/records/sd1/sweep/87c8d12c8af2690e85e6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87c8d12c8af2690e85e6",
+ "values": {
+  "std": {
+   "value": 39.093,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0098,
+   "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": 31.774,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.385861,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.510618,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.362558,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.093034,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0726,
+   "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"
+  },
+  "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/87c8d12c8af2690e85e6/record.json b/records/sd1/sweep/87c8d12c8af2690e85e6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0b7c5ca078d123190b066c17d3863a08cde4737
--- /dev/null
+++ b/records/sd1/sweep/87c8d12c8af2690e85e6/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "87c8d12c8af2690e85e6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:36:41Z",
+ "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 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": 0.0,
+   "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": [
+   "in.lo",
+   "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": "7ad99ad5333ac9095860"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87c8d12c8af2690e85e6/workflow.json b/records/sd1/sweep/87c8d12c8af2690e85e6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2edb2e4503c3e5e45e26fa2aa10ff1c1482a051e
--- /dev/null
+++ b/records/sd1/sweep/87c8d12c8af2690e85e6/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 87c8d12c8af2690e85e6)"
+  },
+  "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}",
+   "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/87dd583cbb1c1b4ade62/measurements.json b/records/sd1/sweep/87dd583cbb1c1b4ade62/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d68d63387ae19d5e842c0f623cfcb605f0a74ee
--- /dev/null
+++ b/records/sd1/sweep/87dd583cbb1c1b4ade62/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87dd583cbb1c1b4ade62",
+ "values": {
+  "std": {
+   "value": 23.109,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00682,
+   "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": 42.874,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.473137,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.737984,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.877562,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.360418,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7431,
+   "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": 0.6557,
+   "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/87dd583cbb1c1b4ade62/record.json b/records/sd1/sweep/87dd583cbb1c1b4ade62/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bcf3a299de455698ad6373e283107ce50a7f42d
--- /dev/null
+++ b/records/sd1/sweep/87dd583cbb1c1b4ade62/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "87dd583cbb1c1b4ade62",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:46:09Z",
+ "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": 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": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "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",
+   "scale",
+   0.7,
+   "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": "84a67197c5e3fa26e14b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87dd583cbb1c1b4ade62/workflow.json b/records/sd1/sweep/87dd583cbb1c1b4ade62/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c13f7d786399341a13e4e5d315887efc239808a
--- /dev/null
+++ b/records/sd1/sweep/87dd583cbb1c1b4ade62/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 87dd583cbb1c1b4ade62)"
+  },
+  "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}",
+   "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/87e0b9b826b3213efbc0/interpretations.jsonl b/records/sd1/sweep/87e0b9b826b3213efbc0/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..16a711d888574c1c388ef35a8e119800bf85654c
--- /dev/null
+++ b/records/sd1/sweep/87e0b9b826b3213efbc0/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "87e0b9b826b3213efbc0"}, "kind": "caption", "value": "Dark glitchy portrait with a ghostly face dissolving into blue-grey glittering static", "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": "fb2d837a0875d2b5"}
+{"target": {"record": "87e0b9b826b3213efbc0"}, "kind": "change", "value": "The painting dissolves into dark blue-grey noisy static, leaving only a faint, distorted face and hair outline made of glittering fragments.", "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": "05f5035073025b53"}
+{"target": {"record": "87e0b9b826b3213efbc0"}, "kind": "keywords", "value": ["glitch", "static", "ghostly face", "dark noise", "degraded portrait"], "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": "81c07a0995267c0e"}
+{"target": {"record": "87e0b9b826b3213efbc0"}, "kind": "effect_tags", "value": ["glitch", "texture", "dark", "fragmented", "distorted", "degenerate"], "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": "c301cfecab564562"}
diff --git a/records/sd1/sweep/87e0b9b826b3213efbc0/measurements.json b/records/sd1/sweep/87e0b9b826b3213efbc0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebb880b93e57841570545b52587e500f03911f19
--- /dev/null
+++ b/records/sd1/sweep/87e0b9b826b3213efbc0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87e0b9b826b3213efbc0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.522132,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.722256,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.458868,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.218041,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.432,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02654,
+   "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": 59.151,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9997,
+   "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": 0.849,
+   "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/87e0b9b826b3213efbc0/record.json b/records/sd1/sweep/87e0b9b826b3213efbc0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..47a2b3b65cb71780decdcc39338bc3b204237585
--- /dev/null
+++ b/records/sd1/sweep/87e0b9b826b3213efbc0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "87e0b9b826b3213efbc0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:36Z",
+ "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": "input_blocks.1.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/a0af04b354ca631ad61265743122a33e1cfdf070e5f232f1a84618a9b3890142.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a0af04b354ca631ad61265743122a33e1cfdf070e5f232f1a84618a9b3890142",
+  "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": "43aa374ed0855f3dbb7a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87e0b9b826b3213efbc0/workflow.json b/records/sd1/sweep/87e0b9b826b3213efbc0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e27be816a769322fcb533a225eba18cca63679c8
--- /dev/null
+++ b/records/sd1/sweep/87e0b9b826b3213efbc0/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 87e0b9b826b3213efbc0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.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/87e87a054f49f79d62de/measurements.json b/records/sd1/sweep/87e87a054f49f79d62de/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b6054258f954b58bcacfc8d17c7c1e4ed28649f
--- /dev/null
+++ b/records/sd1/sweep/87e87a054f49f79d62de/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87e87a054f49f79d62de",
+ "values": {
+  "std": {
+   "value": 58.897,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0099,
+   "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": 1.619,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.002017,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.004855,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.006906,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000833,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0048,
+   "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": 0.3244,
+   "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/87e87a054f49f79d62de/record.json b/records/sd1/sweep/87e87a054f49f79d62de/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8025037b8f4162b7ee4214c12f50783a72d486f4
--- /dev/null
+++ b/records/sd1/sweep/87e87a054f49f79d62de/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "87e87a054f49f79d62de",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:52:51Z",
+ "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": 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.6.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "out.mid",
+   "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": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_noise",
+   0.2,
+   "middle"
+  ],
+  "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"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87e87a054f49f79d62de/workflow.json b/records/sd1/sweep/87e87a054f49f79d62de/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..81604481e3d2ce26b5820eb3c1cd8a34e55595cd
--- /dev/null
+++ b/records/sd1/sweep/87e87a054f49f79d62de/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 87e87a054f49f79d62de)"
+  },
+  "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\": 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": 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/87ec0909a3754a86d14b/interpretations.jsonl b/records/sd1/sweep/87ec0909a3754a86d14b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..dd5de50fc3fa8b7e1af6f2d283a6751f3beb316a
--- /dev/null
+++ b/records/sd1/sweep/87ec0909a3754a86d14b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "87ec0909a3754a86d14b"}, "kind": "caption", "value": "Wrinkled old man with white beard and tweed 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": "2026-10-05T04:48:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5717ab5c34c76815"}
+{"target": {"record": "87ec0909a3754a86d14b"}, "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": "2026-10-05T04:48:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1931206a9f1f2e4f"}
+{"target": {"record": "87ec0909a3754a86d14b"}, "kind": "keywords", "value": ["old man portrait", "tweed hat", "white beard", "no change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:48:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "dee5938d34407302"}
+{"target": {"record": "87ec0909a3754a86d14b"}, "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": "2026-10-05T04:48:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e808802e5ba6eba8"}
diff --git a/records/sd1/sweep/87ec0909a3754a86d14b/measurements.json b/records/sd1/sweep/87ec0909a3754a86d14b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..adf59df59716074e4a26c10e3e3ae4c6c4e770b3
--- /dev/null
+++ b/records/sd1/sweep/87ec0909a3754a86d14b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87ec0909a3754a86d14b",
+ "values": {
+  "std": {
+   "value": 58.861,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00948,
+   "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": 4.394,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.007547,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.016553,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.003653,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002101,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0082,
+   "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": 0.4338,
+   "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/87ec0909a3754a86d14b/record.json b/records/sd1/sweep/87ec0909a3754a86d14b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e13a58d7f6d0c510349ff55fb98a96fe3fbed5e
--- /dev/null
+++ b/records/sd1/sweep/87ec0909a3754a86d14b/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "87ec0909a3754a86d14b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:41:01Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "mid",
+   "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": [
+   "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": "9d30f2adff4cf7ac4294"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87ec0909a3754a86d14b/workflow.json b/records/sd1/sweep/87ec0909a3754a86d14b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1be33438d8e5b618affc19dd566927c9b88dce9a
--- /dev/null
+++ b/records/sd1/sweep/87ec0909a3754a86d14b/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 87ec0909a3754a86d14b)"
+  },
+  "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 \"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/87f4d359e69356161afa/measurements.json b/records/sd1/sweep/87f4d359e69356161afa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..12cd067748b54d1c7ae9136d3644ba79914c0e3c
--- /dev/null
+++ b/records/sd1/sweep/87f4d359e69356161afa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87f4d359e69356161afa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.437692,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.466692,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.391142,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.249145,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.999,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0091,
+   "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": 48.119,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9429,
+   "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": 0.5133,
+   "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/87f4d359e69356161afa/record.json b/records/sd1/sweep/87f4d359e69356161afa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a557372d363feb1a13c5cbf5620557d3de34c1d7
--- /dev/null
+++ b/records/sd1/sweep/87f4d359e69356161afa/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "87f4d359e69356161afa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:32Z",
+ "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": "input_blocks.2.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.hi",
+   "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/ad9ccf150f8ff7b47506f53470e9eaee5a3f6f898b69a108df1b204c9423ea00.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ad9ccf150f8ff7b47506f53470e9eaee5a3f6f898b69a108df1b204c9423ea00",
+  "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": "5d3e1622221359d7628c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87f4d359e69356161afa/workflow.json b/records/sd1/sweep/87f4d359e69356161afa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9df093511faffe5c9dddc042f9f274279f83fab
--- /dev/null
+++ b/records/sd1/sweep/87f4d359e69356161afa/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 87f4d359e69356161afa)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.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/88007c6be83ff3626d9f/measurements.json b/records/sd1/sweep/88007c6be83ff3626d9f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..935c972e379cc0e484add31342197ec7135de2f3
--- /dev/null
+++ b/records/sd1/sweep/88007c6be83ff3626d9f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88007c6be83ff3626d9f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.123114,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.140961,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.203042,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.162382,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.33,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00399,
+   "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": 15.001,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9772,
+   "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": 0.6208,
+   "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/88007c6be83ff3626d9f/record.json b/records/sd1/sweep/88007c6be83ff3626d9f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..96a8c6ff44e4bbecf7ce9544f2e1fb4639dae2c1
--- /dev/null
+++ b/records/sd1/sweep/88007c6be83ff3626d9f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "88007c6be83ff3626d9f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:38Z",
+ "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.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "mid",
+   "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/79f1c418e94f61b90c02d3c2c82a91d4122537392df90d4513e67e403573e3cf.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "79f1c418e94f61b90c02d3c2c82a91d4122537392df90d4513e67e403573e3cf",
+  "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"
+    ],
+    "previous_id": "34d19340c29ad14f79a7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88007c6be83ff3626d9f/workflow.json b/records/sd1/sweep/88007c6be83ff3626d9f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2226fd13dfd69e24dc55ea65793be267c1b2cc0
--- /dev/null
+++ b/records/sd1/sweep/88007c6be83ff3626d9f/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 88007c6be83ff3626d9f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.out_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": [
+    "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/88038de93a06524e90e7/measurements.json b/records/sd1/sweep/88038de93a06524e90e7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..01d7e694c2233ce4b8606b8033981d77b388aea6
--- /dev/null
+++ b/records/sd1/sweep/88038de93a06524e90e7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88038de93a06524e90e7",
+ "values": {
+  "std": {
+   "value": 50.917,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01048,
+   "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": 2.784,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.006315,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.022428,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.032087,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002131,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9898,
+   "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": 0.0145,
+   "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/88038de93a06524e90e7/record.json b/records/sd1/sweep/88038de93a06524e90e7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d39dfb012985c63a7344c8c4b772efca5a6dd6ad
--- /dev/null
+++ b/records/sd1/sweep/88038de93a06524e90e7/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "88038de93a06524e90e7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:41:12Z",
+ "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 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": 0.8,
+    "seed": 1234
+   },
+   "group": "mid",
+   "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": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "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/88038de93a06524e90e7/workflow.json b/records/sd1/sweep/88038de93a06524e90e7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e000044b88fccdea820d1e8ace91f869b758275e
--- /dev/null
+++ b/records/sd1/sweep/88038de93a06524e90e7/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 88038de93a06524e90e7)"
+  },
+  "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\": 0.8,\n    \"seed\": 1234\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/880ad3e45d1d36722c1d/measurements.json b/records/sd1/sweep/880ad3e45d1d36722c1d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e92abbf852a3537f46089f1b22d599d4aec93f6
--- /dev/null
+++ b/records/sd1/sweep/880ad3e45d1d36722c1d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "880ad3e45d1d36722c1d",
+ "values": {
+  "std": {
+   "value": 35.528,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02211,
+   "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": 29.025,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.341033,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.477261,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.150093,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.063951,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0014,
+   "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": 0.0681,
+   "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/880ad3e45d1d36722c1d/record.json b/records/sd1/sweep/880ad3e45d1d36722c1d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..27430195986cdf6efb441ac43d6b23918cbc4a76
--- /dev/null
+++ b/records/sd1/sweep/880ad3e45d1d36722c1d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "880ad3e45d1d36722c1d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:09:13Z",
+ "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": 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": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.lo",
+   "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": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "rotate",
+   30.0,
+   "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": "21a1f6da4329cfaf98d3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/880ad3e45d1d36722c1d/workflow.json b/records/sd1/sweep/880ad3e45d1d36722c1d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a71c4a8823f22af51228aa097e973ed4d584154f
--- /dev/null
+++ b/records/sd1/sweep/880ad3e45d1d36722c1d/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 880ad3e45d1d36722c1d)"
+  },
+  "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\": 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": "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/880b439a3498fc5b2f09/measurements.json b/records/sd1/sweep/880b439a3498fc5b2f09/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef5ad0c508b8342a4cd3d5798d466729a0e1d53f
--- /dev/null
+++ b/records/sd1/sweep/880b439a3498fc5b2f09/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "880b439a3498fc5b2f09",
+ "values": {
+  "std": {
+   "value": 55.903,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01064,
+   "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": 17.147,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.108903,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.124282,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.09405,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02279,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9918,
+   "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": 0.2665,
+   "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/880b439a3498fc5b2f09/record.json b/records/sd1/sweep/880b439a3498fc5b2f09/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfeb816389a921b70879bf0521e58d707989000a
--- /dev/null
+++ b/records/sd1/sweep/880b439a3498fc5b2f09/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "880b439a3498fc5b2f09",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:37:18Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "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": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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/880b439a3498fc5b2f09/workflow.json b/records/sd1/sweep/880b439a3498fc5b2f09/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4be9f3ed60be59957f755c710250a5f32b1654f
--- /dev/null
+++ b/records/sd1/sweep/880b439a3498fc5b2f09/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 880b439a3498fc5b2f09)"
+  },
+  "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\": 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": 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/880e84db10b626e1657b/measurements.json b/records/sd1/sweep/880e84db10b626e1657b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9e1675687ccdd4e94cb499020f4c3c37f75aadd
--- /dev/null
+++ b/records/sd1/sweep/880e84db10b626e1657b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "880e84db10b626e1657b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.048429,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.072911,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038024,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.011807,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.874,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00605,
+   "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": 10.299,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0241,
+   "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": 0.1273,
+   "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/880e84db10b626e1657b/record.json b/records/sd1/sweep/880e84db10b626e1657b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4323d8217c04fb686bcd634278a39f751a921953
--- /dev/null
+++ b/records/sd1/sweep/880e84db10b626e1657b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "880e84db10b626e1657b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:34Z",
+ "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": 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.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "cross_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/248a0594b9036a6c1a8fbf6eedc3db2e8dd6e82be9b3951250dcca32fa4a5e24.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "248a0594b9036a6c1a8fbf6eedc3db2e8dd6e82be9b3951250dcca32fa4a5e24",
+  "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": "b1a196b8e3d5bef15d7f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/880e84db10b626e1657b/workflow.json b/records/sd1/sweep/880e84db10b626e1657b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1356a7665f1ea6c400f04a8af4947296bbfdc6a2
--- /dev/null
+++ b/records/sd1/sweep/880e84db10b626e1657b/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 880e84db10b626e1657b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.attn2.to_v\",\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/8811569b4d0343197749/measurements.json b/records/sd1/sweep/8811569b4d0343197749/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..acc044a01a8f19bba0ad11da0b0193d4193c3f1d
--- /dev/null
+++ b/records/sd1/sweep/8811569b4d0343197749/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8811569b4d0343197749",
+ "values": {
+  "cosine_distance": {
+   "value": 0.307697,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.340532,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.202164,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.178722,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.21,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00258,
+   "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": 34.391,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.015,
+   "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": 0.0424,
+   "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/8811569b4d0343197749/record.json b/records/sd1/sweep/8811569b4d0343197749/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..584d8be6b7dbcc0d9e07423f0bb0100f48ca7883
--- /dev/null
+++ b/records/sd1/sweep/8811569b4d0343197749/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8811569b4d0343197749",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:55Z",
+ "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": "output_blocks.3.1.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/0c6532d115166d840bc812b734adb84c5857c95b72227aa910181c4699d139d4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0c6532d115166d840bc812b734adb84c5857c95b72227aa910181c4699d139d4",
+  "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": "b3986fb68472caa8fe2e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8811569b4d0343197749/workflow.json b/records/sd1/sweep/8811569b4d0343197749/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..abd688dcae4d2f3f382f02a471ab094a0c8dd6d7
--- /dev/null
+++ b/records/sd1/sweep/8811569b4d0343197749/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 8811569b4d0343197749)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.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/881ec0a46c38dc167a13/measurements.json b/records/sd1/sweep/881ec0a46c38dc167a13/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd0c9b57ec3d712d6ad72cac8e61b5ff539ff1e9
--- /dev/null
+++ b/records/sd1/sweep/881ec0a46c38dc167a13/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "881ec0a46c38dc167a13",
+ "values": {
+  "cosine_distance": {
+   "value": 0.310687,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.298465,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.177752,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.1437,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.964,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00431,
+   "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": 33.314,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9535,
+   "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": 0.5514,
+   "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/881ec0a46c38dc167a13/record.json b/records/sd1/sweep/881ec0a46c38dc167a13/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d18feb3d2dd687e05bcb11b9a19ec92189ebd41
--- /dev/null
+++ b/records/sd1/sweep/881ec0a46c38dc167a13/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "881ec0a46c38dc167a13",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:34Z",
+ "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": "output_blocks.6.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "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/5cf62241975914fa694fbfe98342a28b8f314cd24133690462f97a770b5be7a1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5cf62241975914fa694fbfe98342a28b8f314cd24133690462f97a770b5be7a1",
+  "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"
+    ],
+    "previous_id": "df1358172acfa7113f7c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/881ec0a46c38dc167a13/workflow.json b/records/sd1/sweep/881ec0a46c38dc167a13/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f83a8bba8c8ba525722d958c813b6e4874272d49
--- /dev/null
+++ b/records/sd1/sweep/881ec0a46c38dc167a13/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 881ec0a46c38dc167a13)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.transformer_blocks.0.attn2.to_k\",\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/8825310a85bd8a4b056b/measurements.json b/records/sd1/sweep/8825310a85bd8a4b056b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd324601c2f127958c8e25458cc8db78a62b89a0
--- /dev/null
+++ b/records/sd1/sweep/8825310a85bd8a4b056b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8825310a85bd8a4b056b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.09775,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.162261,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.191637,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.110414,
+   "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": 0.00324,
+   "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": 16.128,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0119,
+   "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/8825310a85bd8a4b056b/record.json b/records/sd1/sweep/8825310a85bd8a4b056b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9f7ac3b42218499084cc2990cb6bcae1455f941
--- /dev/null
+++ b/records/sd1/sweep/8825310a85bd8a4b056b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8825310a85bd8a4b056b",
+ "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.in_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.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/47afe0a04e98a003b258da4c3fd58d441688497c986285da7e1c9216f4c5fd00.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "47afe0a04e98a003b258da4c3fd58d441688497c986285da7e1c9216f4c5fd00",
+  "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": "c5c408fd55059134c642"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8825310a85bd8a4b056b/workflow.json b/records/sd1/sweep/8825310a85bd8a4b056b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d300de1d8acfba040948f2f54bfe43dff4293b37
--- /dev/null
+++ b/records/sd1/sweep/8825310a85bd8a4b056b/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 8825310a85bd8a4b056b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.in_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/88253955942818eb7d0d/measurements.json b/records/sd1/sweep/88253955942818eb7d0d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fdaf1254a218888e6e7cb38e0d9ee8f1160c886b
--- /dev/null
+++ b/records/sd1/sweep/88253955942818eb7d0d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88253955942818eb7d0d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.161361,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.327071,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.224538,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.096692,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.485,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00398,
+   "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": 21.308,
+   "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": 0.0744,
+   "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/88253955942818eb7d0d/record.json b/records/sd1/sweep/88253955942818eb7d0d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c0ed0273fc3b5401e3a55a99419e7feb4b39ced
--- /dev/null
+++ b/records/sd1/sweep/88253955942818eb7d0d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88253955942818eb7d0d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:40Z",
+ "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": 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.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ed850a08b13a19445a6cc8882242f3c4b3ff9ac6cbf50571cd831eacaecf00ad.png",
+  "batch_id": "kb_lcm_h2h_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",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6d0fc06d8e95fd8fb243"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88253955942818eb7d0d/workflow.json b/records/sd1/sweep/88253955942818eb7d0d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..48a45e389a70b08e63fe24b2b3590192372a177a
--- /dev/null
+++ b/records/sd1/sweep/88253955942818eb7d0d/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 88253955942818eb7d0d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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": [
+    "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/8826770dc2113bc2ae95/measurements.json b/records/sd1/sweep/8826770dc2113bc2ae95/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4ebc4a93fa1025a6ab0379c56ac9095a35f4f57
--- /dev/null
+++ b/records/sd1/sweep/8826770dc2113bc2ae95/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8826770dc2113bc2ae95",
+ "values": {
+  "cosine_distance": {
+   "value": 0.241022,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.297077,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.100384,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.049524,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.015,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00679,
+   "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": 30.426,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9214,
+   "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": 0.1616,
+   "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/8826770dc2113bc2ae95/record.json b/records/sd1/sweep/8826770dc2113bc2ae95/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..05098ac1539f56a0d4627b16c4c28d3562b34cb5
--- /dev/null
+++ b/records/sd1/sweep/8826770dc2113bc2ae95/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8826770dc2113bc2ae95",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:54Z",
+ "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.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/6880f912a5ed035c162980950ec959dbea2f768d282c1cfd7adc2f8ec2a133a4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6880f912a5ed035c162980950ec959dbea2f768d282c1cfd7adc2f8ec2a133a4",
+  "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": "65830a3e077bce1f488c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8826770dc2113bc2ae95/workflow.json b/records/sd1/sweep/8826770dc2113bc2ae95/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e416d0b520a87c42ec66288e4d6752dc7cb9c3cb
--- /dev/null
+++ b/records/sd1/sweep/8826770dc2113bc2ae95/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 8826770dc2113bc2ae95)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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/8831ada35eefdc90c9e0/measurements.json b/records/sd1/sweep/8831ada35eefdc90c9e0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb026c8adbd79ace3e28e47d83f68ba5bfa4e3b4
--- /dev/null
+++ b/records/sd1/sweep/8831ada35eefdc90c9e0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8831ada35eefdc90c9e0",
+ "values": {
+  "std": {
+   "value": 49.68,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00877,
+   "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": 10.392,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.061941,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.19714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.178402,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.072229,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0474,
+   "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": 0.0466,
+   "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/8831ada35eefdc90c9e0/record.json b/records/sd1/sweep/8831ada35eefdc90c9e0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4da929ea4921ac0e6e84ba7be196151b1ead6ff0
--- /dev/null
+++ b/records/sd1/sweep/8831ada35eefdc90c9e0/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8831ada35eefdc90c9e0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:42:39Z",
+ "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": "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": 30.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "slight",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "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",
+   "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": "9132a0fc45556634b6e4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8831ada35eefdc90c9e0/workflow.json b/records/sd1/sweep/8831ada35eefdc90c9e0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0b7cddcdb11d1e605be9da977e6026114f96119
--- /dev/null
+++ b/records/sd1/sweep/8831ada35eefdc90c9e0/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 8831ada35eefdc90c9e0)"
+  },
+  "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\": 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": 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/884224f6170f841d1877/measurements.json b/records/sd1/sweep/884224f6170f841d1877/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e20081f391572bb7c3dcbc0fb6969ee4c4b18b42
--- /dev/null
+++ b/records/sd1/sweep/884224f6170f841d1877/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "884224f6170f841d1877",
+ "values": {
+  "cosine_distance": {
+   "value": 0.042556,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.090165,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.031542,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02496,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.001,
+   "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": [],
+   "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": 14.02,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.0938,
+   "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/884224f6170f841d1877/record.json b/records/sd1/sweep/884224f6170f841d1877/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0697dc68b4669d6a12e169313d2e8e430a0094e8
--- /dev/null
+++ b/records/sd1/sweep/884224f6170f841d1877/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "884224f6170f841d1877",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:27Z",
+ "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": 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.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "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/0e98a427e79e7b8ace19da6185ea9429bc163fedb758891964d602997f338c54.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0e98a427e79e7b8ace19da6185ea9429bc163fedb758891964d602997f338c54",
+  "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": "0fe8b8e27fde77973694"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/884224f6170f841d1877/workflow.json b/records/sd1/sweep/884224f6170f841d1877/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5b68e6962c3c01629106e88048986b543b18f19
--- /dev/null
+++ b/records/sd1/sweep/884224f6170f841d1877/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 884224f6170f841d1877)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.attn1.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/8842bce56aca2b06d79c/measurements.json b/records/sd1/sweep/8842bce56aca2b06d79c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff2394393e9b5e9ed670486bee217c4eceab2c0d
--- /dev/null
+++ b/records/sd1/sweep/8842bce56aca2b06d79c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8842bce56aca2b06d79c",
+ "values": {
+  "std": {
+   "value": 38.172,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01446,
+   "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": 2.766,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.004681,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.016697,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.002357,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001082,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0004,
+   "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": 0.0149,
+   "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/8842bce56aca2b06d79c/record.json b/records/sd1/sweep/8842bce56aca2b06d79c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0073373f165396cd82e23c225a54cf2d4aaf9d0
--- /dev/null
+++ b/records/sd1/sweep/8842bce56aca2b06d79c/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "8842bce56aca2b06d79c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:08:25Z",
+ "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": 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": "middle_block.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "mid",
+   "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": [
+   "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/8842bce56aca2b06d79c/workflow.json b/records/sd1/sweep/8842bce56aca2b06d79c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8349c58b5b2c1c1aa20a4b05440924cc88db675f
--- /dev/null
+++ b/records/sd1/sweep/8842bce56aca2b06d79c/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 8842bce56aca2b06d79c)"
+  },
+  "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 \"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/884b1aa86057ae71ed7b/measurements.json b/records/sd1/sweep/884b1aa86057ae71ed7b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a0a996b06f982b0ef77eddb58968426d5ab8b6a
--- /dev/null
+++ b/records/sd1/sweep/884b1aa86057ae71ed7b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "884b1aa86057ae71ed7b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.064415,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.075714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.056001,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.018168,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.721,
+   "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": 10.652,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0021,
+   "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/884b1aa86057ae71ed7b/record.json b/records/sd1/sweep/884b1aa86057ae71ed7b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b10f9c1cc052a81e32dd004803372012e9f4297e
--- /dev/null
+++ b/records/sd1/sweep/884b1aa86057ae71ed7b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "884b1aa86057ae71ed7b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:35Z",
+ "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": "input_blocks.8.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "cross_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/1352dcf8310a3ae8e066bb9e7530646a6e2f5165473ddf64686c0932eb8f84c0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1352dcf8310a3ae8e066bb9e7530646a6e2f5165473ddf64686c0932eb8f84c0",
+  "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": "9075239150e0847b1f96"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/884b1aa86057ae71ed7b/workflow.json b/records/sd1/sweep/884b1aa86057ae71ed7b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa3e596635e7eac269f528d8d6efa89bcc30b176
--- /dev/null
+++ b/records/sd1/sweep/884b1aa86057ae71ed7b/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 884b1aa86057ae71ed7b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.attn2.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/884e68e9c823d21e4557/measurements.json b/records/sd1/sweep/884e68e9c823d21e4557/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6a373776ad1fa54ce5348675a5d43f76cd218e0
--- /dev/null
+++ b/records/sd1/sweep/884e68e9c823d21e4557/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "884e68e9c823d21e4557",
+ "values": {
+  "std": {
+   "value": 30.163,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01584,
+   "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": 33.601,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.363145,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.600714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.387508,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.144162,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0477,
+   "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": 0.0483,
+   "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/884e68e9c823d21e4557/record.json b/records/sd1/sweep/884e68e9c823d21e4557/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e84e8fd6a4173d616f06b0f04081bb385f777da9
--- /dev/null
+++ b/records/sd1/sweep/884e68e9c823d21e4557/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "884e68e9c823d21e4557",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:37:30Z",
+ "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": 0.7
+   },
+   "group": "in.lo",
+   "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": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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/884e68e9c823d21e4557/workflow.json b/records/sd1/sweep/884e68e9c823d21e4557/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..745c8b03f206db4529f168ab38de7f8904cc0973
--- /dev/null
+++ b/records/sd1/sweep/884e68e9c823d21e4557/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 884e68e9c823d21e4557)"
+  },
+  "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\": 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": 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/885043175c0b4ca34f78/measurements.json b/records/sd1/sweep/885043175c0b4ca34f78/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2266ca55e2cf170a3dbcb2ead505ddb2156b11d3
--- /dev/null
+++ b/records/sd1/sweep/885043175c0b4ca34f78/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "885043175c0b4ca34f78",
+ "values": {
+  "cosine_distance": {
+   "value": 0.190054,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.275113,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.120326,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.061475,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.937,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00473,
+   "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": 31.204,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.1118,
+   "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/885043175c0b4ca34f78/record.json b/records/sd1/sweep/885043175c0b4ca34f78/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bd885dff6f293e448aa8c4788b84e8bbe3a0282
--- /dev/null
+++ b/records/sd1/sweep/885043175c0b4ca34f78/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "885043175c0b4ca34f78",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:03Z",
+ "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": "output_blocks.10.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "skip",
+   "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/b0079c8a6b9a0b23939a1113a0c6b2174588527c986061f9463b4bb92f602f5b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b0079c8a6b9a0b23939a1113a0c6b2174588527c986061f9463b4bb92f602f5b",
+  "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",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "528447feeaa3c7dbd1e6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/885043175c0b4ca34f78/workflow.json b/records/sd1/sweep/885043175c0b4ca34f78/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..67a4464aab91ab127751d20b5e310ba7671534e5
--- /dev/null
+++ b/records/sd1/sweep/885043175c0b4ca34f78/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 885043175c0b4ca34f78)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.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/885199b41453f7ca180e/measurements.json b/records/sd1/sweep/885199b41453f7ca180e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f31385801365ea7c16b39f84d121f301afebfcc1
--- /dev/null
+++ b/records/sd1/sweep/885199b41453f7ca180e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "885199b41453f7ca180e",
+ "values": {
+  "std": {
+   "value": 32.061,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02669,
+   "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": 45.384,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.50434,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.584191,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.521321,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.295636,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8161,
+   "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": 0.4758,
+   "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/885199b41453f7ca180e/record.json b/records/sd1/sweep/885199b41453f7ca180e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e714b7ea54ae5335fc9432ce1928876fd536168
--- /dev/null
+++ b/records/sd1/sweep/885199b41453f7ca180e/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "885199b41453f7ca180e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:15:00Z",
+ "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 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": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "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",
+   "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": "22fe3891a19e55c66d6b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/885199b41453f7ca180e/workflow.json b/records/sd1/sweep/885199b41453f7ca180e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ac3e37300984e50c6531855f5ef8186ed75dd9a
--- /dev/null
+++ b/records/sd1/sweep/885199b41453f7ca180e/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 885199b41453f7ca180e)"
+  },
+  "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}",
+   "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/8856fe53794a3bd92986/measurements.json b/records/sd1/sweep/8856fe53794a3bd92986/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..26a4999f3064a7e076b3cc3a4bcfb4492f7b4a89
--- /dev/null
+++ b/records/sd1/sweep/8856fe53794a3bd92986/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8856fe53794a3bd92986",
+ "values": {
+  "cosine_distance": {
+   "value": 0.096351,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.101009,
+   "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": 0.019115,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.875,
+   "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": 12.766,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9715,
+   "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": 0.1606,
+   "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/8856fe53794a3bd92986/record.json b/records/sd1/sweep/8856fe53794a3bd92986/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..aada38956e56664458dd74ced5de1a462ea5c15f
--- /dev/null
+++ b/records/sd1/sweep/8856fe53794a3bd92986/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8856fe53794a3bd92986",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:18Z",
+ "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": 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.8.0.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/846c64afb1a9c7c062006c168c4f7a9a220a9c0474b29ec72bf57e3ad809617f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "846c64afb1a9c7c062006c168c4f7a9a220a9c0474b29ec72bf57e3ad809617f",
+  "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": "e3e67c773ce2dac11f3d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8856fe53794a3bd92986/workflow.json b/records/sd1/sweep/8856fe53794a3bd92986/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc31c6d7d3988bc5048abcb79f6a3e4ab4e8db03
--- /dev/null
+++ b/records/sd1/sweep/8856fe53794a3bd92986/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 8856fe53794a3bd92986)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\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/885d86df09339439edf6/interpretations.jsonl b/records/sd1/sweep/885d86df09339439edf6/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ff36ddbb09011ba181cea2c7c058497129141f0d
--- /dev/null
+++ b/records/sd1/sweep/885d86df09339439edf6/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "885d86df09339439edf6"}, "kind": "caption", "value": "Blurred portrait of a woman with curly brown hair, face softened into an indistinct haze", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:32:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b467ebc40b524041"}
+{"target": {"record": "885d86df09339439edf6"}, "kind": "change", "value": "Hair turns curlier and fuller while the face and lower body blur heavily into soft smears, and the palette goes muted grey-beige, like an out-of-focus painting.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:32:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cc75a9f5d296e6aa"}
+{"target": {"record": "885d86df09339439edf6"}, "kind": "keywords", "value": ["blurred face", "soft focus portrait", "curly hair", "hazy painting", "muted beige"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:32:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0639bdc970c72470"}
+{"target": {"record": "885d86df09339439edf6"}, "kind": "effect_tags", "value": ["blur", "washed_out", "painterly", "distorted", "desaturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:32:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2b801db6b68a5466"}
diff --git a/records/sd1/sweep/885d86df09339439edf6/measurements.json b/records/sd1/sweep/885d86df09339439edf6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08d550d1352d89b2c9621e529cab54178147de2a
--- /dev/null
+++ b/records/sd1/sweep/885d86df09339439edf6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "885d86df09339439edf6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.169368,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.333064,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.220635,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.118741,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.815,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00592,
+   "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": 23.999,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.0729,
+   "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/885d86df09339439edf6/record.json b/records/sd1/sweep/885d86df09339439edf6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0ceee84fea8120452bb6480c4b57a92098caabb
--- /dev/null
+++ b/records/sd1/sweep/885d86df09339439edf6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "885d86df09339439edf6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:58Z",
+ "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": 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.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/315bdd2b34d54fa085d9d990d1d309e22f86b593f27fae0cd0a078be32d7f45f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "315bdd2b34d54fa085d9d990d1d309e22f86b593f27fae0cd0a078be32d7f45f",
+  "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": "2482ea9a49f96ab77b65"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/885d86df09339439edf6/workflow.json b/records/sd1/sweep/885d86df09339439edf6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd6f0792be9151789818586d049fb9d13b7bc289
--- /dev/null
+++ b/records/sd1/sweep/885d86df09339439edf6/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 885d86df09339439edf6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/886ae0e4c42ae7f9fa5d/measurements.json b/records/sd1/sweep/886ae0e4c42ae7f9fa5d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..747375083c335535ce997fdd817210e28dda8efe
--- /dev/null
+++ b/records/sd1/sweep/886ae0e4c42ae7f9fa5d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "886ae0e4c42ae7f9fa5d",
+ "values": {
+  "std": {
+   "value": 59.188,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00913,
+   "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": 16.765,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.102168,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.142195,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.053146,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01819,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0147,
+   "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": 0.3023,
+   "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/886ae0e4c42ae7f9fa5d/record.json b/records/sd1/sweep/886ae0e4c42ae7f9fa5d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc0daf2100a35064f48abf070a6c9cbb3cc6f898
--- /dev/null
+++ b/records/sd1/sweep/886ae0e4c42ae7f9fa5d/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "886ae0e4c42ae7f9fa5d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:38:43Z",
+ "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": 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.8.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 1234
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "heavy",
+   "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": [
+   "in.lo",
+   "add_noise",
+   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": "f531bdf0c12f1ee9ad96"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/886ae0e4c42ae7f9fa5d/workflow.json b/records/sd1/sweep/886ae0e4c42ae7f9fa5d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..96f5d6d39b52fc0435096a3a9ae614e7c8276b00
--- /dev/null
+++ b/records/sd1/sweep/886ae0e4c42ae7f9fa5d/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 886ae0e4c42ae7f9fa5d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 1234\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/886fd060579c0527e51b/measurements.json b/records/sd1/sweep/886fd060579c0527e51b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ce32751dc36bef63fdbad94d16669b8b3bb97ef
--- /dev/null
+++ b/records/sd1/sweep/886fd060579c0527e51b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "886fd060579c0527e51b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.17648,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.206057,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.232503,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.234124,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.56,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00518,
+   "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": 19.861,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9718,
+   "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": 0.1575,
+   "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/886fd060579c0527e51b/record.json b/records/sd1/sweep/886fd060579c0527e51b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7898294b3a6ec0418bcd673b1851c4f0d2cd3ba
--- /dev/null
+++ b/records/sd1/sweep/886fd060579c0527e51b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "886fd060579c0527e51b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:11Z",
+ "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.5.0.in_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "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/b5d9729fe8c5a9b8bea22340d3de9ba6fdc9b02da1dea00eb504d8029ed187ee.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b5d9729fe8c5a9b8bea22340d3de9ba6fdc9b02da1dea00eb504d8029ed187ee",
+  "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": "b80780d3d501a61b5ad6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/886fd060579c0527e51b/workflow.json b/records/sd1/sweep/886fd060579c0527e51b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dd6ff2e8abbb643785a5d3009e9b9ab3e9e9e34
--- /dev/null
+++ b/records/sd1/sweep/886fd060579c0527e51b/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 886fd060579c0527e51b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.in_layers.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/88771e3d6373e971ef44/interpretations.jsonl b/records/sd1/sweep/88771e3d6373e971ef44/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..568a7fc5b3303921d1d86bd7facd0fb283f84ace
--- /dev/null
+++ b/records/sd1/sweep/88771e3d6373e971ef44/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "88771e3d6373e971ef44"}, "kind": "caption", "value": "Stippled pointillist portrait of a woman with black bob on speckled light 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": "48bc84bff872dc3f"}
+{"target": {"record": "88771e3d6373e971ef44"}, "kind": "change", "value": "The painting turns into a stippled, pointillist texture of fine dots, with bold black hair, high contrast, and a speckled pale green-white 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": "02595c95669c976c"}
+{"target": {"record": "88771e3d6373e971ef44"}, "kind": "keywords", "value": ["pointillism", "stippled", "black bob", "dotted texture", "high contrast"], "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": "4d424441ea87c3ff"}
+{"target": {"record": "88771e3d6373e971ef44"}, "kind": "effect_tags", "value": ["texture", "pattern", "style_change", "high_contrast", "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": "db1c0e4fe43e89f1"}
diff --git a/records/sd1/sweep/88771e3d6373e971ef44/measurements.json b/records/sd1/sweep/88771e3d6373e971ef44/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..785eba76a6437dc0cb21dc7bfcecf2cd166e392e
--- /dev/null
+++ b/records/sd1/sweep/88771e3d6373e971ef44/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88771e3d6373e971ef44",
+ "values": {
+  "cosine_distance": {
+   "value": 0.290843,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.418737,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.297361,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.116326,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 80.25,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00781,
+   "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": 43.266,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0627,
+   "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": 0.0232,
+   "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/88771e3d6373e971ef44/record.json b/records/sd1/sweep/88771e3d6373e971ef44/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..348030a6b8980329fd7d234105a49cee7271227d
--- /dev/null
+++ b/records/sd1/sweep/88771e3d6373e971ef44/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88771e3d6373e971ef44",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:36Z",
+ "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": "input_blocks.1.0.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/2c2fd5acf1a659e4f307d005309648967b4d5dae04c4f21cc2cc4e7ca6cd487a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2c2fd5acf1a659e4f307d005309648967b4d5dae04c4f21cc2cc4e7ca6cd487a",
+  "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": "837b7833b578d1b857c1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88771e3d6373e971ef44/workflow.json b/records/sd1/sweep/88771e3d6373e971ef44/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6e8c39a05aa3e772f0fcb69dc8d48a694698df1
--- /dev/null
+++ b/records/sd1/sweep/88771e3d6373e971ef44/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 88771e3d6373e971ef44)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.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/8879ad770404a2b3ca36/interpretations.jsonl b/records/sd1/sweep/8879ad770404a2b3ca36/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..61374b7fca3a6d0a225d0be8b552f3330e217e25
--- /dev/null
+++ b/records/sd1/sweep/8879ad770404a2b3ca36/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8879ad770404a2b3ca36"}, "kind": "caption", "value": "Illustrated portrait of a woman with wavy hair and black turtleneck, nearly identical", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:21:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "999174a8326427ae"}
+{"target": {"record": "8879ad770404a2b3ca36"}, "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": "2026-10-05T21:21:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e1e7a44a35445e56"}
+{"target": {"record": "8879ad770404a2b3ca36"}, "kind": "keywords", "value": ["no change", "portrait", "identical", "turtleneck"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:21:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8763cb45e7c3f5f3"}
+{"target": {"record": "8879ad770404a2b3ca36"}, "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": "2026-10-05T21:21:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cb1b6672a9852a42"}
diff --git a/records/sd1/sweep/8879ad770404a2b3ca36/measurements.json b/records/sd1/sweep/8879ad770404a2b3ca36/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e76b8e84e9b975c3e6300e2cf1d28cfb05230fd8
--- /dev/null
+++ b/records/sd1/sweep/8879ad770404a2b3ca36/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8879ad770404a2b3ca36",
+ "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/8879ad770404a2b3ca36/record.json b/records/sd1/sweep/8879ad770404a2b3ca36/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..66ee263c30140db7da03d8bab26074857a31c299
--- /dev/null
+++ b/records/sd1/sweep/8879ad770404a2b3ca36/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8879ad770404a2b3ca36",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:49Z",
+ "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": "input_blocks.4.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "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/e5a153402391148d338bc1b059ad0c538dda0e755c71bfb5232700b565621794.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e5a153402391148d338bc1b059ad0c538dda0e755c71bfb5232700b565621794",
+  "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": "4aaecbbf69cb5baa8237"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8879ad770404a2b3ca36/workflow.json b/records/sd1/sweep/8879ad770404a2b3ca36/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c50d02c8cd5f6ca13bf6ac4b4bcaf8ad250f713d
--- /dev/null
+++ b/records/sd1/sweep/8879ad770404a2b3ca36/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 8879ad770404a2b3ca36)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.h_upd\",\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/887b6e540a4a482f7c31/measurements.json b/records/sd1/sweep/887b6e540a4a482f7c31/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08d98be55bb5f7250bb157d568573478e4053b4c
--- /dev/null
+++ b/records/sd1/sweep/887b6e540a4a482f7c31/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "887b6e540a4a482f7c31",
+ "values": {
+  "std": {
+   "value": 82.163,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00251,
+   "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": 74.503,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.649452,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.738867,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 1.028484,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.396344,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.51,
+   "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": 0.8629,
+   "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/887b6e540a4a482f7c31/record.json b/records/sd1/sweep/887b6e540a4a482f7c31/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..90dbf12da2ee620c09702bf82dd51607fa9768e6
--- /dev/null
+++ b/records/sd1/sweep/887b6e540a4a482f7c31/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "887b6e540a4a482f7c31",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:30:58Z",
+ "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": 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": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "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,
+   "all"
+  ],
+  "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": "21c0e052db6090362942"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/887b6e540a4a482f7c31/workflow.json b/records/sd1/sweep/887b6e540a4a482f7c31/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..74ec448a8ff8bceba8a74b505cd3717336c7e4a4
--- /dev/null
+++ b/records/sd1/sweep/887b6e540a4a482f7c31/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 887b6e540a4a482f7c31)"
+  },
+  "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}",
+   "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/887f3066e6097a800c02/measurements.json b/records/sd1/sweep/887f3066e6097a800c02/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..81feeb4bf16a440d4a02c426f0218fbf670d005d
--- /dev/null
+++ b/records/sd1/sweep/887f3066e6097a800c02/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "887f3066e6097a800c02",
+ "values": {
+  "std": {
+   "value": 31.804,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02081,
+   "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": 14.171,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.083743,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.131804,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.064062,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.017058,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.3199,
+   "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/887f3066e6097a800c02/record.json b/records/sd1/sweep/887f3066e6097a800c02/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..583cf1e4aee6f61758bf32cce8c79a7ad17ee563
--- /dev/null
+++ b/records/sd1/sweep/887f3066e6097a800c02/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "887f3066e6097a800c02",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:21:08Z",
+ "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": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "damp",
+   "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": [
+   "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/887f3066e6097a800c02/workflow.json b/records/sd1/sweep/887f3066e6097a800c02/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6676b7af0611455be2d43e99c9eb6acbea827f23
--- /dev/null
+++ b/records/sd1/sweep/887f3066e6097a800c02/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 887f3066e6097a800c02)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\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": 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/888304ea83e7d17e06c7/measurements.json b/records/sd1/sweep/888304ea83e7d17e06c7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8bdf37eec763ca5c985ae6e77afa5ba5dc99d506
--- /dev/null
+++ b/records/sd1/sweep/888304ea83e7d17e06c7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "888304ea83e7d17e06c7",
+ "values": {
+  "std": {
+   "value": 28.362,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03364,
+   "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": 24.096,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.20654,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.401789,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.199138,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.093174,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9887,
+   "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": 0.4372,
+   "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/888304ea83e7d17e06c7/record.json b/records/sd1/sweep/888304ea83e7d17e06c7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..86f64ebf3bf582abb88aa1b687682afaf630f371
--- /dev/null
+++ b/records/sd1/sweep/888304ea83e7d17e06c7/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "888304ea83e7d17e06c7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:15:43Z",
+ "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": 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": "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": "81ef55543b3b3c080780",
+  "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/888304ea83e7d17e06c7/workflow.json b/records/sd1/sweep/888304ea83e7d17e06c7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4603a207ac01cc87bcef11d322231a6188291026
--- /dev/null
+++ b/records/sd1/sweep/888304ea83e7d17e06c7/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 888304ea83e7d17e06c7)"
+  },
+  "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": 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/888575dd9e8c52e0e6e9/measurements.json b/records/sd1/sweep/888575dd9e8c52e0e6e9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..13fc1d982b9b582b04258aaecab84e8001a449fb
--- /dev/null
+++ b/records/sd1/sweep/888575dd9e8c52e0e6e9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "888575dd9e8c52e0e6e9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.310151,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.434424,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.335389,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.125142,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 42.814,
+   "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": [],
+   "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": 34.83,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9736,
+   "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": 0.0792,
+   "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/888575dd9e8c52e0e6e9/record.json b/records/sd1/sweep/888575dd9e8c52e0e6e9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e74842d68b9efc01caca1daade0327a225b3b200
--- /dev/null
+++ b/records/sd1/sweep/888575dd9e8c52e0e6e9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "888575dd9e8c52e0e6e9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:39Z",
+ "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.8.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.lo",
+   "kind": "skip",
+   "module_type": "Identity",
+   "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/4f6832277aa04bb1df2c752289c785b3bf61e7e64d4770cb9da7f65da855b680.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4f6832277aa04bb1df2c752289c785b3bf61e7e64d4770cb9da7f65da855b680",
+  "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": "be2c4b8c0348724ec0db"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/888575dd9e8c52e0e6e9/workflow.json b/records/sd1/sweep/888575dd9e8c52e0e6e9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae1fd16c2cb066a2d7dc9a873b41f6891b5a7c4a
--- /dev/null
+++ b/records/sd1/sweep/888575dd9e8c52e0e6e9/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 888575dd9e8c52e0e6e9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.skip_connection\",\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/8885ac86d3ff5e7fde3b/measurements.json b/records/sd1/sweep/8885ac86d3ff5e7fde3b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8afd745874bf4ab50006b8e87e6e0dc455a2e62f
--- /dev/null
+++ b/records/sd1/sweep/8885ac86d3ff5e7fde3b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8885ac86d3ff5e7fde3b",
+ "values": {
+  "std": {
+   "value": 50.949,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01108,
+   "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": 4.42,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.011399,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.046034,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.034306,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.005959,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0005,
+   "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": 0.0187,
+   "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/8885ac86d3ff5e7fde3b/record.json b/records/sd1/sweep/8885ac86d3ff5e7fde3b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b484abee3de4c87e3e7836e2b469984c26ad6db7
--- /dev/null
+++ b/records/sd1/sweep/8885ac86d3ff5e7fde3b/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8885ac86d3ff5e7fde3b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:36:24Z",
+ "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": 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": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "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": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.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": "947072118c4be01f182f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8885ac86d3ff5e7fde3b/workflow.json b/records/sd1/sweep/8885ac86d3ff5e7fde3b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fbfc50f0e3e2642d0db9982fbf9f37948287cbc
--- /dev/null
+++ b/records/sd1/sweep/8885ac86d3ff5e7fde3b/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 8885ac86d3ff5e7fde3b)"
+  },
+  "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\": 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": 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/8886a18f0303aa3ab4c9/measurements.json b/records/sd1/sweep/8886a18f0303aa3ab4c9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..842d38daf19f456fdc0fa71f32d2d07aa7e8d940
--- /dev/null
+++ b/records/sd1/sweep/8886a18f0303aa3ab4c9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8886a18f0303aa3ab4c9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.414501,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.483954,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.276687,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.108085,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.756,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00164,
+   "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": 44.078,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9968,
+   "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": 0.1809,
+   "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/8886a18f0303aa3ab4c9/record.json b/records/sd1/sweep/8886a18f0303aa3ab4c9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfe776064330851fd73436a1ef7606dc3d37a24e
--- /dev/null
+++ b/records/sd1/sweep/8886a18f0303aa3ab4c9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8886a18f0303aa3ab4c9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:31Z",
+ "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": "input_blocks.5.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "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/3d8ce76970e6fa70b17585dd90e29a4bc8cb9496632acf22e960cdc5c7bbd2b0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3d8ce76970e6fa70b17585dd90e29a4bc8cb9496632acf22e960cdc5c7bbd2b0",
+  "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": "da13537481f3fe75be8a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8886a18f0303aa3ab4c9/workflow.json b/records/sd1/sweep/8886a18f0303aa3ab4c9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6030ad69c4b55f261645549d64b6382e56a901e1
--- /dev/null
+++ b/records/sd1/sweep/8886a18f0303aa3ab4c9/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 8886a18f0303aa3ab4c9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.proj_in\",\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/8887eb092ee7e894d5e8/measurements.json b/records/sd1/sweep/8887eb092ee7e894d5e8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a6caa3b3eba0f44269926f6f516c4d89e41b5c4
--- /dev/null
+++ b/records/sd1/sweep/8887eb092ee7e894d5e8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8887eb092ee7e894d5e8",
+ "values": {
+  "std": {
+   "value": 64.388,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00994,
+   "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": 14.976,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.05084,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.15396,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.046818,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.046128,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.4178,
+   "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/8887eb092ee7e894d5e8/record.json b/records/sd1/sweep/8887eb092ee7e894d5e8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bed2cc25efc8b83360c24f406dad449b0bd3fa7
--- /dev/null
+++ b/records/sd1/sweep/8887eb092ee7e894d5e8/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8887eb092ee7e894d5e8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:33:09Z",
+ "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": 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.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "boost",
+   "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.hi",
+   "multiply",
+   2.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 2.0,
+    "applied": 2.0,
+    "reason": "outside safe range [0.6, 1.6] 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": "00ce0384a53ef28ab4a7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8887eb092ee7e894d5e8/workflow.json b/records/sd1/sweep/8887eb092ee7e894d5e8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bf732687b11de11f074c821c9b601f541051cb1
--- /dev/null
+++ b/records/sd1/sweep/8887eb092ee7e894d5e8/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 8887eb092ee7e894d5e8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\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/88895764f6b4d33498ca/measurements.json b/records/sd1/sweep/88895764f6b4d33498ca/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a071ca1712a7422bc28ee6de4bb8b43e03ab954
--- /dev/null
+++ b/records/sd1/sweep/88895764f6b4d33498ca/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88895764f6b4d33498ca",
+ "values": {
+  "cosine_distance": {
+   "value": 0.119875,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.122294,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.053401,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.068777,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.057,
+   "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": 13.186,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0057,
+   "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": 0.4427,
+   "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/88895764f6b4d33498ca/record.json b/records/sd1/sweep/88895764f6b4d33498ca/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b535026497ed51d4eafe748998c24ee7212ed395
--- /dev/null
+++ b/records/sd1/sweep/88895764f6b4d33498ca/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "88895764f6b4d33498ca",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:36Z",
+ "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.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/1ee06e7cf667818cef96609796ff52c2c0e4013938875bc60e97930630aee926.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1ee06e7cf667818cef96609796ff52c2c0e4013938875bc60e97930630aee926",
+  "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"
+    ],
+    "previous_id": "acbae8ca4b784bd37fbc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88895764f6b4d33498ca/workflow.json b/records/sd1/sweep/88895764f6b4d33498ca/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..55cf183849f299bf4d3e9b717fcc39247fc6af59
--- /dev/null
+++ b/records/sd1/sweep/88895764f6b4d33498ca/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 88895764f6b4d33498ca)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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": [
+    "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/888e509e850249360579/measurements.json b/records/sd1/sweep/888e509e850249360579/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf7b9edd9148cd356d178f9163fa0bb7d1caa7af
--- /dev/null
+++ b/records/sd1/sweep/888e509e850249360579/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "888e509e850249360579",
+ "values": {
+  "std": {
+   "value": 50.87,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01497,
+   "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": 1.648,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001762,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.007606,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004615,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001638,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0064,
+   "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": 0.1119,
+   "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/888e509e850249360579/record.json b/records/sd1/sweep/888e509e850249360579/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..acc1bd2e5e24ceb5d1dcd83d88dd7c0ec356fdf9
--- /dev/null
+++ b/records/sd1/sweep/888e509e850249360579/record.json
@@ -0,0 +1,86 @@
+{
+ "id": "888e509e850249360579",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:12:14Z",
+ "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 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": 0.333,
+   "window": "early",
+   "bucket": "medium",
+   "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": [
+   "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": "7051e76ff49983349131"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/888e509e850249360579/workflow.json b/records/sd1/sweep/888e509e850249360579/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f12e5fdc93a36a321a3ab1a04cfed9fbb3f178f
--- /dev/null
+++ b/records/sd1/sweep/888e509e850249360579/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 888e509e850249360579)"
+  },
+  "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 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/889c21cb6c133ee4b84e/interpretations.jsonl b/records/sd1/sweep/889c21cb6c133ee4b84e/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a100c9aa481dbc440ffb2f30b3a3fe601951917a
--- /dev/null
+++ b/records/sd1/sweep/889c21cb6c133ee4b84e/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "889c21cb6c133ee4b84e"}, "kind": "caption", "value": "Young man with brown swept hair in dark suit jacket and tie, frontal studio portrait", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cd1a95bd19dfd643"}
+{"target": {"record": "889c21cb6c133ee4b84e"}, "kind": "change", "value": "Similar to a formal headshot: the woman becomes a young man in jacket and tie facing more frontally, with warmer skin tones and a cleaner modern photographic look.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7cfcc0dcbd871686"}
+{"target": {"record": "889c21cb6c133ee4b84e"}, "kind": "keywords", "value": ["formal headshot", "suit", "gender swap", "young man", "studio portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c3161cc48dfa7f6d"}
+{"target": {"record": "889c21cb6c133ee4b84e"}, "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": "2026-10-05T08:26:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9bfed0cd634c2942"}
diff --git a/records/sd1/sweep/889c21cb6c133ee4b84e/measurements.json b/records/sd1/sweep/889c21cb6c133ee4b84e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5704af343fcce8c618001f1735f8f28de3e0e599
--- /dev/null
+++ b/records/sd1/sweep/889c21cb6c133ee4b84e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "889c21cb6c133ee4b84e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.218529,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.218823,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.257986,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.209659,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.897,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0041,
+   "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": 21.559,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9328,
+   "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": 0.6728,
+   "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/889c21cb6c133ee4b84e/record.json b/records/sd1/sweep/889c21cb6c133ee4b84e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..32e40e71b9e98e71fb2a5e896b629896fbab9326
--- /dev/null
+++ b/records/sd1/sweep/889c21cb6c133ee4b84e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "889c21cb6c133ee4b84e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:03Z",
+ "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": "middle_block.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "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/dc0e59f6a0e7f9b28afd601cea65553e666dbbad3456a2cceedd051532e57fb0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "dc0e59f6a0e7f9b28afd601cea65553e666dbbad3456a2cceedd051532e57fb0",
+  "type_path": "UNetModel.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"
+    ],
+    "previous_id": "193e31b71f1566ffcc60"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/889c21cb6c133ee4b84e/workflow.json b/records/sd1/sweep/889c21cb6c133ee4b84e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..954798a502d5415b83488abfe319c5389b6f7228
--- /dev/null
+++ b/records/sd1/sweep/889c21cb6c133ee4b84e/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 889c21cb6c133ee4b84e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.attn1.to_out.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/88a24e30414f28baa034/measurements.json b/records/sd1/sweep/88a24e30414f28baa034/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6865271ddaf0be5f74a69750264170aac732c299
--- /dev/null
+++ b/records/sd1/sweep/88a24e30414f28baa034/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88a24e30414f28baa034",
+ "values": {
+  "cosine_distance": {
+   "value": 0.326692,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.40342,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.387114,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.231434,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.161,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00375,
+   "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": 33.142,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9988,
+   "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": 0.2314,
+   "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/88a24e30414f28baa034/record.json b/records/sd1/sweep/88a24e30414f28baa034/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed9facd31112eb3895328c49c1eb64f8cdf53e79
--- /dev/null
+++ b/records/sd1/sweep/88a24e30414f28baa034/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "88a24e30414f28baa034",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:03Z",
+ "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": "middle_block.1.transformer_blocks.0.attn1.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "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/7d55dbb647ec2d34a86d587b38b1bcf2e937bcbe35461c98ff6d429775ea9233.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7d55dbb647ec2d34a86d587b38b1bcf2e937bcbe35461c98ff6d429775ea9233",
+  "type_path": "UNetModel.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"
+    ],
+    "previous_id": "00427776dc7c5926e85a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88a24e30414f28baa034/workflow.json b/records/sd1/sweep/88a24e30414f28baa034/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f915294f204999ab91fced8e78f639a53a6a0f0a
--- /dev/null
+++ b/records/sd1/sweep/88a24e30414f28baa034/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 88a24e30414f28baa034)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.attn1.to_v\",\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/88a610abe6d52500bef8/measurements.json b/records/sd1/sweep/88a610abe6d52500bef8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce8f85af1cdcfa4c1350627f5cc99837ebed17ed
--- /dev/null
+++ b/records/sd1/sweep/88a610abe6d52500bef8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88a610abe6d52500bef8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.127612,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.187141,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.052449,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.051221,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.038,
+   "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"
+  },
+  "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": 18.997,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9371,
+   "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": 0.0653,
+   "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/88a610abe6d52500bef8/record.json b/records/sd1/sweep/88a610abe6d52500bef8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..db34c1874c7523098758691270dc2264c9abea05
--- /dev/null
+++ b/records/sd1/sweep/88a610abe6d52500bef8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88a610abe6d52500bef8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:47Z",
+ "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.10.0.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9078cc7bb515d57081b05e03a39561338cf41ab5923920ccf460c937a5ec4ab4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9078cc7bb515d57081b05e03a39561338cf41ab5923920ccf460c937a5ec4ab4",
+  "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": "dc0d01fb780d8e32fad5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88a610abe6d52500bef8/workflow.json b/records/sd1/sweep/88a610abe6d52500bef8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..05d58440981a577703eb91d38267461917b2a360
--- /dev/null
+++ b/records/sd1/sweep/88a610abe6d52500bef8/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 88a610abe6d52500bef8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.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": [
+    "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/88b4225d7f914395afaf/measurements.json b/records/sd1/sweep/88b4225d7f914395afaf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..818f495d5d756bf61f1c496aa4ebe51aaaf8231c
--- /dev/null
+++ b/records/sd1/sweep/88b4225d7f914395afaf/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88b4225d7f914395afaf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.073329,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.1082,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.045628,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.022285,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.662,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00587,
+   "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": 14.022,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0513,
+   "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": 0.169,
+   "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/88b4225d7f914395afaf/record.json b/records/sd1/sweep/88b4225d7f914395afaf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4dce080812bf20cedb839b37656b55e961d19419
--- /dev/null
+++ b/records/sd1/sweep/88b4225d7f914395afaf/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88b4225d7f914395afaf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:23Z",
+ "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": "middle_block.1.proj_out",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "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/ceeb92eb73601d509ccbb512c9ddbbb62104cfb7afa169375fdb1ea46ecf8aec.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ceeb92eb73601d509ccbb512c9ddbbb62104cfb7afa169375fdb1ea46ecf8aec",
+  "type_path": "UNetModel.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": "daef8a9a33bd0719ef5f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88b4225d7f914395afaf/workflow.json b/records/sd1/sweep/88b4225d7f914395afaf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bbdc8602db3caa3adcb5153aab3eb6dfb28db3c
--- /dev/null
+++ b/records/sd1/sweep/88b4225d7f914395afaf/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 88b4225d7f914395afaf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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": [
+    "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/88b83efd0d367ab73a9c/interpretations.jsonl b/records/sd1/sweep/88b83efd0d367ab73a9c/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..3e28bfe9260308a5aa6047e08e1ce021b77bde57
--- /dev/null
+++ b/records/sd1/sweep/88b83efd0d367ab73a9c/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "88b83efd0d367ab73a9c"}, "kind": "caption", "value": "Extreme close-up of a pale face with one blue-grey eye and soft blurred edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:44Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0eeb93a3869e9ca3"}
+{"target": {"record": "88b83efd0d367ab73a9c"}, "kind": "change", "value": "The framing jumps into an extreme, tilted close-up of a single eye, nose and lips in soft focus, with washed-out warm skin.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:44Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6215fbd6400be438"}
+{"target": {"record": "88b83efd0d367ab73a9c"}, "kind": "keywords", "value": ["extreme close-up", "single eye", "soft focus", "cropped face", "macro portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:44Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e17340531201388a"}
+{"target": {"record": "88b83efd0d367ab73a9c"}, "kind": "effect_tags", "value": ["zoom", "new_composition", "blur", "washed_out"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:28:44Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "39bdf4cdb5acc451"}
diff --git a/records/sd1/sweep/88b83efd0d367ab73a9c/measurements.json b/records/sd1/sweep/88b83efd0d367ab73a9c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..716b32231fb7363d8b895121663f99027ae17f33
--- /dev/null
+++ b/records/sd1/sweep/88b83efd0d367ab73a9c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88b83efd0d367ab73a9c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.438136,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.689476,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.653603,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.388495,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.133,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00259,
+   "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": 47.558,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8716,
+   "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": 0.7676,
+   "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/88b83efd0d367ab73a9c/record.json b/records/sd1/sweep/88b83efd0d367ab73a9c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c830e5bdb63f7fbceed3955e24b5bea6ac9a8c9a
--- /dev/null
+++ b/records/sd1/sweep/88b83efd0d367ab73a9c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "88b83efd0d367ab73a9c",
+ "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.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.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/2a5c96477a554be4d3f2035e0830d1facfa21143a3974f0a9abbcbfd4bba2c18.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2a5c96477a554be4d3f2035e0830d1facfa21143a3974f0a9abbcbfd4bba2c18",
+  "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": "28ed3d4f8693745f7dd9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88b83efd0d367ab73a9c/workflow.json b/records/sd1/sweep/88b83efd0d367ab73a9c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6db37425410e3aa22ee39b10eb982c7e6f653083
--- /dev/null
+++ b/records/sd1/sweep/88b83efd0d367ab73a9c/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 88b83efd0d367ab73a9c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.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/88d03ba9c36c1f92c969/measurements.json b/records/sd1/sweep/88d03ba9c36c1f92c969/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..75c983c33fb5417833983d0db1112953363e41dc
--- /dev/null
+++ b/records/sd1/sweep/88d03ba9c36c1f92c969/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88d03ba9c36c1f92c969",
+ "values": {
+  "std": {
+   "value": 71.353,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01237,
+   "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": 37.948,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.284192,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.404997,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.355976,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.097166,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9474,
+   "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": 0.3671,
+   "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/88d03ba9c36c1f92c969/record.json b/records/sd1/sweep/88d03ba9c36c1f92c969/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..147f2b4592fd00a1ff04ae5546159a5ccda72032
--- /dev/null
+++ b/records/sd1/sweep/88d03ba9c36c1f92c969/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "88d03ba9c36c1f92c969",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:18:06Z",
+ "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": 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": 2.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "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",
+   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": "ede70f75bf3af7edeb7a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88d03ba9c36c1f92c969/workflow.json b/records/sd1/sweep/88d03ba9c36c1f92c969/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8154880d27eecf4c117e850d6dbe89c5dba50221
--- /dev/null
+++ b/records/sd1/sweep/88d03ba9c36c1f92c969/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 88d03ba9c36c1f92c969)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\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": "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/88d08ef705c8df39388b/measurements.json b/records/sd1/sweep/88d08ef705c8df39388b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..901c35b7496991189a8619c28759e988380c7f7f
--- /dev/null
+++ b/records/sd1/sweep/88d08ef705c8df39388b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88d08ef705c8df39388b",
+ "values": {
+  "std": {
+   "value": 22.69,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01701,
+   "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": 39.462,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.418613,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.65813,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.594302,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.20125,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9795,
+   "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": 0.0453,
+   "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/88d08ef705c8df39388b/record.json b/records/sd1/sweep/88d08ef705c8df39388b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c62f5d7bc663e0a76ae61576772598ab7b48516
--- /dev/null
+++ b/records/sd1/sweep/88d08ef705c8df39388b/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "88d08ef705c8df39388b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:37:23Z",
+ "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": "input_blocks.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "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",
+   0.7,
+   "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": "ec2c93d03c09720065e7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88d08ef705c8df39388b/workflow.json b/records/sd1/sweep/88d08ef705c8df39388b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ce2965c8a5787563e946476a75212520ad30923
--- /dev/null
+++ b/records/sd1/sweep/88d08ef705c8df39388b/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 88d08ef705c8df39388b)"
+  },
+  "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\": 0.7\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": 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/88d7f025ebb1672a95d8/measurements.json b/records/sd1/sweep/88d7f025ebb1672a95d8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..828ecf896cfe487b77197868dff8d6039db08a4f
--- /dev/null
+++ b/records/sd1/sweep/88d7f025ebb1672a95d8/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88d7f025ebb1672a95d8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.103645,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.161278,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.079381,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.049927,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.818,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00375,
+   "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": 17.005,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0749,
+   "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": 0.1309,
+   "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/88d7f025ebb1672a95d8/record.json b/records/sd1/sweep/88d7f025ebb1672a95d8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fdc4f0b14f9391cdee32c14f20a828b3f46ac3d
--- /dev/null
+++ b/records/sd1/sweep/88d7f025ebb1672a95d8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88d7f025ebb1672a95d8",
+ "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": 0.5
+   },
+   "group": "out.hi",
+   "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/a2b1a3df7c5ffce896308f983be506e2415041e11d8f17731d26e422c2bbbd84.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a2b1a3df7c5ffce896308f983be506e2415041e11d8f17731d26e422c2bbbd84",
+  "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": "c75ec07e35c6ce26bf79"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88d7f025ebb1672a95d8/workflow.json b/records/sd1/sweep/88d7f025ebb1672a95d8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ecc18fc0deaf420074b46df3c919b0337982779
--- /dev/null
+++ b/records/sd1/sweep/88d7f025ebb1672a95d8/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 88d7f025ebb1672a95d8)"
+  },
+  "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\": 0.5\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/88f57280be9f6757f472/interpretations.jsonl b/records/sd1/sweep/88f57280be9f6757f472/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..847c80538283ec80c5f56e0e7b76ea8fc0246244
--- /dev/null
+++ b/records/sd1/sweep/88f57280be9f6757f472/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "88f57280be9f6757f472"}, "kind": "caption", "value": "Translucent ghostly face with doubled features and loose hair strands on pale green", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e0610e407a2a15b8"}
+{"target": {"record": "88f57280be9f6757f472"}, "kind": "change", "value": "The face becomes a see-through ghost with duplicated smudged features and stray hair, in a lighter palette.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "66fd7edb470d9ce2"}
+{"target": {"record": "88f57280be9f6757f472"}, "kind": "keywords", "value": ["ghostly face", "double exposure", "translucent", "loose hair"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7526d84986abdaec"}
+{"target": {"record": "88f57280be9f6757f472"}, "kind": "effect_tags", "value": ["distorted", "washed_out", "surreal", "blur"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6e8088586d89af46"}
diff --git a/records/sd1/sweep/88f57280be9f6757f472/measurements.json b/records/sd1/sweep/88f57280be9f6757f472/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f315104f31024d8e6fdf9319e21e30aac2dfe91
--- /dev/null
+++ b/records/sd1/sweep/88f57280be9f6757f472/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88f57280be9f6757f472",
+ "values": {
+  "cosine_distance": {
+   "value": 0.23302,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.399536,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.406719,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.267907,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.643,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00352,
+   "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": 29.208,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9684,
+   "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": 0.6081,
+   "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/88f57280be9f6757f472/record.json b/records/sd1/sweep/88f57280be9f6757f472/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2a313b9d0ddd15a935108da280fb0c09f3ac77e
--- /dev/null
+++ b/records/sd1/sweep/88f57280be9f6757f472/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "88f57280be9f6757f472",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:06Z",
+ "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.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e1df3bf4441e94230402934bb9b13a07332e4861bf4c893d327184b776e71022.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e1df3bf4441e94230402934bb9b13a07332e4861bf4c893d327184b776e71022",
+  "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": "2664ede34a7c672e7550"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88f57280be9f6757f472/workflow.json b/records/sd1/sweep/88f57280be9f6757f472/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bafbccadc5acd224b59d7fa4423aea2e14402df8
--- /dev/null
+++ b/records/sd1/sweep/88f57280be9f6757f472/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 88f57280be9f6757f472)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.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": [
+    "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/88f9844e18f6aa9b98ea/measurements.json b/records/sd1/sweep/88f9844e18f6aa9b98ea/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ec7633d49a1c6faa061425d284e2712d1475556
--- /dev/null
+++ b/records/sd1/sweep/88f9844e18f6aa9b98ea/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88f9844e18f6aa9b98ea",
+ "values": {
+  "std": {
+   "value": 13.675,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03718,
+   "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": 47.609,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.585252,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.699431,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.590898,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.338263,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.82,
+   "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": 0.1165,
+   "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/88f9844e18f6aa9b98ea/record.json b/records/sd1/sweep/88f9844e18f6aa9b98ea/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca0699829d3eb24ee1e2407ab05051056cc660b5
--- /dev/null
+++ b/records/sd1/sweep/88f9844e18f6aa9b98ea/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "88f9844e18f6aa9b98ea",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:41:26Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.lo",
+   "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": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   0.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for out.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": "e8ede41e98b362ee9ef9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88f9844e18f6aa9b98ea/workflow.json b/records/sd1/sweep/88f9844e18f6aa9b98ea/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5dc3c4ab2bc9c758e431dbd067a757754d657bbb
--- /dev/null
+++ b/records/sd1/sweep/88f9844e18f6aa9b98ea/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 88f9844e18f6aa9b98ea)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\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": 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/88fbb07521474706f778/measurements.json b/records/sd1/sweep/88fbb07521474706f778/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad15dc66e72d74ed5931b69a21054af3e3c40b2
--- /dev/null
+++ b/records/sd1/sweep/88fbb07521474706f778/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88fbb07521474706f778",
+ "values": {
+  "cosine_distance": {
+   "value": 0.107049,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.17405,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.133625,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.03225,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.162,
+   "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"
+  },
+  "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": 21.641,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9832,
+   "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": 0.0776,
+   "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/88fbb07521474706f778/record.json b/records/sd1/sweep/88fbb07521474706f778/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..450c3d784bb74f4572a9c398b4aa99373000c4ff
--- /dev/null
+++ b/records/sd1/sweep/88fbb07521474706f778/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88fbb07521474706f778",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:54Z",
+ "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": "output_blocks.4.1.transformer_blocks.0.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "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/e1d146cd7f4b40f966a4d5f66e3408f86c0e1d7c00f264f3885160a470a02818.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e1d146cd7f4b40f966a4d5f66e3408f86c0e1d7c00f264f3885160a470a02818",
+  "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": "17a156766529cbb6a053"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88fbb07521474706f778/workflow.json b/records/sd1/sweep/88fbb07521474706f778/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..abce9be181eb3ff96712539b23d6a011ff5918e7
--- /dev/null
+++ b/records/sd1/sweep/88fbb07521474706f778/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 88fbb07521474706f778)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.transformer_blocks.0.attn2.to_v\",\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/88ff24fb313eed723b97/measurements.json b/records/sd1/sweep/88ff24fb313eed723b97/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..04d6086842d9cafe6b978901bb7406b2701db8da
--- /dev/null
+++ b/records/sd1/sweep/88ff24fb313eed723b97/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "88ff24fb313eed723b97",
+ "values": {
+  "cosine_distance": {
+   "value": 0.08044,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0879,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.119764,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.181559,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.96,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00601,
+   "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": 11.334,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8803,
+   "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": 0.4529,
+   "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/88ff24fb313eed723b97/record.json b/records/sd1/sweep/88ff24fb313eed723b97/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f77410c54add5a27b26645ad2ca17ad1e9bdc7ae
--- /dev/null
+++ b/records/sd1/sweep/88ff24fb313eed723b97/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "88ff24fb313eed723b97",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:35Z",
+ "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": "output_blocks.6.1.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.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/7d55ce650ed5f7e320abb7cb04c0f19f2ae9b0bc54fad467f61c17f0f7f0b347.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7d55ce650ed5f7e320abb7cb04c0f19f2ae9b0bc54fad467f61c17f0f7f0b347",
+  "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": "95755c20839191be57ea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88ff24fb313eed723b97/workflow.json b/records/sd1/sweep/88ff24fb313eed723b97/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6dd5e6cd726534bc6ced798ffb7eaea06b1fa7b5
--- /dev/null
+++ b/records/sd1/sweep/88ff24fb313eed723b97/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 88ff24fb313eed723b97)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.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/88ffd5d3559fe8ae9a04/measurements.json b/records/sd1/sweep/88ffd5d3559fe8ae9a04/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..19a5d0a1b15147031b932b36b823a8ae3bdad5eb
--- /dev/null
+++ b/records/sd1/sweep/88ffd5d3559fe8ae9a04/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "88ffd5d3559fe8ae9a04",
+ "values": {
+  "cosine_distance": {
+   "value": 0.004073,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.012199,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.003321,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001498,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.619,
+   "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": [
+    "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": 2.753,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0152,
+   "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": 0.118,
+   "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/88ffd5d3559fe8ae9a04/record.json b/records/sd1/sweep/88ffd5d3559fe8ae9a04/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..38322574817cbc6906e1352b36132b2387ebb124
--- /dev/null
+++ b/records/sd1/sweep/88ffd5d3559fe8ae9a04/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "88ffd5d3559fe8ae9a04",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:51Z",
+ "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": "input_blocks.11.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.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/b4a8c79b797671f6c402e090d97dfc495f12b0c66ca61d625702e001b2660bbe.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b4a8c79b797671f6c402e090d97dfc495f12b0c66ca61d625702e001b2660bbe",
+  "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": "f3134eac072548933b3e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/88ffd5d3559fe8ae9a04/workflow.json b/records/sd1/sweep/88ffd5d3559fe8ae9a04/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e3986354eba5951cf992adcc2e2c43e36de7765
--- /dev/null
+++ b/records/sd1/sweep/88ffd5d3559fe8ae9a04/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 88ffd5d3559fe8ae9a04)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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": [
+    "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/891249fc1b12f5621c4d/interpretations.jsonl b/records/sd1/sweep/891249fc1b12f5621c4d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..eadbd51602e47f5c936548d1c63f91ef3e56a5a6
--- /dev/null
+++ b/records/sd1/sweep/891249fc1b12f5621c4d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "891249fc1b12f5621c4d"}, "kind": "caption", "value": "Frontal portrait of woman with tousled mid-length brown hair on plain beige background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6f98f0ddca0b3d12"}
+{"target": {"record": "891249fc1b12f5621c4d"}, "kind": "change", "value": "The frame disappears and the image becomes a soft, warm beige studio portrait of a frontal woman with tousled longer hair and a dark collared shirt.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6ec02127f616c707"}
+{"target": {"record": "891249fc1b12f5621c4d"}, "kind": "keywords", "value": ["beige background", "borderless", "tousled hair", "soft studio portrait", "frontal pose"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8b543565aa66d3f9"}
+{"target": {"record": "891249fc1b12f5621c4d"}, "kind": "effect_tags", "value": ["new_composition", "colour_shift", "framing", "bright", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "044be4a3222c9e88"}
diff --git a/records/sd1/sweep/891249fc1b12f5621c4d/measurements.json b/records/sd1/sweep/891249fc1b12f5621c4d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3520489f886de105352982e01bb890761276cc80
--- /dev/null
+++ b/records/sd1/sweep/891249fc1b12f5621c4d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "891249fc1b12f5621c4d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.38461,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.448981,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.364679,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.155488,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.073,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00551,
+   "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": 46.948,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9818,
+   "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": 0.2506,
+   "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/891249fc1b12f5621c4d/record.json b/records/sd1/sweep/891249fc1b12f5621c4d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..80fa53cb625df735cfd726b816823a2403f413f4
--- /dev/null
+++ b/records/sd1/sweep/891249fc1b12f5621c4d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "891249fc1b12f5621c4d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:43Z",
+ "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.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "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/edc27e93ead2cfe3ff77daffc37c90c2029e20af56cc02584687a767274fc866.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "edc27e93ead2cfe3ff77daffc37c90c2029e20af56cc02584687a767274fc866",
+  "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": "b713d4c5d50dc151d650"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/891249fc1b12f5621c4d/workflow.json b/records/sd1/sweep/891249fc1b12f5621c4d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea18e2a2f9210f080751d2a7a42d30af176cce14
--- /dev/null
+++ b/records/sd1/sweep/891249fc1b12f5621c4d/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 891249fc1b12f5621c4d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.transformer_blocks.0.attn1.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": [
+    "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/891810cbaf1c9b4a6cf7/measurements.json b/records/sd1/sweep/891810cbaf1c9b4a6cf7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..13581607b71af216c4a75a18e0b518eb33abbddb
--- /dev/null
+++ b/records/sd1/sweep/891810cbaf1c9b4a6cf7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "891810cbaf1c9b4a6cf7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.32317,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.352749,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.166078,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.119957,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.199,
+   "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": 36.69,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9794,
+   "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": 0.227,
+   "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/891810cbaf1c9b4a6cf7/record.json b/records/sd1/sweep/891810cbaf1c9b4a6cf7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2b6d12df1fcc88b4f18a37fe78a91262b83b8da
--- /dev/null
+++ b/records/sd1/sweep/891810cbaf1c9b4a6cf7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "891810cbaf1c9b4a6cf7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:22Z",
+ "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": "input_blocks.7.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "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/f44ce276d01a80c5b8384958daa2d9d5dcf162422706624176ce7b9ba5993f49.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f44ce276d01a80c5b8384958daa2d9d5dcf162422706624176ce7b9ba5993f49",
+  "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": "7b2d8b7e7da9d11f1140"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/891810cbaf1c9b4a6cf7/workflow.json b/records/sd1/sweep/891810cbaf1c9b4a6cf7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bb143b843c4707d478464a5a0b148d00a60ea8a
--- /dev/null
+++ b/records/sd1/sweep/891810cbaf1c9b4a6cf7/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 891810cbaf1c9b4a6cf7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.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": [
+    "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/892e5218fd08083aaaf6/measurements.json b/records/sd1/sweep/892e5218fd08083aaaf6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..14042c77e508ab2d36aa72597bbb96b48091bd7e
--- /dev/null
+++ b/records/sd1/sweep/892e5218fd08083aaaf6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "892e5218fd08083aaaf6",
+ "values": {
+  "std": {
+   "value": 50.674,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01566,
+   "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": 14.318,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.149314,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.212741,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.179633,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.030622,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9842,
+   "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": 0.1126,
+   "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/892e5218fd08083aaaf6/record.json b/records/sd1/sweep/892e5218fd08083aaaf6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cc5252967f19a2c7eecb0e0858f4b373aaaf432
--- /dev/null
+++ b/records/sd1/sweep/892e5218fd08083aaaf6/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "892e5218fd08083aaaf6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:04:58Z",
+ "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 still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.mid",
+   "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.mid",
+   "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/892e5218fd08083aaaf6/workflow.json b/records/sd1/sweep/892e5218fd08083aaaf6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a33a096bdd03230a74fbf8fec3cbba26781f986b
--- /dev/null
+++ b/records/sd1/sweep/892e5218fd08083aaaf6/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 892e5218fd08083aaaf6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\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/89624e7064bb5f821740/measurements.json b/records/sd1/sweep/89624e7064bb5f821740/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c5f0838a1a95b0b34735b856ff5749dd43f872c
--- /dev/null
+++ b/records/sd1/sweep/89624e7064bb5f821740/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "89624e7064bb5f821740",
+ "values": {
+  "cosine_distance": {
+   "value": 0.241909,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.296999,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.23947,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.077108,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.304,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0048,
+   "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": 25.136,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0042,
+   "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": 0.2449,
+   "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/89624e7064bb5f821740/record.json b/records/sd1/sweep/89624e7064bb5f821740/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d651fab8521e06d9121ce19b131ba4c9c6412f2
--- /dev/null
+++ b/records/sd1/sweep/89624e7064bb5f821740/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "89624e7064bb5f821740",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:34Z",
+ "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": 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.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "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/65c3e2e02d1164f774a2f12244bf2177e7df06c041440e6824f1e6d888de0c94.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "65c3e2e02d1164f774a2f12244bf2177e7df06c041440e6824f1e6d888de0c94",
+  "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": "e5b4b42d4e5f532c3a30"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89624e7064bb5f821740/workflow.json b/records/sd1/sweep/89624e7064bb5f821740/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c315a8a73aa89223e82152260d7eac6aede1685
--- /dev/null
+++ b/records/sd1/sweep/89624e7064bb5f821740/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 89624e7064bb5f821740)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\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/8964bfaa626ff630b90c/measurements.json b/records/sd1/sweep/8964bfaa626ff630b90c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e51edf2a5628adadd966fa960b6c199c95c461e
--- /dev/null
+++ b/records/sd1/sweep/8964bfaa626ff630b90c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8964bfaa626ff630b90c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.306973,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.549661,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.397966,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.324441,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.48,
+   "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": [
+    "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": 31.827,
+   "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": 0.8673,
+   "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/8964bfaa626ff630b90c/record.json b/records/sd1/sweep/8964bfaa626ff630b90c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf5fc0f706fd8c267287dc253b01fb1421060c07
--- /dev/null
+++ b/records/sd1/sweep/8964bfaa626ff630b90c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8964bfaa626ff630b90c",
+ "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.out_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "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/f2b8877ed33716e194031fa6263544c777b977f163969797da0ff5cf50979828.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f2b8877ed33716e194031fa6263544c777b977f163969797da0ff5cf50979828",
+  "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": "8711923e776b0e62c498"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8964bfaa626ff630b90c/workflow.json b/records/sd1/sweep/8964bfaa626ff630b90c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2a7793dc01444fbe4ee8bd187e099c72aaf683b
--- /dev/null
+++ b/records/sd1/sweep/8964bfaa626ff630b90c/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 8964bfaa626ff630b90c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.out_layers.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/89693b7d6a9a5c13e436/measurements.json b/records/sd1/sweep/89693b7d6a9a5c13e436/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d3b27eb791afb4d12d7b99707ed8567dd31b7c8
--- /dev/null
+++ b/records/sd1/sweep/89693b7d6a9a5c13e436/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "89693b7d6a9a5c13e436",
+ "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/89693b7d6a9a5c13e436/record.json b/records/sd1/sweep/89693b7d6a9a5c13e436/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8bb3100dbf65622967fe2c0a85efc07d793a031b
--- /dev/null
+++ b/records/sd1/sweep/89693b7d6a9a5c13e436/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "89693b7d6a9a5c13e436",
+ "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": 90
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "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/99ed29d1161b137008a2a01fc0d9f1886924c8e992a9c09a5f7d6a0702cab8ce.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "99ed29d1161b137008a2a01fc0d9f1886924c8e992a9c09a5f7d6a0702cab8ce",
+  "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": "df4726ca0e93286713d4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89693b7d6a9a5c13e436/workflow.json b/records/sd1/sweep/89693b7d6a9a5c13e436/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..232bd6cb52b7618899d6c6f866450287e6c9f5bb
--- /dev/null
+++ b/records/sd1/sweep/89693b7d6a9a5c13e436/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 89693b7d6a9a5c13e436)"
+  },
+  "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\": 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/896c460d6bfcb8f9b61f/measurements.json b/records/sd1/sweep/896c460d6bfcb8f9b61f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0be4787dbb4ef6e0955d79ae979bf8f9a78fc933
--- /dev/null
+++ b/records/sd1/sweep/896c460d6bfcb8f9b61f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "896c460d6bfcb8f9b61f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.293208,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.516291,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.350766,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.214433,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 31.46,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00311,
+   "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": 37.243,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0514,
+   "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": 0.4949,
+   "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/896c460d6bfcb8f9b61f/record.json b/records/sd1/sweep/896c460d6bfcb8f9b61f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7429fd31dd7a3f288fcc98e67ebd0ad4605a89a
--- /dev/null
+++ b/records/sd1/sweep/896c460d6bfcb8f9b61f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "896c460d6bfcb8f9b61f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:44Z",
+ "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.4.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "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/b09af194fc6bd5f30c3a2fdec25895fd178a67e37e0abd4a08e3bfdc2faf5a14.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b09af194fc6bd5f30c3a2fdec25895fd178a67e37e0abd4a08e3bfdc2faf5a14",
+  "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": "94311737ebfbd28cdd17"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/896c460d6bfcb8f9b61f/workflow.json b/records/sd1/sweep/896c460d6bfcb8f9b61f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..53b5efc6686b18411b4e5f2be306fc3d38191d9b
--- /dev/null
+++ b/records/sd1/sweep/896c460d6bfcb8f9b61f/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 896c460d6bfcb8f9b61f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.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/8973e3643a1d3bc40c9c/measurements.json b/records/sd1/sweep/8973e3643a1d3bc40c9c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7cea91dddc38bbbc5bc1f1345486247f69353afe
--- /dev/null
+++ b/records/sd1/sweep/8973e3643a1d3bc40c9c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8973e3643a1d3bc40c9c",
+ "values": {
+  "std": {
+   "value": 36.031,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01265,
+   "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": 13.983,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.109303,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.193204,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.036162,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.011118,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0273,
+   "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/8973e3643a1d3bc40c9c/record.json b/records/sd1/sweep/8973e3643a1d3bc40c9c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..457b7c95506217942a575de61db5b6093a334ff9
--- /dev/null
+++ b/records/sd1/sweep/8973e3643a1d3bc40c9c/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8973e3643a1d3bc40c9c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:06:17Z",
+ "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": 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": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.lo",
+   "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": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "5f5f3a3a61d64c0d6bc4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8973e3643a1d3bc40c9c/workflow.json b/records/sd1/sweep/8973e3643a1d3bc40c9c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e833f92476a9f33054d0099ce1e50da44b48780
--- /dev/null
+++ b/records/sd1/sweep/8973e3643a1d3bc40c9c/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 8973e3643a1d3bc40c9c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\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": 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/8979f0a07c5c912d39f2/interpretations.jsonl b/records/sd1/sweep/8979f0a07c5c912d39f2/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2187007eb4ca8c4fe42f6d5f968f8b5585abc832
--- /dev/null
+++ b/records/sd1/sweep/8979f0a07c5c912d39f2/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8979f0a07c5c912d39f2"}, "kind": "caption", "value": "Neon night street with orange and red signs, darker blue sky and wet reflections", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:43:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "99c4a935f3641d92"}
+{"target": {"record": "8979f0a07c5c912d39f2"}, "kind": "change", "value": "Same composition with slightly redrawn signs, a darker and bluer cast, and a softer, more painterly finish.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:43:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6f41f5c1afb6b169"}
+{"target": {"record": "8979f0a07c5c912d39f2"}, "kind": "keywords", "value": ["neon street", "night city", "blue tones", "reflections"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:43:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9d5c0508f9326139"}
+{"target": {"record": "8979f0a07c5c912d39f2"}, "kind": "effect_tags", "value": ["subject_kept", "subtle", "dark", "painterly"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:43:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fe7c16c7c2979aae"}
diff --git a/records/sd1/sweep/8979f0a07c5c912d39f2/measurements.json b/records/sd1/sweep/8979f0a07c5c912d39f2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..943ed653732917a4cbf4af0cd202996ab042cae7
--- /dev/null
+++ b/records/sd1/sweep/8979f0a07c5c912d39f2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8979f0a07c5c912d39f2",
+ "values": {
+  "std": {
+   "value": 33.322,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01399,
+   "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": 12.356,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.040735,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.161127,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.093732,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036903,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9414,
+   "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": 0.552,
+   "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/8979f0a07c5c912d39f2/record.json b/records/sd1/sweep/8979f0a07c5c912d39f2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1950ef916f80060054b7a997d78d7efe1c15236b
--- /dev/null
+++ b/records/sd1/sweep/8979f0a07c5c912d39f2/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8979f0a07c5c912d39f2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:13:09Z",
+ "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": 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.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "boost",
+   "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": [
+   "in.hi",
+   "multiply",
+   2.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 2.0,
+    "applied": 2.0,
+    "reason": "outside safe range [0.6, 1.6] 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": "1bd07440062ca2280d5e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8979f0a07c5c912d39f2/workflow.json b/records/sd1/sweep/8979f0a07c5c912d39f2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6ebedad12e82e164183e617796893c79cfd5425
--- /dev/null
+++ b/records/sd1/sweep/8979f0a07c5c912d39f2/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 8979f0a07c5c912d39f2)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\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 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/897e2f9e6b07656076d2/measurements.json b/records/sd1/sweep/897e2f9e6b07656076d2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf6519bb00d325bd755ea5b710f09e5aa0dcbcc2
--- /dev/null
+++ b/records/sd1/sweep/897e2f9e6b07656076d2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "897e2f9e6b07656076d2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.648348,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.549277,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.339359,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.137915,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.051,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00521,
+   "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": 59.306,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0174,
+   "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": 0.5157,
+   "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/897e2f9e6b07656076d2/record.json b/records/sd1/sweep/897e2f9e6b07656076d2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e76b1c682d1baac58fb87b69b46ab3951e68df31
--- /dev/null
+++ b/records/sd1/sweep/897e2f9e6b07656076d2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "897e2f9e6b07656076d2",
+ "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.norm",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "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/bb9bf5a31f755045e2fb365fd1c2389008f2d5b22db27fd02609e3f088751578.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "bb9bf5a31f755045e2fb365fd1c2389008f2d5b22db27fd02609e3f088751578",
+  "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": "908d6fbc593c3e5bc461"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/897e2f9e6b07656076d2/workflow.json b/records/sd1/sweep/897e2f9e6b07656076d2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac067e5daa2c4f01d71f9e79d8ebe6bd238bcac0
--- /dev/null
+++ b/records/sd1/sweep/897e2f9e6b07656076d2/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 897e2f9e6b07656076d2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.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/8986db62456fb893f28a/measurements.json b/records/sd1/sweep/8986db62456fb893f28a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..95ad3dcb99dde95b64e7a12bdb88c0329af9f8f2
--- /dev/null
+++ b/records/sd1/sweep/8986db62456fb893f28a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8986db62456fb893f28a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.333762,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.44247,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.233875,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.078993,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.378,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00468,
+   "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": 36.214,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9815,
+   "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/8986db62456fb893f28a/record.json b/records/sd1/sweep/8986db62456fb893f28a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f025686e276fcad3fc9779cbba01a038bc6e09bf
--- /dev/null
+++ b/records/sd1/sweep/8986db62456fb893f28a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8986db62456fb893f28a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:49Z",
+ "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": "input_blocks.10.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.lo",
+   "kind": "skip",
+   "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/89d2d2b09676e794f0be873491fb5716ce8e2664c5c222587ec6045dbc1f37f1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "89d2d2b09676e794f0be873491fb5716ce8e2664c5c222587ec6045dbc1f37f1",
+  "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": "835c87b9d360f30dad03"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8986db62456fb893f28a/workflow.json b/records/sd1/sweep/8986db62456fb893f28a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c3a39b1de542f376b115522b8f140a7119d8665
--- /dev/null
+++ b/records/sd1/sweep/8986db62456fb893f28a/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 8986db62456fb893f28a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.skip_connection\",\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/898d203411f53341ea8b/measurements.json b/records/sd1/sweep/898d203411f53341ea8b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c33a4c3bec82c7dcce8ebe4a79ac964db30e6589
--- /dev/null
+++ b/records/sd1/sweep/898d203411f53341ea8b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "898d203411f53341ea8b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.544085,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.570147,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.627877,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.319624,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.463,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00459,
+   "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": 48.697,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0055,
+   "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": 0.4661,
+   "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/898d203411f53341ea8b/record.json b/records/sd1/sweep/898d203411f53341ea8b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa4ee9d97ab457d14b3248f0565b4c409ed555bd
--- /dev/null
+++ b/records/sd1/sweep/898d203411f53341ea8b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "898d203411f53341ea8b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:08Z",
+ "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": "output_blocks.4.0.in_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.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/c807b9ab9759a498b2197cabf758536bc9f3af81379268303d84b28a456acee2.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c807b9ab9759a498b2197cabf758536bc9f3af81379268303d84b28a456acee2",
+  "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": "0ae7587d863214486989"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/898d203411f53341ea8b/workflow.json b/records/sd1/sweep/898d203411f53341ea8b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..15db62f14a1e1529d7fde4732424dd0901ed5932
--- /dev/null
+++ b/records/sd1/sweep/898d203411f53341ea8b/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 898d203411f53341ea8b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.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/899375abe7d9988e24d6/measurements.json b/records/sd1/sweep/899375abe7d9988e24d6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..27a6d4c15bfbcf0dec5ea0715727e8630b16146a
--- /dev/null
+++ b/records/sd1/sweep/899375abe7d9988e24d6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "899375abe7d9988e24d6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.06585,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.053009,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.032951,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044172,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.93,
+   "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": 10.006,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0197,
+   "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": 0.1783,
+   "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/899375abe7d9988e24d6/record.json b/records/sd1/sweep/899375abe7d9988e24d6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b758cac67b21525179433933d4e4ec7771992e86
--- /dev/null
+++ b/records/sd1/sweep/899375abe7d9988e24d6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "899375abe7d9988e24d6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:57Z",
+ "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.10.1.transformer_blocks.0.attn2.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.hi",
+   "kind": "cross_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/d28a992bfd37b211d465917c19b9dcce4065a55f4974079c4db47c4ee041fecf.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d28a992bfd37b211d465917c19b9dcce4065a55f4974079c4db47c4ee041fecf",
+  "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": "881f4400e73275494853"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/899375abe7d9988e24d6/workflow.json b/records/sd1/sweep/899375abe7d9988e24d6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..20826278c4c3fc71febf42ea30bb6ee7088f3c00
--- /dev/null
+++ b/records/sd1/sweep/899375abe7d9988e24d6/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 899375abe7d9988e24d6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.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": [
+    "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/89944b7db24aaa40f68c/measurements.json b/records/sd1/sweep/89944b7db24aaa40f68c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b236cf36964d3f246f26742e3a8645a42375492c
--- /dev/null
+++ b/records/sd1/sweep/89944b7db24aaa40f68c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "89944b7db24aaa40f68c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.18257,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.284903,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.152341,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036584,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.218,
+   "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": 25.771,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9847,
+   "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": 0.0498,
+   "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/89944b7db24aaa40f68c/record.json b/records/sd1/sweep/89944b7db24aaa40f68c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..51ec0bb0024b4abd0725e1520bc0cf428dfdbf6d
--- /dev/null
+++ b/records/sd1/sweep/89944b7db24aaa40f68c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "89944b7db24aaa40f68c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:15Z",
+ "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": "input_blocks.2.0.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": "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/6e4caad40ee0278fceb541f3c6f0f73e66e2575ea6789c421967c2bc077f746e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6e4caad40ee0278fceb541f3c6f0f73e66e2575ea6789c421967c2bc077f746e",
+  "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": "45bd1bc6c98204745489"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89944b7db24aaa40f68c/workflow.json b/records/sd1/sweep/89944b7db24aaa40f68c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bb5da962f0ffb4d5173318f162a0c722b35424f
--- /dev/null
+++ b/records/sd1/sweep/89944b7db24aaa40f68c/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 89944b7db24aaa40f68c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.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": [
+    "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/899a0744f3d43036d2f5/measurements.json b/records/sd1/sweep/899a0744f3d43036d2f5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..84d71abd9833950ede7ad46ecd8967d7c48e297c
--- /dev/null
+++ b/records/sd1/sweep/899a0744f3d43036d2f5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "899a0744f3d43036d2f5",
+ "values": {
+  "std": {
+   "value": 45.076,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01606,
+   "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": 16.675,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.124434,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.228024,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.081094,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.038641,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.011,
+   "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/899a0744f3d43036d2f5/record.json b/records/sd1/sweep/899a0744f3d43036d2f5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1e458b76db5c20a44c62b1f5a97843135b7a896
--- /dev/null
+++ b/records/sd1/sweep/899a0744f3d43036d2f5/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "899a0744f3d43036d2f5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:48:24Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "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": "a77bd19c2cc87d81e6d0",
+  "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/899a0744f3d43036d2f5/workflow.json b/records/sd1/sweep/899a0744f3d43036d2f5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2ba89674f9ba645a695ce07019d04e3d12120cb
--- /dev/null
+++ b/records/sd1/sweep/899a0744f3d43036d2f5/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 899a0744f3d43036d2f5)"
+  },
+  "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 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/89c1dc84c145b74f0628/measurements.json b/records/sd1/sweep/89c1dc84c145b74f0628/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0affe377979993552782191da4e27cf84de01ab9
--- /dev/null
+++ b/records/sd1/sweep/89c1dc84c145b74f0628/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "89c1dc84c145b74f0628",
+ "values": {
+  "cosine_distance": {
+   "value": 0.076123,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.121322,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.045664,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02825,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.674,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00741,
+   "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": 13.308,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9721,
+   "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": 0.0808,
+   "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/89c1dc84c145b74f0628/record.json b/records/sd1/sweep/89c1dc84c145b74f0628/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9307fa012fd339925c1c1cbc66f662a3faaf07d
--- /dev/null
+++ b/records/sd1/sweep/89c1dc84c145b74f0628/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "89c1dc84c145b74f0628",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:08Z",
+ "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_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.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/4ec1fb1230bc71f56df982229dd002b5a11f1d641c294d170a5ccfdbc62e802c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4ec1fb1230bc71f56df982229dd002b5a11f1d641c294d170a5ccfdbc62e802c",
+  "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": "095a9384144d7e5c7816"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89c1dc84c145b74f0628/workflow.json b/records/sd1/sweep/89c1dc84c145b74f0628/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5fa55e9a28e3a11296f5315d8f6b0f87aed186c
--- /dev/null
+++ b/records/sd1/sweep/89c1dc84c145b74f0628/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 89c1dc84c145b74f0628)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.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/89c4d506d1a0c5caa6e8/measurements.json b/records/sd1/sweep/89c4d506d1a0c5caa6e8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..680ab64844d536aa52509ef094fa5f2bda93e6bb
--- /dev/null
+++ b/records/sd1/sweep/89c4d506d1a0c5caa6e8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "89c4d506d1a0c5caa6e8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.259543,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.499485,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.574006,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.348321,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.128,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00454,
+   "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": 31.204,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.979,
+   "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": 0.8767,
+   "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/89c4d506d1a0c5caa6e8/record.json b/records/sd1/sweep/89c4d506d1a0c5caa6e8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebbdede6f06bf25762b36114205cdec3029a7012
--- /dev/null
+++ b/records/sd1/sweep/89c4d506d1a0c5caa6e8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "89c4d506d1a0c5caa6e8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:07Z",
+ "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.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.mid",
+   "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/abf865ebb8609e666dd768aedee370bc98307139bfd6b46652428b64bcdd1ec2.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "abf865ebb8609e666dd768aedee370bc98307139bfd6b46652428b64bcdd1ec2",
+  "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": "0168960f3d72fe0a7167"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89c4d506d1a0c5caa6e8/workflow.json b/records/sd1/sweep/89c4d506d1a0c5caa6e8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..272795cb0fc8ca8c5eced79a90dae7300bed50ee
--- /dev/null
+++ b/records/sd1/sweep/89c4d506d1a0c5caa6e8/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 89c4d506d1a0c5caa6e8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_layers.3\",\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/89ca45340fdcec4c0268/measurements.json b/records/sd1/sweep/89ca45340fdcec4c0268/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..23fdc4e19a1fa996b654a087eadfab62776d62ef
--- /dev/null
+++ b/records/sd1/sweep/89ca45340fdcec4c0268/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "89ca45340fdcec4c0268",
+ "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/89ca45340fdcec4c0268/record.json b/records/sd1/sweep/89ca45340fdcec4c0268/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e8a7a55ba99da3daed14d9be1161746300f7364
--- /dev/null
+++ b/records/sd1/sweep/89ca45340fdcec4c0268/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "89ca45340fdcec4c0268",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:38Z",
+ "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.3.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "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/0d48552f254fd687c6e865337bf2c1ca3a051f34027e57fb86327f9953d57af0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0d48552f254fd687c6e865337bf2c1ca3a051f34027e57fb86327f9953d57af0",
+  "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": "2454ce3a9ade7dff3b3a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89ca45340fdcec4c0268/workflow.json b/records/sd1/sweep/89ca45340fdcec4c0268/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbd7eb85255d5db78d695b9c43adbc4f92312563
--- /dev/null
+++ b/records/sd1/sweep/89ca45340fdcec4c0268/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 89ca45340fdcec4c0268)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.h_upd\",\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/89de5fa728f4b13d4cf4/interpretations.jsonl b/records/sd1/sweep/89de5fa728f4b13d4cf4/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..44df73eaf3afbdd3e4ba45966b437853f1652108
--- /dev/null
+++ b/records/sd1/sweep/89de5fa728f4b13d4cf4/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "89de5fa728f4b13d4cf4"}, "kind": "caption", "value": "Red-haired woman with long hair against painterly teal and peach brushstroke backdrop", "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": "60c9bb9536408b3a"}
+{"target": {"record": "89de5fa728f4b13d4cf4"}, "kind": "change", "value": "The subject becomes a different woman with long auburn hair and warmer skin, and the plain grey background turns into a loose teal-and-salmon painted wall with torn-paper edges.", "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": "3da80199a3e802af"}
+{"target": {"record": "89de5fa728f4b13d4cf4"}, "kind": "keywords", "value": ["red hair", "painted background", "teal and peach", "new person", "warm palette", "mixed media"], "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": "21188af3a2f3594b"}
+{"target": {"record": "89de5fa728f4b13d4cf4"}, "kind": "effect_tags", "value": ["colour_shift", "painterly", "style_change", "new_composition", "saturated"], "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": "6324e56d697a142f"}
diff --git a/records/sd1/sweep/89de5fa728f4b13d4cf4/measurements.json b/records/sd1/sweep/89de5fa728f4b13d4cf4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7c9d06248a0e5aebc2ede87d4aebf4638ab0c56
--- /dev/null
+++ b/records/sd1/sweep/89de5fa728f4b13d4cf4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "89de5fa728f4b13d4cf4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.272482,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.417035,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.296273,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.141593,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.078,
+   "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": 28.563,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9788,
+   "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/89de5fa728f4b13d4cf4/record.json b/records/sd1/sweep/89de5fa728f4b13d4cf4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6654735a3f34f3c5d1cb9d96d13321ebd1f7f08d
--- /dev/null
+++ b/records/sd1/sweep/89de5fa728f4b13d4cf4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "89de5fa728f4b13d4cf4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:36Z",
+ "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.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "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/f89aad4c30f19d9677ab756acdf61638c2902d4ed8cc8f87f867c71dee92dab1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f89aad4c30f19d9677ab756acdf61638c2902d4ed8cc8f87f867c71dee92dab1",
+  "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": "340e64f571dbe7e0cd67"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89de5fa728f4b13d4cf4/workflow.json b/records/sd1/sweep/89de5fa728f4b13d4cf4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a20141f85d5b1cc50e64eef4cfc38b194e156bc8
--- /dev/null
+++ b/records/sd1/sweep/89de5fa728f4b13d4cf4/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 89de5fa728f4b13d4cf4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/89e0b8291be1f783101f/measurements.json b/records/sd1/sweep/89e0b8291be1f783101f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a3c7f5d9e2acc9159a38363b0c380440f1a3b44
--- /dev/null
+++ b/records/sd1/sweep/89e0b8291be1f783101f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "89e0b8291be1f783101f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.354522,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.397408,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.159601,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.069658,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.749,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00356,
+   "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": 36.724,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9531,
+   "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": 0.3734,
+   "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/89e0b8291be1f783101f/record.json b/records/sd1/sweep/89e0b8291be1f783101f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..de2630e84a3d481e070f226aae2d8c9e8a983716
--- /dev/null
+++ b/records/sd1/sweep/89e0b8291be1f783101f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "89e0b8291be1f783101f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:30Z",
+ "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_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "kind": "self_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/e7da359c3a81eb2bbebdb63b61f2b341f7f4508abe75ba0fe65bfe714ecec2ac.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e7da359c3a81eb2bbebdb63b61f2b341f7f4508abe75ba0fe65bfe714ecec2ac",
+  "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": "09e8ec649582411509d4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89e0b8291be1f783101f/workflow.json b/records/sd1/sweep/89e0b8291be1f783101f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..244e1ba3b4e071180d472d67c9f611395047db3e
--- /dev/null
+++ b/records/sd1/sweep/89e0b8291be1f783101f/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 89e0b8291be1f783101f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.attn1.to_v\",\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/89ea870b92df6b1f7741/interpretations.jsonl b/records/sd1/sweep/89ea870b92df6b1f7741/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c883baacd7dedd5e558ac126e9fcaa6a6c0c1583
--- /dev/null
+++ b/records/sd1/sweep/89ea870b92df6b1f7741/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "89ea870b92df6b1f7741"}, "kind": "caption", "value": "Close frontal portrait of woman with tied-back hair, teal sweater, greenish backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "83a8213bf541181e"}
+{"target": {"record": "89ea870b92df6b1f7741"}, "kind": "change", "value": "The face is enlarged and frontal with a different identity, ponytail and teal sweater, and the backdrop shifts to a greener tone with a soft border.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9be412d7cbe79c29"}
+{"target": {"record": "89ea870b92df6b1f7741"}, "kind": "keywords", "value": ["closer crop", "frontal portrait", "teal sweater", "green backdrop", "identity change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c1a3805f73cc69ee"}
+{"target": {"record": "89ea870b92df6b1f7741"}, "kind": "effect_tags", "value": ["zoom", "subject_kept", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a671fbab74d52986"}
diff --git a/records/sd1/sweep/89ea870b92df6b1f7741/measurements.json b/records/sd1/sweep/89ea870b92df6b1f7741/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..771b0db46714c3e0484c1569cbd1ceef47a74f89
--- /dev/null
+++ b/records/sd1/sweep/89ea870b92df6b1f7741/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "89ea870b92df6b1f7741",
+ "values": {
+  "cosine_distance": {
+   "value": 0.261367,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.321832,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.160609,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.091275,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.831,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00544,
+   "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": 27.314,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0106,
+   "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": 0.1602,
+   "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/89ea870b92df6b1f7741/record.json b/records/sd1/sweep/89ea870b92df6b1f7741/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5bf88a1630f2625eb429d8af170129de21d97ed
--- /dev/null
+++ b/records/sd1/sweep/89ea870b92df6b1f7741/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "89ea870b92df6b1f7741",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:54Z",
+ "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": "middle_block.0.out_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "mid",
+   "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/d7f7a9847718c45a0dff7b16d25904926bab5d6f19b39efc6c784abb52b8738f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d7f7a9847718c45a0dff7b16d25904926bab5d6f19b39efc6c784abb52b8738f",
+  "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"
+    ],
+    "previous_id": "68edba869486c5f9b32b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89ea870b92df6b1f7741/workflow.json b/records/sd1/sweep/89ea870b92df6b1f7741/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9dd5beb4257f7f86e88bf97d28c7a8fcb1f4273f
--- /dev/null
+++ b/records/sd1/sweep/89ea870b92df6b1f7741/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 89ea870b92df6b1f7741)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.out_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/89f7a17efde591ffd3da/interpretations.jsonl b/records/sd1/sweep/89f7a17efde591ffd3da/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..137e0ad889fa7e1086bd24ce2007f22d71484f02
--- /dev/null
+++ b/records/sd1/sweep/89f7a17efde591ffd3da/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "89f7a17efde591ffd3da"}, "kind": "caption", "value": "Woman with auburn bob and dark blazer, head slightly tilted, grey-green aged print", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b2e6700bb0fc27dd"}
+{"target": {"record": "89f7a17efde591ffd3da"}, "kind": "change", "value": "The woman becomes a different person with a warmer auburn bob and tilted head, set against a slightly greener backdrop with worn print edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4c841fb4dd9ada28"}
+{"target": {"record": "89f7a17efde591ffd3da"}, "kind": "keywords", "value": ["auburn bob", "head tilt", "aged print", "dark blazer", "identity change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1f4517f0aed2a8a0"}
+{"target": {"record": "89f7a17efde591ffd3da"}, "kind": "effect_tags", "value": ["colour_shift", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7aa5b9e2d60b1549"}
diff --git a/records/sd1/sweep/89f7a17efde591ffd3da/measurements.json b/records/sd1/sweep/89f7a17efde591ffd3da/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2d33175c2c8416b1babf65cf327935ec845f70
--- /dev/null
+++ b/records/sd1/sweep/89f7a17efde591ffd3da/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "89f7a17efde591ffd3da",
+ "values": {
+  "cosine_distance": {
+   "value": 0.309281,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.362365,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.226791,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0779,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.61,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00747,
+   "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": 37.423,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0251,
+   "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": 0.18,
+   "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/89f7a17efde591ffd3da/record.json b/records/sd1/sweep/89f7a17efde591ffd3da/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..13837bd9199a9b947c7c341fb843cc8dd4f6c6b1
--- /dev/null
+++ b/records/sd1/sweep/89f7a17efde591ffd3da/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "89f7a17efde591ffd3da",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:14Z",
+ "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": "input_blocks.4.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e66bd807750caf222e1421a09bd49cb86417dcdcdb49731d72973c5b171335d0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e66bd807750caf222e1421a09bd49cb86417dcdcdb49731d72973c5b171335d0",
+  "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": "e151d7ca72766d9d1ff4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89f7a17efde591ffd3da/workflow.json b/records/sd1/sweep/89f7a17efde591ffd3da/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e96419a055fac772a778f427b1fd50780dc21123
--- /dev/null
+++ b/records/sd1/sweep/89f7a17efde591ffd3da/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 89f7a17efde591ffd3da)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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": [
+    "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/89f92f3ab045fed749d4/measurements.json b/records/sd1/sweep/89f92f3ab045fed749d4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..89fcbee590e7c444456dfac0c617ca9e98adc8b6
--- /dev/null
+++ b/records/sd1/sweep/89f92f3ab045fed749d4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "89f92f3ab045fed749d4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.098938,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.156014,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.089725,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040117,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.737,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0029,
+   "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": 14.464,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9386,
+   "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": 0.3128,
+   "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/89f92f3ab045fed749d4/record.json b/records/sd1/sweep/89f92f3ab045fed749d4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1141a6146fec27ecc03bf3368d84224532d155f8
--- /dev/null
+++ b/records/sd1/sweep/89f92f3ab045fed749d4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "89f92f3ab045fed749d4",
+ "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.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "ff",
+   "module_type": "Linear",
+   "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/3b41cfbbe0d8eb60586ecf4fba20e7589c2a2e6f85f5708291bf87695b711da8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3b41cfbbe0d8eb60586ecf4fba20e7589c2a2e6f85f5708291bf87695b711da8",
+  "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": "073ee4cffaef5bf7f166"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/89f92f3ab045fed749d4/workflow.json b/records/sd1/sweep/89f92f3ab045fed749d4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..85b683a46fe40085e4d5a410bc5e5f0bffa95b69
--- /dev/null
+++ b/records/sd1/sweep/89f92f3ab045fed749d4/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 89f92f3ab045fed749d4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.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": [
+    "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/8a0167f542a990b6c98f/measurements.json b/records/sd1/sweep/8a0167f542a990b6c98f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ded595beff74f21f5553345ac710c415f0d6b04
--- /dev/null
+++ b/records/sd1/sweep/8a0167f542a990b6c98f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a0167f542a990b6c98f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.251067,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.405734,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.133812,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.104611,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.355,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00364,
+   "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": 31.373,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0151,
+   "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": 0.6515,
+   "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/8a0167f542a990b6c98f/record.json b/records/sd1/sweep/8a0167f542a990b6c98f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..99ac82644a14db8863649dd43209405275752a89
--- /dev/null
+++ b/records/sd1/sweep/8a0167f542a990b6c98f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a0167f542a990b6c98f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:06Z",
+ "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.2.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/5df8c7819012e375515fbf6b876a3685a5949c7ba491b79ce30ec85113f477d3.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5df8c7819012e375515fbf6b876a3685a5949c7ba491b79ce30ec85113f477d3",
+  "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": "ae28c567163590725c37"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a0167f542a990b6c98f/workflow.json b/records/sd1/sweep/8a0167f542a990b6c98f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c3ec0ab2514389ac312f505d75715dbcf85f39bd
--- /dev/null
+++ b/records/sd1/sweep/8a0167f542a990b6c98f/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 8a0167f542a990b6c98f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.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/8a036491ea0eebdf0920/measurements.json b/records/sd1/sweep/8a036491ea0eebdf0920/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3597e36df65049430fd2e516c54c5589d33edbc4
--- /dev/null
+++ b/records/sd1/sweep/8a036491ea0eebdf0920/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a036491ea0eebdf0920",
+ "values": {
+  "cosine_distance": {
+   "value": 0.134632,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.149742,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.086002,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.12077,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.27,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00522,
+   "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": 16.498,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9257,
+   "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": 0.338,
+   "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/8a036491ea0eebdf0920/record.json b/records/sd1/sweep/8a036491ea0eebdf0920/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e645efdc3957d8c4535a550d09e40c91506ae19
--- /dev/null
+++ b/records/sd1/sweep/8a036491ea0eebdf0920/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a036491ea0eebdf0920",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:34Z",
+ "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.11.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "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/6ec49ff397dbff0d642c9d76c889dda692a0f2d282ea76cd218414695cf4172d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6ec49ff397dbff0d642c9d76c889dda692a0f2d282ea76cd218414695cf4172d",
+  "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": "494cb6d87bc822633f52"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a036491ea0eebdf0920/workflow.json b/records/sd1/sweep/8a036491ea0eebdf0920/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..70d84f0f5eb2182c57349d0a78a846e41978d92e
--- /dev/null
+++ b/records/sd1/sweep/8a036491ea0eebdf0920/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 8a036491ea0eebdf0920)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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": [
+    "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/8a06acfaacd0fd4f6d4d/measurements.json b/records/sd1/sweep/8a06acfaacd0fd4f6d4d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dad8a3e80dfa203adb3fe34b01edf9ecb5741d89
--- /dev/null
+++ b/records/sd1/sweep/8a06acfaacd0fd4f6d4d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a06acfaacd0fd4f6d4d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.148172,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.235363,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.078064,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040863,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.244,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00654,
+   "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": 21.67,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9497,
+   "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": 0.0838,
+   "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/8a06acfaacd0fd4f6d4d/record.json b/records/sd1/sweep/8a06acfaacd0fd4f6d4d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2fb5dd5c5214fa647f063648ccd1119e177b0daf
--- /dev/null
+++ b/records/sd1/sweep/8a06acfaacd0fd4f6d4d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a06acfaacd0fd4f6d4d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:36Z",
+ "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.6.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "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/34f5468deea25ae07dfa6d5ec5bbc632105d7fe944c08d7ecbe285301e3f515f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "34f5468deea25ae07dfa6d5ec5bbc632105d7fe944c08d7ecbe285301e3f515f",
+  "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": "cd02f5e2525b62abd29c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a06acfaacd0fd4f6d4d/workflow.json b/records/sd1/sweep/8a06acfaacd0fd4f6d4d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed4f738ef9a82a230a7eaaeb76e92445ab0fc3a2
--- /dev/null
+++ b/records/sd1/sweep/8a06acfaacd0fd4f6d4d/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 8a06acfaacd0fd4f6d4d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.transformer_blocks.0.attn1.to_k\",\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/8a0d41160d867b0d176c/measurements.json b/records/sd1/sweep/8a0d41160d867b0d176c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..af0f29b999b8c5cf97ce5bd38a983ce04b26196f
--- /dev/null
+++ b/records/sd1/sweep/8a0d41160d867b0d176c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a0d41160d867b0d176c",
+ "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/8a0d41160d867b0d176c/record.json b/records/sd1/sweep/8a0d41160d867b0d176c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b086bd5ddadc392f700cc3d257d864ab533167b
--- /dev/null
+++ b/records/sd1/sweep/8a0d41160d867b0d176c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a0d41160d867b0d176c",
+ "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": 1.5
+   },
+   "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/b965aa1042d1f3327cb98d4a30496f97bd75e0bed903b29a1285e103fe2c8772.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b965aa1042d1f3327cb98d4a30496f97bd75e0bed903b29a1285e103fe2c8772",
+  "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": "6158b3d0215fba5da0b7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a0d41160d867b0d176c/workflow.json b/records/sd1/sweep/8a0d41160d867b0d176c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b32d0298ce0dd9123d705665d3bfb1be056522bf
--- /dev/null
+++ b/records/sd1/sweep/8a0d41160d867b0d176c/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 8a0d41160d867b0d176c)"
+  },
+  "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\": 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/8a11ba33af3225e2d241/interpretations.jsonl b/records/sd1/sweep/8a11ba33af3225e2d241/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..8da27a68bfe601475e67c29171fa10bed6b0ef1b
--- /dev/null
+++ b/records/sd1/sweep/8a11ba33af3225e2d241/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8a11ba33af3225e2d241"}, "kind": "caption", "value": "Abstract teal, green and amber brush texture with dark shapes, no subject", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3bf0f6b60ed070a0"}
+{"target": {"record": "8a11ba33af3225e2d241"}, "kind": "change", "value": "The portrait is replaced by a loose abstract of teal, dark green and amber strokes, like foliage, with no figure.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9005a6a6c5855300"}
+{"target": {"record": "8a11ba33af3225e2d241"}, "kind": "keywords", "value": ["abstract foliage", "teal green", "amber", "brush texture", "no subject"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bedb878a940b68c8"}
+{"target": {"record": "8a11ba33af3225e2d241"}, "kind": "effect_tags", "value": ["abstract", "painterly", "colour_shift", "subject_lost", "texture"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "022a3e31f2e9eb25"}
diff --git a/records/sd1/sweep/8a11ba33af3225e2d241/measurements.json b/records/sd1/sweep/8a11ba33af3225e2d241/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9267f9e9a1a857505cc6bd283524d81de304465d
--- /dev/null
+++ b/records/sd1/sweep/8a11ba33af3225e2d241/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a11ba33af3225e2d241",
+ "values": {
+  "cosine_distance": {
+   "value": 0.561045,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.70299,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.912077,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.446904,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.058,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00557,
+   "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": 58.527,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.758,
+   "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": 0.2281,
+   "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/8a11ba33af3225e2d241/record.json b/records/sd1/sweep/8a11ba33af3225e2d241/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fdf3315348fb79bc6e753f351c1fea708e3fe48
--- /dev/null
+++ b/records/sd1/sweep/8a11ba33af3225e2d241/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a11ba33af3225e2d241",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:16Z",
+ "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.8.2.conv",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "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/e1e46018424f4624579313b55539a551fae56ef7900cd58ef26a3a4ad6788d38.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e1e46018424f4624579313b55539a551fae56ef7900cd58ef26a3a4ad6788d38",
+  "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": "c768bfc26a4b3e94ce68"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a11ba33af3225e2d241/workflow.json b/records/sd1/sweep/8a11ba33af3225e2d241/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c14b2b60b8280440b898c932df04d71fbf9036cd
--- /dev/null
+++ b/records/sd1/sweep/8a11ba33af3225e2d241/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 8a11ba33af3225e2d241)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.2.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/8a1a543611716693ec9e/interpretations.jsonl b/records/sd1/sweep/8a1a543611716693ec9e/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..abb605e1cdc86d24004009da10c3388404f39472
--- /dev/null
+++ b/records/sd1/sweep/8a1a543611716693ec9e/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8a1a543611716693ec9e"}, "kind": "caption", "value": "Frontal portrait of woman with wavy dark hair in black turtleneck", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c540679f9fa71e70"}
+{"target": {"record": "8a1a543611716693ec9e"}, "kind": "change", "value": "Almost no visible change; the portrait is essentially identical.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "818b9be8f8946137"}
+{"target": {"record": "8a1a543611716693ec9e"}, "kind": "keywords", "value": ["identical", "no change", "portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "906694a0e7cd3a3f"}
+{"target": {"record": "8a1a543611716693ec9e"}, "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": "2026-10-05T21:30:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "31af8da4d435bc94"}
diff --git a/records/sd1/sweep/8a1a543611716693ec9e/measurements.json b/records/sd1/sweep/8a1a543611716693ec9e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..985692ab217eb880388c6c99665f66135c64a0ca
--- /dev/null
+++ b/records/sd1/sweep/8a1a543611716693ec9e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a1a543611716693ec9e",
+ "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/8a1a543611716693ec9e/record.json b/records/sd1/sweep/8a1a543611716693ec9e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..022e1cd0a02e906c04beb92ec18b375324986953
--- /dev/null
+++ b/records/sd1/sweep/8a1a543611716693ec9e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a1a543611716693ec9e",
+ "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.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "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/54b57e969c64dc3defd493dc01b31869a014aa919338df4ea028cc046309f5e2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "54b57e969c64dc3defd493dc01b31869a014aa919338df4ea028cc046309f5e2",
+  "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": "3c1435db4ab133cc97f3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a1a543611716693ec9e/workflow.json b/records/sd1/sweep/8a1a543611716693ec9e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..60071551d7ede79209975507f296773773e8f729
--- /dev/null
+++ b/records/sd1/sweep/8a1a543611716693ec9e/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 8a1a543611716693ec9e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.h_upd\",\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/8a1e9200a59538fa2f4f/measurements.json b/records/sd1/sweep/8a1e9200a59538fa2f4f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7194e78703ebe8fcc3ff5e66641703f54373858
--- /dev/null
+++ b/records/sd1/sweep/8a1e9200a59538fa2f4f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a1e9200a59538fa2f4f",
+ "values": {
+  "std": {
+   "value": 32.192,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01961,
+   "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": 1.007,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000264,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.00116,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000709,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000309,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9974,
+   "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": 0.2336,
+   "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/8a1e9200a59538fa2f4f/record.json b/records/sd1/sweep/8a1e9200a59538fa2f4f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4615c34d1e75bb8d707a29cde1dbc7aa676c0b5d
--- /dev/null
+++ b/records/sd1/sweep/8a1e9200a59538fa2f4f/record.json
@@ -0,0 +1,86 @@
+{
+ "id": "8a1e9200a59538fa2f4f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:29:19Z",
+ "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": 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": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "medium",
+   "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": [
+   "out.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 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": "912d90f59a361cfa968c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a1e9200a59538fa2f4f/workflow.json b/records/sd1/sweep/8a1e9200a59538fa2f4f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..626f9e6b70b4347e32fbd2b64c16e2b94968a76e
--- /dev/null
+++ b/records/sd1/sweep/8a1e9200a59538fa2f4f/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 8a1e9200a59538fa2f4f)"
+  },
+  "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\": 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/8a1f92528c369b8d65c8/interpretations.jsonl b/records/sd1/sweep/8a1f92528c369b8d65c8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c07cf95a3589cb54b31399a3b71f07a4fff4c28d
--- /dev/null
+++ b/records/sd1/sweep/8a1f92528c369b8d65c8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8a1f92528c369b8d65c8"}, "kind": "caption", "value": "Old bearded man in a knit beanie and sweater before a choppy sea", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:53:46Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "46b62f2073a2dab4"}
+{"target": {"record": "8a1f92528c369b8d65c8"}, "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": "2026-10-05T04:53:46Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f8718a7dc4a8b283"}
+{"target": {"record": "8a1f92528c369b8d65c8"}, "kind": "keywords", "value": ["old man portrait", "no change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:53:46Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7dc74af8b699da1a"}
+{"target": {"record": "8a1f92528c369b8d65c8"}, "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": "2026-10-05T04:53:46Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9a15c356adf8e64e"}
diff --git a/records/sd1/sweep/8a1f92528c369b8d65c8/measurements.json b/records/sd1/sweep/8a1f92528c369b8d65c8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..de2fa9c6e74069d029398c66f87ffdba806e1162
--- /dev/null
+++ b/records/sd1/sweep/8a1f92528c369b8d65c8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a1f92528c369b8d65c8",
+ "values": {
+  "std": {
+   "value": 45.606,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01375,
+   "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": 5.338,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013859,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.025129,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.003273,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004501,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0064,
+   "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": 0.1912,
+   "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/8a1f92528c369b8d65c8/record.json b/records/sd1/sweep/8a1f92528c369b8d65c8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..18e0244ed4922fbc2ca2abb8730569633ba53a3f
--- /dev/null
+++ b/records/sd1/sweep/8a1f92528c369b8d65c8/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8a1f92528c369b8d65c8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:30:07Z",
+ "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": "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": "2976534844fe80be26b4",
+  "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": "3acb046700b1560ebcd1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a1f92528c369b8d65c8/workflow.json b/records/sd1/sweep/8a1f92528c369b8d65c8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7cd54777d6847a4355cac892dac31c027c7afa37
--- /dev/null
+++ b/records/sd1/sweep/8a1f92528c369b8d65c8/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 8a1f92528c369b8d65c8)"
+  },
+  "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": "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/8a24b387e60d5f49c61c/measurements.json b/records/sd1/sweep/8a24b387e60d5f49c61c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6a97f4fe64b386dc4d21a22db36a649ba0e30d6
--- /dev/null
+++ b/records/sd1/sweep/8a24b387e60d5f49c61c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a24b387e60d5f49c61c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.246538,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.308422,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.301083,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.156034,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.047,
+   "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": 29.236,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9412,
+   "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": 0.6558,
+   "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/8a24b387e60d5f49c61c/record.json b/records/sd1/sweep/8a24b387e60d5f49c61c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..102749a7d1359410d7774f98cbeaa7cf45330c26
--- /dev/null
+++ b/records/sd1/sweep/8a24b387e60d5f49c61c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a24b387e60d5f49c61c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:18Z",
+ "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.3.1.transformer_blocks.0.attn2.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "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/c92875a5af7f97d17d528ff7cc7a841a9389afd5a1f2c5083703b2edfd1d6561.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c92875a5af7f97d17d528ff7cc7a841a9389afd5a1f2c5083703b2edfd1d6561",
+  "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"
+    ],
+    "previous_id": "2ff488f22d50588c0603"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a24b387e60d5f49c61c/workflow.json b/records/sd1/sweep/8a24b387e60d5f49c61c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e5bd7898044edf5453dee830b5eca78767d39b5
--- /dev/null
+++ b/records/sd1/sweep/8a24b387e60d5f49c61c/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 8a24b387e60d5f49c61c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.transformer_blocks.0.attn2.to_out.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/8a31180f485aa0bf8d00/measurements.json b/records/sd1/sweep/8a31180f485aa0bf8d00/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..77d5ed605b779474ca5bb9a23d1089cd8819c3cf
--- /dev/null
+++ b/records/sd1/sweep/8a31180f485aa0bf8d00/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a31180f485aa0bf8d00",
+ "values": {
+  "cosine_distance": {
+   "value": 0.114231,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.14252,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.072635,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027059,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.46,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00466,
+   "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": 17.503,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0121,
+   "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": 0.1249,
+   "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/8a31180f485aa0bf8d00/record.json b/records/sd1/sweep/8a31180f485aa0bf8d00/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..279682d47374dcb0a58d03e30d6d0a8434b36f43
--- /dev/null
+++ b/records/sd1/sweep/8a31180f485aa0bf8d00/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a31180f485aa0bf8d00",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:03Z",
+ "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.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "mid",
+   "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/fadee6e054d87af4e7e741a4a08eeee898299639211ef9212f1e53e6f4a1b595.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "fadee6e054d87af4e7e741a4a08eeee898299639211ef9212f1e53e6f4a1b595",
+  "type_path": "UNetModel.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": "c16b715d28801ab13ffe"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a31180f485aa0bf8d00/workflow.json b/records/sd1/sweep/8a31180f485aa0bf8d00/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ecb4a8a1e2cf99af0ec3d97c8c4caee8483e630
--- /dev/null
+++ b/records/sd1/sweep/8a31180f485aa0bf8d00/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 8a31180f485aa0bf8d00)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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": [
+    "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/8a35e9f5b581b35e1b85/measurements.json b/records/sd1/sweep/8a35e9f5b581b35e1b85/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b96bcf3ac1641bf2feb5e287497ccb24e2d7f4a
--- /dev/null
+++ b/records/sd1/sweep/8a35e9f5b581b35e1b85/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a35e9f5b581b35e1b85",
+ "values": {
+  "cosine_distance": {
+   "value": 0.17619,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.259984,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.164203,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036223,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.999,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00383,
+   "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": 25.059,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0137,
+   "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": 0.0929,
+   "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/8a35e9f5b581b35e1b85/record.json b/records/sd1/sweep/8a35e9f5b581b35e1b85/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe334f689652fa0fe996440ae8ca5187181fd08a
--- /dev/null
+++ b/records/sd1/sweep/8a35e9f5b581b35e1b85/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a35e9f5b581b35e1b85",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:58Z",
+ "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": 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.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.hi",
+   "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/b7a5596a15dee6d201b80ef7a13a505c17abdc2144c110cda2ded9a9d3d1b262.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b7a5596a15dee6d201b80ef7a13a505c17abdc2144c110cda2ded9a9d3d1b262",
+  "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": "0f2fd73a120c38b7d77e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a35e9f5b581b35e1b85/workflow.json b/records/sd1/sweep/8a35e9f5b581b35e1b85/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..486542840c59d74a98a8e512444e392a04f67357
--- /dev/null
+++ b/records/sd1/sweep/8a35e9f5b581b35e1b85/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 8a35e9f5b581b35e1b85)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.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/8a36c1948a21f9534dff/measurements.json b/records/sd1/sweep/8a36c1948a21f9534dff/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc0713ab07784847db88241d209435e28a1c1fc2
--- /dev/null
+++ b/records/sd1/sweep/8a36c1948a21f9534dff/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a36c1948a21f9534dff",
+ "values": {
+  "cosine_distance": {
+   "value": 0.348699,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.55607,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.678999,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.303056,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 38.918,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00455,
+   "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": 39.118,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9333,
+   "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": 0.6067,
+   "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/8a36c1948a21f9534dff/record.json b/records/sd1/sweep/8a36c1948a21f9534dff/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfbed10dd13b4fd688546b8dd002a3252aab9158
--- /dev/null
+++ b/records/sd1/sweep/8a36c1948a21f9534dff/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a36c1948a21f9534dff",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:33Z",
+ "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": "output_blocks.6.1.transformer_blocks.0.attn1.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "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/e6f638c1ed27c36f9bc06b4db050bfa40d0a61b1636e4cc6e3a667178bb08b37.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e6f638c1ed27c36f9bc06b4db050bfa40d0a61b1636e4cc6e3a667178bb08b37",
+  "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"
+    ],
+    "previous_id": "e7994debfd6191ee00e7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a36c1948a21f9534dff/workflow.json b/records/sd1/sweep/8a36c1948a21f9534dff/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..76c51b64952e7922b5729929393c90203ec5bd27
--- /dev/null
+++ b/records/sd1/sweep/8a36c1948a21f9534dff/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 8a36c1948a21f9534dff)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.transformer_blocks.0.attn1.to_v\",\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/8a3e2b347043db5e1e6b/measurements.json b/records/sd1/sweep/8a3e2b347043db5e1e6b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a107c0737d069447b59773924025390adb15f79
--- /dev/null
+++ b/records/sd1/sweep/8a3e2b347043db5e1e6b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a3e2b347043db5e1e6b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.421457,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.56848,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.345423,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.264816,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 35.807,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00293,
+   "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": 39.485,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0513,
+   "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": 0.3555,
+   "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/8a3e2b347043db5e1e6b/record.json b/records/sd1/sweep/8a3e2b347043db5e1e6b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..299f89e9f07da44ffe2a3b91ef9b48c2b12e8f7e
--- /dev/null
+++ b/records/sd1/sweep/8a3e2b347043db5e1e6b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a3e2b347043db5e1e6b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:44Z",
+ "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.4.0.in_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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d962d3948323dd6e5138f35e3bdb01094279bced66af0651db5a5792d592af01.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d962d3948323dd6e5138f35e3bdb01094279bced66af0651db5a5792d592af01",
+  "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": "062031c5bd050c280b57"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a3e2b347043db5e1e6b/workflow.json b/records/sd1/sweep/8a3e2b347043db5e1e6b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d5ffb533bf259d8fea2c581c1fde10331831668
--- /dev/null
+++ b/records/sd1/sweep/8a3e2b347043db5e1e6b/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 8a3e2b347043db5e1e6b)"
+  },
+  "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\": 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/8a425ed5c6b3ec134b49/measurements.json b/records/sd1/sweep/8a425ed5c6b3ec134b49/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5317ba7559fb5f71d1065bb6224f624caa09965f
--- /dev/null
+++ b/records/sd1/sweep/8a425ed5c6b3ec134b49/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a425ed5c6b3ec134b49",
+ "values": {
+  "cosine_distance": {
+   "value": 0.010258,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.018288,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.007458,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003249,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.364,
+   "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": 4.4,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9984,
+   "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": 0.1516,
+   "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/8a425ed5c6b3ec134b49/record.json b/records/sd1/sweep/8a425ed5c6b3ec134b49/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5de2dcb0759ff0ff4870e949e7a8d9a0eecab404
--- /dev/null
+++ b/records/sd1/sweep/8a425ed5c6b3ec134b49/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a425ed5c6b3ec134b49",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:44Z",
+ "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": "input_blocks.10.0.emb_layers.0",
+   "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/9f6a88ef162907dbb17f3b21fabbd72fb2c75c03554c683789d80650befb8607.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9f6a88ef162907dbb17f3b21fabbd72fb2c75c03554c683789d80650befb8607",
+  "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": "cbd51dd8d843ba662035"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a425ed5c6b3ec134b49/workflow.json b/records/sd1/sweep/8a425ed5c6b3ec134b49/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..de400842394f75bd66c0c1595bc5303fc2b0bb6d
--- /dev/null
+++ b/records/sd1/sweep/8a425ed5c6b3ec134b49/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 8a425ed5c6b3ec134b49)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.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/8a47e73782a6a1f274d9/measurements.json b/records/sd1/sweep/8a47e73782a6a1f274d9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..638a585fa552bb44e007cdd0365618e61628bee0
--- /dev/null
+++ b/records/sd1/sweep/8a47e73782a6a1f274d9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a47e73782a6a1f274d9",
+ "values": {
+  "std": {
+   "value": 38.17,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01319,
+   "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": 9.67,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.052507,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.156433,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.061387,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.015548,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0233,
+   "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/8a47e73782a6a1f274d9/record.json b/records/sd1/sweep/8a47e73782a6a1f274d9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..427694e594cd2311e8a44a5960e9df11fce97f6b
--- /dev/null
+++ b/records/sd1/sweep/8a47e73782a6a1f274d9/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8a47e73782a6a1f274d9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:00:56Z",
+ "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": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "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": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.5,
+    "applied": 0.5,
+    "reason": "outside safe range [0.6, 1.6] 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": "fd1f0ae899d66c3481d3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a47e73782a6a1f274d9/workflow.json b/records/sd1/sweep/8a47e73782a6a1f274d9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..16ad7c4ab83ba4f31b4f0fe766491cd99a097142
--- /dev/null
+++ b/records/sd1/sweep/8a47e73782a6a1f274d9/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 8a47e73782a6a1f274d9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\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": 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/8a535135b6bc004a8810/measurements.json b/records/sd1/sweep/8a535135b6bc004a8810/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d51539e272e6ef4dab3672cc50ccbe3a752fd61
--- /dev/null
+++ b/records/sd1/sweep/8a535135b6bc004a8810/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a535135b6bc004a8810",
+ "values": {
+  "cosine_distance": {
+   "value": 0.065881,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.084215,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.037558,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025554,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.78,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00505,
+   "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": 10.494,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9889,
+   "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": 0.0862,
+   "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/8a535135b6bc004a8810/record.json b/records/sd1/sweep/8a535135b6bc004a8810/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fe1b69c8bec84c7795c10be0f1f182a6ef79a25
--- /dev/null
+++ b/records/sd1/sweep/8a535135b6bc004a8810/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a535135b6bc004a8810",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:20Z",
+ "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": "output_blocks.5.1.transformer_blocks.0.attn2.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "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/1d6273590960acd2fa5ea7662a4846230f80cfefe39c2c776b7dcc7650b0cad3.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1d6273590960acd2fa5ea7662a4846230f80cfefe39c2c776b7dcc7650b0cad3",
+  "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": "98eda19901d279c4480c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a535135b6bc004a8810/workflow.json b/records/sd1/sweep/8a535135b6bc004a8810/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..58abbd5e75e1130da4a9eec7088b1be8871a6f43
--- /dev/null
+++ b/records/sd1/sweep/8a535135b6bc004a8810/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 8a535135b6bc004a8810)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.transformer_blocks.0.attn2.to_out.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/8a54df35a2eb5a599850/measurements.json b/records/sd1/sweep/8a54df35a2eb5a599850/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9231576dca109aad17997ca93fda4f5d3efa6213
--- /dev/null
+++ b/records/sd1/sweep/8a54df35a2eb5a599850/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a54df35a2eb5a599850",
+ "values": {
+  "cosine_distance": {
+   "value": 0.386559,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.380837,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.358868,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.22606,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.648,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00658,
+   "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": 39.355,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9491,
+   "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": 0.7004,
+   "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/8a54df35a2eb5a599850/record.json b/records/sd1/sweep/8a54df35a2eb5a599850/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e43c7644d81167149a6255474cf30cb41fd71903
--- /dev/null
+++ b/records/sd1/sweep/8a54df35a2eb5a599850/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a54df35a2eb5a599850",
+ "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.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/77107992556d560b5b2fd909936d6192872eb1bc73879f0001b3a9a5e6a05137.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "77107992556d560b5b2fd909936d6192872eb1bc73879f0001b3a9a5e6a05137",
+  "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": "0b8fa528ebdf97040e3b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a54df35a2eb5a599850/workflow.json b/records/sd1/sweep/8a54df35a2eb5a599850/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..54ea940d65751ba76b6d3ddbe1b2d5775ae8b4d2
--- /dev/null
+++ b/records/sd1/sweep/8a54df35a2eb5a599850/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 8a54df35a2eb5a599850)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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/8a6849852fc54e75f61e/measurements.json b/records/sd1/sweep/8a6849852fc54e75f61e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..50130072d7d3bf2a94282e8dbb726b0b450661dd
--- /dev/null
+++ b/records/sd1/sweep/8a6849852fc54e75f61e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a6849852fc54e75f61e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.153106,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.35036,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.334715,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.167026,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.043,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00352,
+   "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": 23.078,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.023,
+   "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": 0.0961,
+   "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/8a6849852fc54e75f61e/record.json b/records/sd1/sweep/8a6849852fc54e75f61e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9fab548559f794bf5f71065f06b4d0d2a289b52
--- /dev/null
+++ b/records/sd1/sweep/8a6849852fc54e75f61e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a6849852fc54e75f61e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:25Z",
+ "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.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f12bebae343149d70338f9354024743d164ce5829df7d566f9f70fde821159ac.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f12bebae343149d70338f9354024743d164ce5829df7d566f9f70fde821159ac",
+  "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": "faf40feb59bdc6f08593"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a6849852fc54e75f61e/workflow.json b/records/sd1/sweep/8a6849852fc54e75f61e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..334e2a6aab7df7db89497d475cd3a557b31bd072
--- /dev/null
+++ b/records/sd1/sweep/8a6849852fc54e75f61e/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 8a6849852fc54e75f61e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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": [
+    "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/8a6e46f92813c19e278d/measurements.json b/records/sd1/sweep/8a6e46f92813c19e278d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..013b89abec8fe82ed676dc9f4eedf2358f50e527
--- /dev/null
+++ b/records/sd1/sweep/8a6e46f92813c19e278d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a6e46f92813c19e278d",
+ "values": {
+  "std": {
+   "value": 36.54,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01591,
+   "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": 11.515,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.083896,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.180785,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.076631,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.017762,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9532,
+   "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": 0.0141,
+   "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/8a6e46f92813c19e278d/record.json b/records/sd1/sweep/8a6e46f92813c19e278d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..10dd9878a319c4a0ecc9e832c9ded24f707e8da9
--- /dev/null
+++ b/records/sd1/sweep/8a6e46f92813c19e278d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8a6e46f92813c19e278d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:16:22Z",
+ "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": 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": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 1234
+   },
+   "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": "de1eccace874429a61dc",
+  "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/8a6e46f92813c19e278d/workflow.json b/records/sd1/sweep/8a6e46f92813c19e278d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f24b7735fd5d178b0d745685a50d50c1a129099d
--- /dev/null
+++ b/records/sd1/sweep/8a6e46f92813c19e278d/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 8a6e46f92813c19e278d)"
+  },
+  "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\": 1234\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/8a79bd4cd9a1f7af3a83/measurements.json b/records/sd1/sweep/8a79bd4cd9a1f7af3a83/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..879a5dd737e879a0b7f6d81bb130a6e0bae01126
--- /dev/null
+++ b/records/sd1/sweep/8a79bd4cd9a1f7af3a83/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a79bd4cd9a1f7af3a83",
+ "values": {
+  "cosine_distance": {
+   "value": 0.460504,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.679735,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.515056,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.468815,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.464,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01206,
+   "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": 49.939,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9437,
+   "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": 0.0273,
+   "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/8a79bd4cd9a1f7af3a83/record.json b/records/sd1/sweep/8a79bd4cd9a1f7af3a83/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8151e780cbecf97c89f6bbf642e99a8a10159425
--- /dev/null
+++ b/records/sd1/sweep/8a79bd4cd9a1f7af3a83/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a79bd4cd9a1f7af3a83",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:28Z",
+ "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.2.0.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.hi",
+   "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/848bd7c553aea4e9d07f050609cbf69f97cdf5a58af8c4b1adfaa3b014dec510.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "848bd7c553aea4e9d07f050609cbf69f97cdf5a58af8c4b1adfaa3b014dec510",
+  "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": "3e3c2c69ac8e4a60171d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a79bd4cd9a1f7af3a83/workflow.json b/records/sd1/sweep/8a79bd4cd9a1f7af3a83/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..eae19f22b0d1ab2286f76c4ac9526a250c563456
--- /dev/null
+++ b/records/sd1/sweep/8a79bd4cd9a1f7af3a83/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 8a79bd4cd9a1f7af3a83)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.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/8a7b977c2018969d79cb/interpretations.jsonl b/records/sd1/sweep/8a7b977c2018969d79cb/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..067fac6035e53b5829d88d506b0d3ca5e8df1017
--- /dev/null
+++ b/records/sd1/sweep/8a7b977c2018969d79cb/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8a7b977c2018969d79cb"}, "kind": "caption", "value": "Grainy monochrome portrait with smeared, fragmented face and ragged ink-splattered border", "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": "4ac5ca85a90fa261"}
+{"target": {"record": "8a7b977c2018969d79cb"}, "kind": "change", "value": "The photo collapses into a gritty desaturated print where the face is broken into displaced, smeared patches, the hair goes curly and the border becomes rough and speckled.", "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": "536f3b630117478d"}
+{"target": {"record": "8a7b977c2018969d79cb"}, "kind": "keywords", "value": ["broken face", "grunge print", "monochrome", "ragged border", "distorted portrait", "glitch"], "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": "d82ba29e03b553c3"}
+{"target": {"record": "8a7b977c2018969d79cb"}, "kind": "effect_tags", "value": ["distorted", "fragmented", "desaturated", "texture", "glitch", "surreal", "degenerate"], "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": "a8e270d38af51e67"}
diff --git a/records/sd1/sweep/8a7b977c2018969d79cb/measurements.json b/records/sd1/sweep/8a7b977c2018969d79cb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8249fae71e9fd7a9caad16a26d15d38204efedde
--- /dev/null
+++ b/records/sd1/sweep/8a7b977c2018969d79cb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a7b977c2018969d79cb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.232892,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.437824,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.269767,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.247297,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.047,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00677,
+   "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": 32.516,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9643,
+   "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": 0.801,
+   "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/8a7b977c2018969d79cb/record.json b/records/sd1/sweep/8a7b977c2018969d79cb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3827250be433151acbd65da35438db34e00360fb
--- /dev/null
+++ b/records/sd1/sweep/8a7b977c2018969d79cb/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a7b977c2018969d79cb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:41Z",
+ "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.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.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/3d53eaef52fda92394aa8a9a2e8635bc41c9fbf8fc5366d65e1b96984ed2db25.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3d53eaef52fda92394aa8a9a2e8635bc41c9fbf8fc5366d65e1b96984ed2db25",
+  "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": "2314d0c95fb490b3b072"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a7b977c2018969d79cb/workflow.json b/records/sd1/sweep/8a7b977c2018969d79cb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f3b62f46aedcbdc52f943ca0c9b77d817e3d8074
--- /dev/null
+++ b/records/sd1/sweep/8a7b977c2018969d79cb/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 8a7b977c2018969d79cb)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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/8a7c6240437e666b4929/measurements.json b/records/sd1/sweep/8a7c6240437e666b4929/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3078e1027dc17f4f7ef72d74bd0392525d6f3ef
--- /dev/null
+++ b/records/sd1/sweep/8a7c6240437e666b4929/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a7c6240437e666b4929",
+ "values": {
+  "cosine_distance": {
+   "value": 0.420367,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.409092,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.369605,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.20475,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.892,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00779,
+   "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": 43.877,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0282,
+   "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": 0.2477,
+   "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/8a7c6240437e666b4929/record.json b/records/sd1/sweep/8a7c6240437e666b4929/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..82f2046431e879a9f2802c04898a8a2729275385
--- /dev/null
+++ b/records/sd1/sweep/8a7c6240437e666b4929/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a7c6240437e666b4929",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:43Z",
+ "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.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "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/36c79902d0d15a265b14767f4313d5a2ad1446d1419ed3c5433a590e8a02ef23.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "36c79902d0d15a265b14767f4313d5a2ad1446d1419ed3c5433a590e8a02ef23",
+  "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"
+    ],
+    "previous_id": "df38359c51286a417f41"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a7c6240437e666b4929/workflow.json b/records/sd1/sweep/8a7c6240437e666b4929/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..71ee01cb1673b3bde8a376afd5fcb9e4ed7f6c75
--- /dev/null
+++ b/records/sd1/sweep/8a7c6240437e666b4929/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 8a7c6240437e666b4929)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.transformer_blocks.0.attn1.to_k\",\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/8a802f749c2a6b9691a3/measurements.json b/records/sd1/sweep/8a802f749c2a6b9691a3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..84d827fcf34a29bf5c817c6bbb57dac09a0cb7a6
--- /dev/null
+++ b/records/sd1/sweep/8a802f749c2a6b9691a3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a802f749c2a6b9691a3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.299189,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.335075,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.229942,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.109741,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.528,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00629,
+   "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": 36.983,
+   "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": 0.3537,
+   "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/8a802f749c2a6b9691a3/record.json b/records/sd1/sweep/8a802f749c2a6b9691a3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..36efcaadf3ecfa666f141b8aaefae8478a1df3ba
--- /dev/null
+++ b/records/sd1/sweep/8a802f749c2a6b9691a3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8a802f749c2a6b9691a3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:13Z",
+ "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.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "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/3af4d16f4f1e281d1737941655a4bc4dd476c6f34e2db447de9831704323fc0a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3af4d16f4f1e281d1737941655a4bc4dd476c6f34e2db447de9831704323fc0a",
+  "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": "86716d74a39ac67db62c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a802f749c2a6b9691a3/workflow.json b/records/sd1/sweep/8a802f749c2a6b9691a3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..36d023c850a6c22fcf45cc988ffd9f4431c42da9
--- /dev/null
+++ b/records/sd1/sweep/8a802f749c2a6b9691a3/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 8a802f749c2a6b9691a3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.in_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/8a88b50ea3146de3f4d5/measurements.json b/records/sd1/sweep/8a88b50ea3146de3f4d5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..894ea6be7392bb6689c9ff1f6047c388f4a3e0f0
--- /dev/null
+++ b/records/sd1/sweep/8a88b50ea3146de3f4d5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a88b50ea3146de3f4d5",
+ "values": {
+  "std": {
+   "value": 30.622,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02961,
+   "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": 13.223,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.078424,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.194895,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131566,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025843,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9796,
+   "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": 0.201,
+   "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/8a88b50ea3146de3f4d5/record.json b/records/sd1/sweep/8a88b50ea3146de3f4d5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9622007502e7fb065ca9ea4da97b756dbfca33e7
--- /dev/null
+++ b/records/sd1/sweep/8a88b50ea3146de3f4d5/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "8a88b50ea3146de3f4d5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:59:42Z",
+ "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": 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.8.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.lo",
+   "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": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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/8a88b50ea3146de3f4d5/workflow.json b/records/sd1/sweep/8a88b50ea3146de3f4d5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..11851542f67549439d0141e79291aade405c895a
--- /dev/null
+++ b/records/sd1/sweep/8a88b50ea3146de3f4d5/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 8a88b50ea3146de3f4d5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\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 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/8a90ecc0819def99e93a/measurements.json b/records/sd1/sweep/8a90ecc0819def99e93a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae6aeb04fc6a9eb552ac045ba923c34924acaf39
--- /dev/null
+++ b/records/sd1/sweep/8a90ecc0819def99e93a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8a90ecc0819def99e93a",
+ "values": {
+  "std": {
+   "value": 51.372,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00403,
+   "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": 22.7,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.25182,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.424195,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.37779,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.094034,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0248,
+   "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": 0.0495,
+   "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/8a90ecc0819def99e93a/record.json b/records/sd1/sweep/8a90ecc0819def99e93a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..001eab196b54a797551950c6922cc07ec51e2407
--- /dev/null
+++ b/records/sd1/sweep/8a90ecc0819def99e93a/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "8a90ecc0819def99e93a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:27:18Z",
+ "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": "input_blocks.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "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": [
+   "in.hi",
+   "multiply",
+   0.5,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.5,
+    "applied": 0.5,
+    "reason": "outside safe range [0.6, 1.6] 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/8a90ecc0819def99e93a/workflow.json b/records/sd1/sweep/8a90ecc0819def99e93a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b0ae23896338d6e0ba50c4c37a1d613bfdc8f65
--- /dev/null
+++ b/records/sd1/sweep/8a90ecc0819def99e93a/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 8a90ecc0819def99e93a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\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/8a93ff6cf071bd401b97/measurements.json b/records/sd1/sweep/8a93ff6cf071bd401b97/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4718cc27d0e43462a7eee7b49f5965725a73e0aa
--- /dev/null
+++ b/records/sd1/sweep/8a93ff6cf071bd401b97/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a93ff6cf071bd401b97",
+ "values": {
+  "std": {
+   "value": 37.712,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02066,
+   "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": 38.485,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.389595,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.607173,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.5747,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.165523,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9621,
+   "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": 0.4706,
+   "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/8a93ff6cf071bd401b97/record.json b/records/sd1/sweep/8a93ff6cf071bd401b97/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4482b9b486c815c2c931918477038d4e94b35cc9
--- /dev/null
+++ b/records/sd1/sweep/8a93ff6cf071bd401b97/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "8a93ff6cf071bd401b97",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:20:43Z",
+ "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": 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": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "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": [
+   "mid",
+   "multiply",
+   0.0,
+   "all"
+  ],
+  "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": "23b8f7f9c6bc00938577"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a93ff6cf071bd401b97/workflow.json b/records/sd1/sweep/8a93ff6cf071bd401b97/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b6cf9f6ebfa5235bc423f66cdf5257d29daadda
--- /dev/null
+++ b/records/sd1/sweep/8a93ff6cf071bd401b97/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 8a93ff6cf071bd401b97)"
+  },
+  "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}",
+   "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/8a9614575a16edd71e14/measurements.json b/records/sd1/sweep/8a9614575a16edd71e14/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cae83ec2fdc2a37cb515cd7b470ebb1923060a8
--- /dev/null
+++ b/records/sd1/sweep/8a9614575a16edd71e14/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a9614575a16edd71e14",
+ "values": {
+  "std": {
+   "value": 49.011,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01303,
+   "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": 15.465,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.124615,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.301469,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.431914,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.130183,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0322,
+   "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": 0.0247,
+   "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/8a9614575a16edd71e14/record.json b/records/sd1/sweep/8a9614575a16edd71e14/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d591e97186e9cf0d2f6d593e52a8b7b6b19f2c71
--- /dev/null
+++ b/records/sd1/sweep/8a9614575a16edd71e14/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8a9614575a16edd71e14",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:20:02Z",
+ "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": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "grow",
+   "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.mid",
+   "scale",
+   1.4,
+   "late"
+  ],
+  "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": "5b5fed59c355fe81f368"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a9614575a16edd71e14/workflow.json b/records/sd1/sweep/8a9614575a16edd71e14/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b956c0bd11937803ca4fc38490c5cb6c308e3652
--- /dev/null
+++ b/records/sd1/sweep/8a9614575a16edd71e14/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 8a9614575a16edd71e14)"
+  },
+  "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\": 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/8a9a91d9493146eb6cfe/measurements.json b/records/sd1/sweep/8a9a91d9493146eb6cfe/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae9ce8d3daa76235faadbe28fa6e5ff1223dd5f2
--- /dev/null
+++ b/records/sd1/sweep/8a9a91d9493146eb6cfe/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8a9a91d9493146eb6cfe",
+ "values": {
+  "cosine_distance": {
+   "value": 0.415586,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.45326,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.351331,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.107697,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 40.309,
+   "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"
+  },
+  "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": 40.497,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9563,
+   "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": 0.0566,
+   "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/8a9a91d9493146eb6cfe/record.json b/records/sd1/sweep/8a9a91d9493146eb6cfe/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a6b7e2219caa10be5d2ccf1d14803507db94034
--- /dev/null
+++ b/records/sd1/sweep/8a9a91d9493146eb6cfe/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8a9a91d9493146eb6cfe",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:34Z",
+ "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": 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.2.1.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/23fdf87f9cc3ba3b15e69f7d3147560a95f339fff2cd5bd12a6b1d5c5dfcfefe.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "23fdf87f9cc3ba3b15e69f7d3147560a95f339fff2cd5bd12a6b1d5c5dfcfefe",
+  "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",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5c0640a0a69a21f3295d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8a9a91d9493146eb6cfe/workflow.json b/records/sd1/sweep/8a9a91d9493146eb6cfe/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..53d3bc8c0ed4112fb8768b4de2a9b13c77a21c43
--- /dev/null
+++ b/records/sd1/sweep/8a9a91d9493146eb6cfe/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 8a9a91d9493146eb6cfe)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.1.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": [
+    "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/8aae1ef9bc045897a766/measurements.json b/records/sd1/sweep/8aae1ef9bc045897a766/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6a995ee50e5ca21a08fa3c6652ee34b24455481
--- /dev/null
+++ b/records/sd1/sweep/8aae1ef9bc045897a766/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8aae1ef9bc045897a766",
+ "values": {
+  "cosine_distance": {
+   "value": 0.189732,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.214403,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.183854,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.16715,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.39,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00578,
+   "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": 21.842,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.002,
+   "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": 0.2053,
+   "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/8aae1ef9bc045897a766/record.json b/records/sd1/sweep/8aae1ef9bc045897a766/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..673377904a8cda0ed79e38af5f8ff91175c298ca
--- /dev/null
+++ b/records/sd1/sweep/8aae1ef9bc045897a766/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8aae1ef9bc045897a766",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:49Z",
+ "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": "output_blocks.2.0.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/c310978c3b376a47710706eda3fa1197cdb67dbd24bce573a2c91bf3ba0d9a56.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c310978c3b376a47710706eda3fa1197cdb67dbd24bce573a2c91bf3ba0d9a56",
+  "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": "571741e5e31a6162fcfc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8aae1ef9bc045897a766/workflow.json b/records/sd1/sweep/8aae1ef9bc045897a766/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..263c63093744353be40c3fea6103bd618a309830
--- /dev/null
+++ b/records/sd1/sweep/8aae1ef9bc045897a766/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 8aae1ef9bc045897a766)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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": [
+    "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/8ab6cee5a2b5525e6c77/measurements.json b/records/sd1/sweep/8ab6cee5a2b5525e6c77/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..98df67aa4683852d786392c657fdddbaed74b02d
--- /dev/null
+++ b/records/sd1/sweep/8ab6cee5a2b5525e6c77/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ab6cee5a2b5525e6c77",
+ "values": {
+  "cosine_distance": {
+   "value": 0.003199,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.007033,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.002906,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001407,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.338,
+   "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": 2.844,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.009,
+   "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": 0.1495,
+   "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/8ab6cee5a2b5525e6c77/record.json b/records/sd1/sweep/8ab6cee5a2b5525e6c77/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cae107733cfa9211e53cfe1d4cabb665967e657d
--- /dev/null
+++ b/records/sd1/sweep/8ab6cee5a2b5525e6c77/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ab6cee5a2b5525e6c77",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:51Z",
+ "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.9.1.transformer_blocks.0.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "cross_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/af01af9380dad2cfa77bfc07d163df77dc17bd99361cb4af77a14ab20cace1c2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "af01af9380dad2cfa77bfc07d163df77dc17bd99361cb4af77a14ab20cace1c2",
+  "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": "8d74b494d47c83eaa8ce"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ab6cee5a2b5525e6c77/workflow.json b/records/sd1/sweep/8ab6cee5a2b5525e6c77/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f6d7e1583d540567d58be5e21064ac7f72bad91
--- /dev/null
+++ b/records/sd1/sweep/8ab6cee5a2b5525e6c77/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 8ab6cee5a2b5525e6c77)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.transformer_blocks.0.attn2.to_v\",\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/8ac1617bd3cae1f7b4b4/measurements.json b/records/sd1/sweep/8ac1617bd3cae1f7b4b4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a64659bfafbe306d04abd49d517487232749b42
--- /dev/null
+++ b/records/sd1/sweep/8ac1617bd3cae1f7b4b4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ac1617bd3cae1f7b4b4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.258067,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.302386,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.102673,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.069991,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 39.726,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0088,
+   "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": 31.381,
+   "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": 0.0795,
+   "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/8ac1617bd3cae1f7b4b4/record.json b/records/sd1/sweep/8ac1617bd3cae1f7b4b4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3bfbcf1cc649c2f8f823c35767f225ef432cc6c
--- /dev/null
+++ b/records/sd1/sweep/8ac1617bd3cae1f7b4b4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ac1617bd3cae1f7b4b4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:25Z",
+ "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.11.1.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/6517e2f7d85eba288b84759113742428a2397610821fb1e921f70029d8762f8a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6517e2f7d85eba288b84759113742428a2397610821fb1e921f70029d8762f8a",
+  "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": "63ec74cbccc0d46b19a2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ac1617bd3cae1f7b4b4/workflow.json b/records/sd1/sweep/8ac1617bd3cae1f7b4b4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d59e6e8ecb5825b120277bce72a44c2851a34c7d
--- /dev/null
+++ b/records/sd1/sweep/8ac1617bd3cae1f7b4b4/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 8ac1617bd3cae1f7b4b4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.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/8ac435644e3503ee5fb9/measurements.json b/records/sd1/sweep/8ac435644e3503ee5fb9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8165055813f5326c237f1af82f55c9cf5aaed3b9
--- /dev/null
+++ b/records/sd1/sweep/8ac435644e3503ee5fb9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ac435644e3503ee5fb9",
+ "values": {
+  "std": {
+   "value": 61.795,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01294,
+   "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": 56.138,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.614743,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.655928,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.619005,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.11712,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0185,
+   "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": 0.0748,
+   "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/8ac435644e3503ee5fb9/record.json b/records/sd1/sweep/8ac435644e3503ee5fb9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c804cbe4dd13c8bd6ce7a8ce517c2b63e64112ac
--- /dev/null
+++ b/records/sd1/sweep/8ac435644e3503ee5fb9/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8ac435644e3503ee5fb9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:59:43Z",
+ "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": 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": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "half",
+   "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": [
+   "in.hi",
+   "rotate",
+   180.0,
+   "early"
+  ],
+  "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": "f010290c9e71cb271293"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ac435644e3503ee5fb9/workflow.json b/records/sd1/sweep/8ac435644e3503ee5fb9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..49c44e35d3842f611954ae554ab6826530e5b586
--- /dev/null
+++ b/records/sd1/sweep/8ac435644e3503ee5fb9/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 8ac435644e3503ee5fb9)"
+  },
+  "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 \"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/8ac6af8af154da53961a/measurements.json b/records/sd1/sweep/8ac6af8af154da53961a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b087ad244d6235cc23ab854781e82e66b30ea2ec
--- /dev/null
+++ b/records/sd1/sweep/8ac6af8af154da53961a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ac6af8af154da53961a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.117074,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.136703,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.122196,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.12948,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.012,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00321,
+   "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": 14.097,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9671,
+   "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": 0.6281,
+   "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/8ac6af8af154da53961a/record.json b/records/sd1/sweep/8ac6af8af154da53961a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a500f0af3216773dde03c2fd5ba46ad154c3f932
--- /dev/null
+++ b/records/sd1/sweep/8ac6af8af154da53961a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8ac6af8af154da53961a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:22Z",
+ "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.8.1.proj_in",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "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/72b7025b6330480f68bc78287dd902109031b693dca9fc621e1c2dcb9afa3984.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "72b7025b6330480f68bc78287dd902109031b693dca9fc621e1c2dcb9afa3984",
+  "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": "9619cd6a7ac20f7b1492"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ac6af8af154da53961a/workflow.json b/records/sd1/sweep/8ac6af8af154da53961a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..37d1a3ebc82cd3e2e64da0683c723fb089b5aaa3
--- /dev/null
+++ b/records/sd1/sweep/8ac6af8af154da53961a/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 8ac6af8af154da53961a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.proj_in\",\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/8acb99a1afdd6e1ade39/measurements.json b/records/sd1/sweep/8acb99a1afdd6e1ade39/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cbc69b927b1ef585e152e93ccb9c8aa4838914f
--- /dev/null
+++ b/records/sd1/sweep/8acb99a1afdd6e1ade39/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8acb99a1afdd6e1ade39",
+ "values": {
+  "std": {
+   "value": 40.512,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01986,
+   "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": 19.33,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.159823,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.266357,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.073652,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023794,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0187,
+   "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/8acb99a1afdd6e1ade39/record.json b/records/sd1/sweep/8acb99a1afdd6e1ade39/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a19833e0ad1504305ac84e446d786280a32c516f
--- /dev/null
+++ b/records/sd1/sweep/8acb99a1afdd6e1ade39/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8acb99a1afdd6e1ade39",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:50:54Z",
+ "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": "input_blocks.8.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "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.lo",
+   "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": "f7c2b3bb04bc28d65a2a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8acb99a1afdd6e1ade39/workflow.json b/records/sd1/sweep/8acb99a1afdd6e1ade39/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7500259085714039d722f4e38d65cba8e2a74928
--- /dev/null
+++ b/records/sd1/sweep/8acb99a1afdd6e1ade39/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 8acb99a1afdd6e1ade39)"
+  },
+  "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}",
+   "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/8acc0db7508721c891e8/measurements.json b/records/sd1/sweep/8acc0db7508721c891e8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d49de467d85c766e80a2a745994867d194817eb1
--- /dev/null
+++ b/records/sd1/sweep/8acc0db7508721c891e8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8acc0db7508721c891e8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.307242,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.309761,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.271708,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.218211,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.746,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00664,
+   "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": 34.434,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9741,
+   "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": 0.1271,
+   "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/8acc0db7508721c891e8/record.json b/records/sd1/sweep/8acc0db7508721c891e8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1fa70491dbb2703f1ecf1aa15af84cd09e68cecb
--- /dev/null
+++ b/records/sd1/sweep/8acc0db7508721c891e8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8acc0db7508721c891e8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:06Z",
+ "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.2.in_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": "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/731732c55f96d0e03b1373ad306ee5885c7523840d81d2df0b6eadeb0cff6314.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "731732c55f96d0e03b1373ad306ee5885c7523840d81d2df0b6eadeb0cff6314",
+  "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"
+    ],
+    "previous_id": "f515431fb7d0780342c7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8acc0db7508721c891e8/workflow.json b/records/sd1/sweep/8acc0db7508721c891e8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbbf95bb34c0e47b798d5c1b7146a9b69fc49242
--- /dev/null
+++ b/records/sd1/sweep/8acc0db7508721c891e8/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 8acc0db7508721c891e8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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": [
+    "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/8ad85eb73bf9f31221fe/measurements.json b/records/sd1/sweep/8ad85eb73bf9f31221fe/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..80a9a0daa3ed956fff3c864f007c798acfc53b8c
--- /dev/null
+++ b/records/sd1/sweep/8ad85eb73bf9f31221fe/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ad85eb73bf9f31221fe",
+ "values": {
+  "std": {
+   "value": 26.201,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0224,
+   "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": 14.634,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.0868,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.301062,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.529699,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.158877,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8565,
+   "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": 0.4119,
+   "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/8ad85eb73bf9f31221fe/record.json b/records/sd1/sweep/8ad85eb73bf9f31221fe/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..af1adb367bf9639e59b43c754d91a2379fcc8c89
--- /dev/null
+++ b/records/sd1/sweep/8ad85eb73bf9f31221fe/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8ad85eb73bf9f31221fe",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:07:44Z",
+ "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": 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.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "half",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "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.mid",
+   "rotate",
+   180.0,
+   "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": "194816d0fa8565f621b5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ad85eb73bf9f31221fe/workflow.json b/records/sd1/sweep/8ad85eb73bf9f31221fe/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0883844319d1e35aa25b0f500b2d4bab9cde4d7e
--- /dev/null
+++ b/records/sd1/sweep/8ad85eb73bf9f31221fe/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 8ad85eb73bf9f31221fe)"
+  },
+  "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 \"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": 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/8adcd287a7192d210445/measurements.json b/records/sd1/sweep/8adcd287a7192d210445/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfcd4701cc834953784fa7f0097f0ec1ad5daf6b
--- /dev/null
+++ b/records/sd1/sweep/8adcd287a7192d210445/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8adcd287a7192d210445",
+ "values": {
+  "cosine_distance": {
+   "value": 0.384655,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.458101,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.249387,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.109772,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 74.656,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00499,
+   "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": 51.227,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0686,
+   "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": 0.1689,
+   "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/8adcd287a7192d210445/record.json b/records/sd1/sweep/8adcd287a7192d210445/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c7633926a49fa275148992fe021443d269a496e
--- /dev/null
+++ b/records/sd1/sweep/8adcd287a7192d210445/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8adcd287a7192d210445",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:32Z",
+ "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": "input_blocks.1.0.out_layers.3",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.hi",
+   "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/a0e6a7ca76c77db77e65e4b296795eaf622c12afea22f367d1af28d2f8055b7d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a0e6a7ca76c77db77e65e4b296795eaf622c12afea22f367d1af28d2f8055b7d",
+  "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": "829f145eba995189c457"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8adcd287a7192d210445/workflow.json b/records/sd1/sweep/8adcd287a7192d210445/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7946c3c11acc7201b9d4813e8abc625dafe0cfea
--- /dev/null
+++ b/records/sd1/sweep/8adcd287a7192d210445/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 8adcd287a7192d210445)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.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/8adfadbb76b864361be0/measurements.json b/records/sd1/sweep/8adfadbb76b864361be0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9e671c1cf588524df09942f92c8da25e72a79a2
--- /dev/null
+++ b/records/sd1/sweep/8adfadbb76b864361be0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8adfadbb76b864361be0",
+ "values": {
+  "std": {
+   "value": 32.491,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01925,
+   "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": 6.22,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013919,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.04607,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038079,
+   "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"
+  },
+  "prompt_retention": {
+   "value": 0.9898,
+   "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"
+  },
+  "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/8adfadbb76b864361be0/record.json b/records/sd1/sweep/8adfadbb76b864361be0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..634c42c082dc76da7805c2664283d8ecfb03cc7c
--- /dev/null
+++ b/records/sd1/sweep/8adfadbb76b864361be0/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8adfadbb76b864361be0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:22:46Z",
+ "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": 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": "middle_block.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "shrink",
+   "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": [
+   "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": "1bf7bf5764aa8ff38afd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8adfadbb76b864361be0/workflow.json b/records/sd1/sweep/8adfadbb76b864361be0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad97e1dfdd4d2a542165fe3633844e4fc5064c31
--- /dev/null
+++ b/records/sd1/sweep/8adfadbb76b864361be0/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 8adfadbb76b864361be0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\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 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/8ae349a85fdba3b92eb6/measurements.json b/records/sd1/sweep/8ae349a85fdba3b92eb6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..023cdbc541de1881ca662b4514ad4bafc853f40e
--- /dev/null
+++ b/records/sd1/sweep/8ae349a85fdba3b92eb6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ae349a85fdba3b92eb6",
+ "values": {
+  "std": {
+   "value": 59.668,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00994,
+   "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": 10.525,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.044539,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.061735,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.011999,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004086,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0213,
+   "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": 0.2798,
+   "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/8ae349a85fdba3b92eb6/record.json b/records/sd1/sweep/8ae349a85fdba3b92eb6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8611fc9c6dea26a7d8c67e5316aa30b833449db6
--- /dev/null
+++ b/records/sd1/sweep/8ae349a85fdba3b92eb6/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8ae349a85fdba3b92eb6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:49:48Z",
+ "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": 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.3.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.lo",
+   "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.lo",
+   "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": "801994f0d2a8b77aebdf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ae349a85fdba3b92eb6/workflow.json b/records/sd1/sweep/8ae349a85fdba3b92eb6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5a619ce5932d645f3765fbe14efa4d06a3d6f5d
--- /dev/null
+++ b/records/sd1/sweep/8ae349a85fdba3b92eb6/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 8ae349a85fdba3b92eb6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\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/8ae4950b41fd6fe3d5b5/interpretations.jsonl b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c33606a69a5300a6948a4c332b1d1fa4861cbebd
--- /dev/null
+++ b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ae4950b41fd6fe3d5b5"}, "kind": "caption", "value": "Blurry, soft-focus close-up of a brown-haired woman with a teal background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4d117929257fe707"}
+{"target": {"record": "8ae4950b41fd6fe3d5b5"}, "kind": "change", "value": "The camera moves closer to a larger, different-looking face, and the image turns into heavy soft focus with smeared hair and a muted teal background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f8b8b578f7c1577a"}
+{"target": {"record": "8ae4950b41fd6fe3d5b5"}, "kind": "keywords", "value": ["soft focus", "blurred portrait", "closer crop", "dreamy", "smeared hair"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f4bd2f35853cf52b"}
+{"target": {"record": "8ae4950b41fd6fe3d5b5"}, "kind": "effect_tags", "value": ["blur", "zoom", "subject_kept", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "219d3c9a0ebbfa46"}
diff --git a/records/sd1/sweep/8ae4950b41fd6fe3d5b5/measurements.json b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e0139bbf8c33a1a931b630be1e5f3f82decc3a7a
--- /dev/null
+++ b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ae4950b41fd6fe3d5b5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.260414,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.331791,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.095271,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.052437,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.813,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00169,
+   "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": 25.991,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0042,
+   "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": 0.2891,
+   "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/8ae4950b41fd6fe3d5b5/record.json b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..09b0a2aafa970839a47084544e5b990091d116a9
--- /dev/null
+++ b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ae4950b41fd6fe3d5b5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:49Z",
+ "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": "input_blocks.6.0.op",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.mid",
+   "kind": "downsample",
+   "module_type": "Conv2d",
+   "container": "Downsample",
+   "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/ac51bde447b2c23c93cb6cc9c8ccef571eb4a8b74b9df028f1e58de28355bd13.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ac51bde447b2c23c93cb6cc9c8ccef571eb4a8b74b9df028f1e58de28355bd13",
+  "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": "9503d8f7ee1392084a4d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ae4950b41fd6fe3d5b5/workflow.json b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d52ab211ccd597004d193c917b6555d29964ee7
--- /dev/null
+++ b/records/sd1/sweep/8ae4950b41fd6fe3d5b5/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 8ae4950b41fd6fe3d5b5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.6.0.op\",\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/8ae54cb5227556b7c48a/measurements.json b/records/sd1/sweep/8ae54cb5227556b7c48a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1708f1f1c125224db098cbca6904e155702b2b8
--- /dev/null
+++ b/records/sd1/sweep/8ae54cb5227556b7c48a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ae54cb5227556b7c48a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.071222,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.085944,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.053347,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.016566,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.83,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00605,
+   "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": 10.795,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0037,
+   "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": 0.1623,
+   "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/8ae54cb5227556b7c48a/record.json b/records/sd1/sweep/8ae54cb5227556b7c48a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e100b656ec4688f5072b72b1e6a5cd26233578d
--- /dev/null
+++ b/records/sd1/sweep/8ae54cb5227556b7c48a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ae54cb5227556b7c48a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:47Z",
+ "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": "output_blocks.9.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.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/4d1562419ca208cafd7715867b81cb2e849a41f1e9a3dc89abd8e226c1baa833.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4d1562419ca208cafd7715867b81cb2e849a41f1e9a3dc89abd8e226c1baa833",
+  "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": "9f25aaae0bbbbaf87a8f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ae54cb5227556b7c48a/workflow.json b/records/sd1/sweep/8ae54cb5227556b7c48a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0752ea0b553a5d9badc0cecc92a84bf4c032b4c4
--- /dev/null
+++ b/records/sd1/sweep/8ae54cb5227556b7c48a/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 8ae54cb5227556b7c48a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.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/8ae8a27e1f8e94132b5f/interpretations.jsonl b/records/sd1/sweep/8ae8a27e1f8e94132b5f/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..4ef3ec846336490b31ba7f52f298a375d1e8d000
--- /dev/null
+++ b/records/sd1/sweep/8ae8a27e1f8e94132b5f/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ae8a27e1f8e94132b5f"}, "kind": "caption", "value": "Loose painterly cluster of colourful fruit and grapes in glass on beige background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0304dda952111ce1"}
+{"target": {"record": "8ae8a27e1f8e94132b5f"}, "kind": "change", "value": "The flowers and table vanish, leaving a loose, top-down jumble of fruit and grapes in translucent glass on a beige wash of paint.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "68a211f1d1f0b8c1"}
+{"target": {"record": "8ae8a27e1f8e94132b5f"}, "kind": "keywords", "value": ["painterly fruit", "grapes", "loose brushwork", "beige wash", "recomposed still life"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "df35dfac017e4298"}
+{"target": {"record": "8ae8a27e1f8e94132b5f"}, "kind": "effect_tags", "value": ["painterly", "new_composition", "style_change", "washed_out", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4af5a92a42594625"}
diff --git a/records/sd1/sweep/8ae8a27e1f8e94132b5f/measurements.json b/records/sd1/sweep/8ae8a27e1f8e94132b5f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fd9941316801f1d9b26170700be92b206ec5086
--- /dev/null
+++ b/records/sd1/sweep/8ae8a27e1f8e94132b5f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ae8a27e1f8e94132b5f",
+ "values": {
+  "std": {
+   "value": 16.076,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01963,
+   "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": 48.057,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.561788,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.65346,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.707619,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.197951,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0544,
+   "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": 0.1794,
+   "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/8ae8a27e1f8e94132b5f/record.json b/records/sd1/sweep/8ae8a27e1f8e94132b5f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..df5e5df89cd533416da41f2813b14f9bcb768923
--- /dev/null
+++ b/records/sd1/sweep/8ae8a27e1f8e94132b5f/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "8ae8a27e1f8e94132b5f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:22Z",
+ "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": 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": 180.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "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",
+   180.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 180.0,
+    "applied": 180.0,
+    "reason": "outside safe range [10.0, 90.0] for out.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": "049db6c6dc5d24f842fd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ae8a27e1f8e94132b5f/workflow.json b/records/sd1/sweep/8ae8a27e1f8e94132b5f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebecc2de3d8aa895cdd9b125f0a221da0feb3d06
--- /dev/null
+++ b/records/sd1/sweep/8ae8a27e1f8e94132b5f/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 8ae8a27e1f8e94132b5f)"
+  },
+  "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\": 180.0\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/8af480aacb2dab44e1d9/measurements.json b/records/sd1/sweep/8af480aacb2dab44e1d9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..afc888577d6b05b618d5fb88b980b1d23388c832
--- /dev/null
+++ b/records/sd1/sweep/8af480aacb2dab44e1d9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8af480aacb2dab44e1d9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.183886,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.244793,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.147053,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.231833,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.615,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00674,
+   "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": 19.909,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8961,
+   "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": 0.3656,
+   "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/8af480aacb2dab44e1d9/record.json b/records/sd1/sweep/8af480aacb2dab44e1d9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6628d92d32085e071f90ced38f48f7a4a31e4253
--- /dev/null
+++ b/records/sd1/sweep/8af480aacb2dab44e1d9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8af480aacb2dab44e1d9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:02Z",
+ "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.5.0.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/d565e137fd66cbd1bb26ca0599f519ce4ca2654c0da8834f870983e180951f79.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d565e137fd66cbd1bb26ca0599f519ce4ca2654c0da8834f870983e180951f79",
+  "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": "a1241d3f64ea89d9eef0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8af480aacb2dab44e1d9/workflow.json b/records/sd1/sweep/8af480aacb2dab44e1d9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..79f94d3c73d0e7068668eb2aec2c5c232e30c1db
--- /dev/null
+++ b/records/sd1/sweep/8af480aacb2dab44e1d9/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 8af480aacb2dab44e1d9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.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/8af7bd821352383cf23f/measurements.json b/records/sd1/sweep/8af7bd821352383cf23f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b87447d6e41c0fcfcbaf5951a2449591aa92bc2
--- /dev/null
+++ b/records/sd1/sweep/8af7bd821352383cf23f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8af7bd821352383cf23f",
+ "values": {
+  "std": {
+   "value": 52.413,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01764,
+   "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": 47.654,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.566091,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.603962,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.416057,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.081392,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.1056,
+   "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": 0.0076,
+   "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/8af7bd821352383cf23f/record.json b/records/sd1/sweep/8af7bd821352383cf23f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..05dc6d03ef7cb7fa36b7495aba9597840e877eac
--- /dev/null
+++ b/records/sd1/sweep/8af7bd821352383cf23f/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8af7bd821352383cf23f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:47:42Z",
+ "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": "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": "27f5baccf7c0d4f1de23",
+  "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": "25065ea1e19cc5a409a3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8af7bd821352383cf23f/workflow.json b/records/sd1/sweep/8af7bd821352383cf23f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f49a50cc2b895b1b0df65486fd2c59cb9864123
--- /dev/null
+++ b/records/sd1/sweep/8af7bd821352383cf23f/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 8af7bd821352383cf23f)"
+  },
+  "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 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/8af834bbcfa3064519c1/measurements.json b/records/sd1/sweep/8af834bbcfa3064519c1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d0d654f2059a9750c6d171b3b803a54eca57f10
--- /dev/null
+++ b/records/sd1/sweep/8af834bbcfa3064519c1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8af834bbcfa3064519c1",
+ "values": {
+  "std": {
+   "value": 47.11,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01368,
+   "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": 2.501,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001459,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.006536,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000997,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00059,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0217,
+   "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": 0.2543,
+   "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/8af834bbcfa3064519c1/record.json b/records/sd1/sweep/8af834bbcfa3064519c1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c25d2e6216b90fa5f0ba40ad7eccae0b9035d6a1
--- /dev/null
+++ b/records/sd1/sweep/8af834bbcfa3064519c1/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "8af834bbcfa3064519c1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:26:59Z",
+ "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": 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": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "heavy",
+   "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",
+   "add_noise",
+   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": "20b185e17042b186db79"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8af834bbcfa3064519c1/workflow.json b/records/sd1/sweep/8af834bbcfa3064519c1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c745567ee84174deb2785c8fbdd1e56ff542ad2
--- /dev/null
+++ b/records/sd1/sweep/8af834bbcfa3064519c1/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 8af834bbcfa3064519c1)"
+  },
+  "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\": 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/8afa536b00b49fa71f77/interpretations.jsonl b/records/sd1/sweep/8afa536b00b49fa71f77/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..8eb2909755e293be71fbae2e5b4a0aafb0c3aa0c
--- /dev/null
+++ b/records/sd1/sweep/8afa536b00b49fa71f77/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8afa536b00b49fa71f77"}, "kind": "caption", "value": "Vertical cluster of glowing neon signs in a dark city canyon, no visible crowd", "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": "24c6b4e8d4718fd6"}
+{"target": {"record": "8afa536b00b49fa71f77"}, "kind": "change", "value": "The street with walkers becomes a dense upward view of stacked neon signs and towers, with the crowd gone and the signage more fragmented.", "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": "decf977f5a0766f7"}
+{"target": {"record": "8afa536b00b49fa71f77"}, "kind": "keywords", "value": ["neon signs", "sign cluster", "night city", "no people", "dense signage"], "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": "88fd920d79554c63"}
+{"target": {"record": "8afa536b00b49fa71f77"}, "kind": "effect_tags", "value": ["new_composition", "fragmented", "saturated"], "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": "2342596a064a75ba"}
diff --git a/records/sd1/sweep/8afa536b00b49fa71f77/measurements.json b/records/sd1/sweep/8afa536b00b49fa71f77/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb1e8314cb54e31a28fb42ecd4c220ee44e16bf1
--- /dev/null
+++ b/records/sd1/sweep/8afa536b00b49fa71f77/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8afa536b00b49fa71f77",
+ "values": {
+  "std": {
+   "value": 31.704,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02805,
+   "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": 25.307,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.226788,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.406942,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.206834,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.039869,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.2785,
+   "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/8afa536b00b49fa71f77/record.json b/records/sd1/sweep/8afa536b00b49fa71f77/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5b7b608cd2a295bb81d14ee011fc260851c7405
--- /dev/null
+++ b/records/sd1/sweep/8afa536b00b49fa71f77/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "8afa536b00b49fa71f77",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:22:57Z",
+ "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 city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "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": "81ef55543b3b3c080780",
+  "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/8afa536b00b49fa71f77/workflow.json b/records/sd1/sweep/8afa536b00b49fa71f77/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d106aa65fb17ef1b520b19e36c5227f6428ec12
--- /dev/null
+++ b/records/sd1/sweep/8afa536b00b49fa71f77/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 8afa536b00b49fa71f77)"
+  },
+  "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 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/8b01dbe37bb1e00d7326/measurements.json b/records/sd1/sweep/8b01dbe37bb1e00d7326/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d463af8f54695feb9aa4edfc8c842c6a6f60feb
--- /dev/null
+++ b/records/sd1/sweep/8b01dbe37bb1e00d7326/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b01dbe37bb1e00d7326",
+ "values": {
+  "cosine_distance": {
+   "value": 0.044461,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.075155,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.050134,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.012772,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.947,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0058,
+   "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": 12.189,
+   "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": 0.1202,
+   "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/8b01dbe37bb1e00d7326/record.json b/records/sd1/sweep/8b01dbe37bb1e00d7326/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd36bcf0cc3f54b1a11e4c0d196d96d941dd11a4
--- /dev/null
+++ b/records/sd1/sweep/8b01dbe37bb1e00d7326/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b01dbe37bb1e00d7326",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:53Z",
+ "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": 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.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.hi",
+   "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/4d5df7c6ff93dc7ada7428c4aea1df5c736c90ff19dfd20a96c44e662a9d977c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4d5df7c6ff93dc7ada7428c4aea1df5c736c90ff19dfd20a96c44e662a9d977c",
+  "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": "1b89212b4088a8632f98"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b01dbe37bb1e00d7326/workflow.json b/records/sd1/sweep/8b01dbe37bb1e00d7326/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e130c8acd7bd42bc70ad58f0911f757bb93a7578
--- /dev/null
+++ b/records/sd1/sweep/8b01dbe37bb1e00d7326/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 8b01dbe37bb1e00d7326)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.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/8b08401a12a5a135985d/measurements.json b/records/sd1/sweep/8b08401a12a5a135985d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e03ee919a1fc06f6497949e61ebfd73f428f0b7
--- /dev/null
+++ b/records/sd1/sweep/8b08401a12a5a135985d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b08401a12a5a135985d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.590313,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.513774,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.375196,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.203445,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.223,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00748,
+   "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": 49.418,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9879,
+   "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": 0.3123,
+   "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/8b08401a12a5a135985d/record.json b/records/sd1/sweep/8b08401a12a5a135985d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fc94981378fb64bd2f93a705b72af2add7799c9
--- /dev/null
+++ b/records/sd1/sweep/8b08401a12a5a135985d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b08401a12a5a135985d",
+ "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.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "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/c98d9802bbea02d856e93c751c213f78ff8a40559445a1ebf8114fa639fa09ce.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c98d9802bbea02d856e93c751c213f78ff8a40559445a1ebf8114fa639fa09ce",
+  "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": "c8b01d9d122c2d42d98e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b08401a12a5a135985d/workflow.json b/records/sd1/sweep/8b08401a12a5a135985d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e19fdffacf2e2d0d0f16da57ba20a145a4d7314e
--- /dev/null
+++ b/records/sd1/sweep/8b08401a12a5a135985d/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 8b08401a12a5a135985d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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/8b0dbd6edd03a83c88e9/interpretations.jsonl b/records/sd1/sweep/8b0dbd6edd03a83c88e9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c5f607d3d7ffb31f0b1b081d78157a299536a970
--- /dev/null
+++ b/records/sd1/sweep/8b0dbd6edd03a83c88e9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b0dbd6edd03a83c88e9"}, "kind": "caption", "value": "Very out-of-focus woman with brown bob in dark collared top on beige 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": "2b2801581efc51e8"}
+{"target": {"record": "8b0dbd6edd03a83c88e9"}, "kind": "change", "value": "The whole image goes heavily out of focus, leaving only a soft blurry frontal woman with a brown bob and muted beige tones.", "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": "d76ba2779ffa54ab"}
+{"target": {"record": "8b0dbd6edd03a83c88e9"}, "kind": "keywords", "value": ["out of focus", "heavy blur", "soft portrait", "beige tones", "bob haircut"], "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": "c0c9ca415d25cbb4"}
+{"target": {"record": "8b0dbd6edd03a83c88e9"}, "kind": "effect_tags", "value": ["blur", "washed_out", "desaturated", "subject_kept"], "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": "80453a177fa711c5"}
diff --git a/records/sd1/sweep/8b0dbd6edd03a83c88e9/measurements.json b/records/sd1/sweep/8b0dbd6edd03a83c88e9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9051e600e7391bbde7cef5eeaf1cda0aa71967eb
--- /dev/null
+++ b/records/sd1/sweep/8b0dbd6edd03a83c88e9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b0dbd6edd03a83c88e9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.414441,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.575013,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.464393,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.231454,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.477,
+   "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": 43.164,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9253,
+   "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": 0.5002,
+   "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/8b0dbd6edd03a83c88e9/record.json b/records/sd1/sweep/8b0dbd6edd03a83c88e9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..31080feb4a814f3f7cb967e1671887e3d52e56d8
--- /dev/null
+++ b/records/sd1/sweep/8b0dbd6edd03a83c88e9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b0dbd6edd03a83c88e9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:32Z",
+ "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": "input_blocks.2.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.hi",
+   "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/549ea8676ecb94aeb7dacea69415f4ad00a6b505262694ef396667a7effdbe7d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "549ea8676ecb94aeb7dacea69415f4ad00a6b505262694ef396667a7effdbe7d",
+  "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": "10cde5bdaec25fc038b2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b0dbd6edd03a83c88e9/workflow.json b/records/sd1/sweep/8b0dbd6edd03a83c88e9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..68663dd638024cd3ec5827b6fa3049e5032eca0f
--- /dev/null
+++ b/records/sd1/sweep/8b0dbd6edd03a83c88e9/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 8b0dbd6edd03a83c88e9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.transformer_blocks.0.attn1.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": [
+    "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/8b0e510f63e50ebd40f9/measurements.json b/records/sd1/sweep/8b0e510f63e50ebd40f9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..abc44d8ffde9f41ce606db6763b76d5b5717d6ed
--- /dev/null
+++ b/records/sd1/sweep/8b0e510f63e50ebd40f9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b0e510f63e50ebd40f9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.188229,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.224108,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.099627,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.072443,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.935,
+   "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": 22.208,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9734,
+   "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": 0.0662,
+   "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/8b0e510f63e50ebd40f9/record.json b/records/sd1/sweep/8b0e510f63e50ebd40f9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7de1dd60cfbeb7aaac8891803a6605cef1d6d0b
--- /dev/null
+++ b/records/sd1/sweep/8b0e510f63e50ebd40f9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b0e510f63e50ebd40f9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:19Z",
+ "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": 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.attn2.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "mid",
+   "kind": "cross_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/7008ba6ebfab346b8dbf4706bff69b4e9c4b556e0c7af70650de84e0956ecc87.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7008ba6ebfab346b8dbf4706bff69b4e9c4b556e0c7af70650de84e0956ecc87",
+  "type_path": "UNetModel.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": "68e641cade866bbf0b68"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b0e510f63e50ebd40f9/workflow.json b/records/sd1/sweep/8b0e510f63e50ebd40f9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa7e447965e481a414cfafa95cafb4463586ec9c
--- /dev/null
+++ b/records/sd1/sweep/8b0e510f63e50ebd40f9/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 8b0e510f63e50ebd40f9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.attn2.to_out.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/8b1602cd280d94eac495/measurements.json b/records/sd1/sweep/8b1602cd280d94eac495/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bed91a357f96efb9d790c6daa1dca1b6988e2d64
--- /dev/null
+++ b/records/sd1/sweep/8b1602cd280d94eac495/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b1602cd280d94eac495",
+ "values": {
+  "cosine_distance": {
+   "value": 0.231586,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.33077,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.228063,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.075009,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.726,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00858,
+   "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": 33.187,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9467,
+   "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": 0.0308,
+   "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/8b1602cd280d94eac495/record.json b/records/sd1/sweep/8b1602cd280d94eac495/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a83ce61fa18ed78c4b0e1031104f68c1f0f59b4d
--- /dev/null
+++ b/records/sd1/sweep/8b1602cd280d94eac495/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b1602cd280d94eac495",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:49Z",
+ "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.4.1.transformer_blocks.0.ff.net.0.proj",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "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/50e98c5d3b057c3c5f66b6a83d2f39e485d5c5eb359f23217b4dac32e083146b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "50e98c5d3b057c3c5f66b6a83d2f39e485d5c5eb359f23217b4dac32e083146b",
+  "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": "6be61b8bd4318b0d1f62"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b1602cd280d94eac495/workflow.json b/records/sd1/sweep/8b1602cd280d94eac495/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4817ffbef13a186ab3e135a5cd389a1345dbec13
--- /dev/null
+++ b/records/sd1/sweep/8b1602cd280d94eac495/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 8b1602cd280d94eac495)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.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/8b1fcbdb3a677266ac0f/measurements.json b/records/sd1/sweep/8b1fcbdb3a677266ac0f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b2a9f565e9af83e3102dc6044e21cc7c082f189
--- /dev/null
+++ b/records/sd1/sweep/8b1fcbdb3a677266ac0f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b1fcbdb3a677266ac0f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.233162,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.267663,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.123334,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.067769,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.879,
+   "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"
+  },
+  "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": 28.195,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0179,
+   "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": 0.2547,
+   "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/8b1fcbdb3a677266ac0f/record.json b/records/sd1/sweep/8b1fcbdb3a677266ac0f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..849770d8d7a16da3ef6410fb1eabcbde409d810e
--- /dev/null
+++ b/records/sd1/sweep/8b1fcbdb3a677266ac0f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b1fcbdb3a677266ac0f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:18Z",
+ "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": "input_blocks.5.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.mid",
+   "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/9cf7e306d1efb09a27de283cbbf632281f5463d2418806713593143f95810908.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9cf7e306d1efb09a27de283cbbf632281f5463d2418806713593143f95810908",
+  "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": "a546f15c227d68fa62ca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b1fcbdb3a677266ac0f/workflow.json b/records/sd1/sweep/8b1fcbdb3a677266ac0f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe286a43fd2432ee025df802ddebae555d04e81e
--- /dev/null
+++ b/records/sd1/sweep/8b1fcbdb3a677266ac0f/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 8b1fcbdb3a677266ac0f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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": [
+    "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/8b25945dcf733f2a0cfc/measurements.json b/records/sd1/sweep/8b25945dcf733f2a0cfc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bed2af9dc1bac574dcaf86c5b224d81bb407ba91
--- /dev/null
+++ b/records/sd1/sweep/8b25945dcf733f2a0cfc/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b25945dcf733f2a0cfc",
+ "values": {
+  "cosine_distance": {
+   "value": 0.181009,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.270205,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.072736,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.042381,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.425,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00634,
+   "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": 22.994,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9687,
+   "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": 0.0992,
+   "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/8b25945dcf733f2a0cfc/record.json b/records/sd1/sweep/8b25945dcf733f2a0cfc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f472654334043bcee6fe61d692a6e1c400c4d94
--- /dev/null
+++ b/records/sd1/sweep/8b25945dcf733f2a0cfc/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b25945dcf733f2a0cfc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:28Z",
+ "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.5.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "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/8e34ba5e2a9a7373e42a469e4bb7be2f2a5da2b8425e4f491f46383e5ddc7332.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8e34ba5e2a9a7373e42a469e4bb7be2f2a5da2b8425e4f491f46383e5ddc7332",
+  "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": "2482bfad595ed42a08ca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b25945dcf733f2a0cfc/workflow.json b/records/sd1/sweep/8b25945dcf733f2a0cfc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..750847dcc916faddd1395899ed3bd69cbffaa3cd
--- /dev/null
+++ b/records/sd1/sweep/8b25945dcf733f2a0cfc/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 8b25945dcf733f2a0cfc)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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/8b263fd312d3517f84b9/measurements.json b/records/sd1/sweep/8b263fd312d3517f84b9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..86e9f7f16c0c216ad8bd8a54e3330d79d43737ee
--- /dev/null
+++ b/records/sd1/sweep/8b263fd312d3517f84b9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b263fd312d3517f84b9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.356025,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.41708,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.263798,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.135062,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.057,
+   "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": 42.696,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0474,
+   "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": 0.2561,
+   "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/8b263fd312d3517f84b9/record.json b/records/sd1/sweep/8b263fd312d3517f84b9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a8b4e69a939ef0354dbb8953ed24b417d7a3c11
--- /dev/null
+++ b/records/sd1/sweep/8b263fd312d3517f84b9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b263fd312d3517f84b9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:15Z",
+ "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": "input_blocks.4.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.mid",
+   "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/c2838658ef7a85506c1660a8b6d2ed52e0bbfb2103010cabc11eb3a2930e0788.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c2838658ef7a85506c1660a8b6d2ed52e0bbfb2103010cabc11eb3a2930e0788",
+  "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": "dec9c1911ad7655dab4e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b263fd312d3517f84b9/workflow.json b/records/sd1/sweep/8b263fd312d3517f84b9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ffcb32b6557a8bfd57c5248aad7f341ac418cba
--- /dev/null
+++ b/records/sd1/sweep/8b263fd312d3517f84b9/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 8b263fd312d3517f84b9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.out_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": [
+    "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/8b26d93513c6c9e6aa35/interpretations.jsonl b/records/sd1/sweep/8b26d93513c6c9e6aa35/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5d2bbcafc7f95757c9d348a187ff883719d18ed2
--- /dev/null
+++ b/records/sd1/sweep/8b26d93513c6c9e6aa35/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b26d93513c6c9e6aa35"}, "kind": "caption", "value": "Young woman with pale skin and red lips in a soft-edged vintage photo frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "912b3ab54ff06681"}
+{"target": {"record": "8b26d93513c6c9e6aa35"}, "kind": "change", "value": "The subject is kept but the face becomes paler and smoother with redder lips, and the frame gains warm, burnt-looking edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9d099c4f896ce4e1"}
+{"target": {"record": "8b26d93513c6c9e6aa35"}, "kind": "keywords", "value": ["pale complexion", "red lips", "vintage frame", "soft portrait", "burnt edges"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ccaa3e252600dcac"}
+{"target": {"record": "8b26d93513c6c9e6aa35"}, "kind": "effect_tags", "value": ["subject_kept", "framing", "subtle", "bright"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "74ebd43be86f35c1"}
diff --git a/records/sd1/sweep/8b26d93513c6c9e6aa35/measurements.json b/records/sd1/sweep/8b26d93513c6c9e6aa35/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..142446d25f6ae78eba6c0e0691733e82cb592e5f
--- /dev/null
+++ b/records/sd1/sweep/8b26d93513c6c9e6aa35/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b26d93513c6c9e6aa35",
+ "values": {
+  "cosine_distance": {
+   "value": 0.16268,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.193837,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.106214,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.131464,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.21,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00398,
+   "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": 19.633,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.918,
+   "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": 0.6121,
+   "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/8b26d93513c6c9e6aa35/record.json b/records/sd1/sweep/8b26d93513c6c9e6aa35/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eebfad4a6d9b50cd10dc1e6cac08d24ee46d9a5b
--- /dev/null
+++ b/records/sd1/sweep/8b26d93513c6c9e6aa35/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b26d93513c6c9e6aa35",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:15Z",
+ "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.8.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.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/3c57a2a6e099461f5ed4e767dcb5145d55fab9db8590ef31aba4ab8231acab88.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3c57a2a6e099461f5ed4e767dcb5145d55fab9db8590ef31aba4ab8231acab88",
+  "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": "03b1edb17bf49ad8f501"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b26d93513c6c9e6aa35/workflow.json b/records/sd1/sweep/8b26d93513c6c9e6aa35/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b0b2facc2c229fa0c3b510f018bade389a98e2e
--- /dev/null
+++ b/records/sd1/sweep/8b26d93513c6c9e6aa35/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 8b26d93513c6c9e6aa35)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.proj_in\",\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/8b29aa73afb58c0d797d/interpretations.jsonl b/records/sd1/sweep/8b29aa73afb58c0d797d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..dab5572722cdda6bad31cc78be21793645a58a62
--- /dev/null
+++ b/records/sd1/sweep/8b29aa73afb58c0d797d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b29aa73afb58c0d797d"}, "kind": "caption", "value": "Young man with swept-back dark hair, dark green top, aged beige photo with worn edges", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "83202bc87e9e848a"}
+{"target": {"record": "8b29aa73afb58c0d797d"}, "kind": "change", "value": "The subject becomes a young man with swept-back hair, with similar framing and a warmer, more worn vintage print.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f02598cd4eff2c25"}
+{"target": {"record": "8b29aa73afb58c0d797d"}, "kind": "keywords", "value": ["young man", "swept-back hair", "aged photo", "vintage print", "gender swap"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "664c97d09f6fbdf5"}
+{"target": {"record": "8b29aa73afb58c0d797d"}, "kind": "effect_tags", "value": ["subject_kept", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:11:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "18fb522795c03470"}
diff --git a/records/sd1/sweep/8b29aa73afb58c0d797d/measurements.json b/records/sd1/sweep/8b29aa73afb58c0d797d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..550dfba7cb1df3469c64ff16b56fa93444aaf1d3
--- /dev/null
+++ b/records/sd1/sweep/8b29aa73afb58c0d797d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b29aa73afb58c0d797d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.308512,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.300239,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.139087,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.1352,
+   "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": 0.00517,
+   "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": 36.945,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.993,
+   "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": 0.3571,
+   "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/8b29aa73afb58c0d797d/record.json b/records/sd1/sweep/8b29aa73afb58c0d797d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bdf066df82f191754dfc784f8666dd110ba85ac
--- /dev/null
+++ b/records/sd1/sweep/8b29aa73afb58c0d797d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b29aa73afb58c0d797d",
+ "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_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.mid",
+   "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/8096a0829ddb994f011bb1fcc3791b2e7f8965d20edf50a2913dfa80934fdaf4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8096a0829ddb994f011bb1fcc3791b2e7f8965d20edf50a2913dfa80934fdaf4",
+  "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": "008880539832f538ab9e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b29aa73afb58c0d797d/workflow.json b/records/sd1/sweep/8b29aa73afb58c0d797d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bcba318f3220deeea5fb75408eec99333a838b60
--- /dev/null
+++ b/records/sd1/sweep/8b29aa73afb58c0d797d/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 8b29aa73afb58c0d797d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.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/8b34ede97f8c98df087e/measurements.json b/records/sd1/sweep/8b34ede97f8c98df087e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f30372e156707eaf8634c53cda742ba46b9e3a8b
--- /dev/null
+++ b/records/sd1/sweep/8b34ede97f8c98df087e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b34ede97f8c98df087e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.116114,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.145367,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.087544,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040243,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.375,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00545,
+   "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": 15.199,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0159,
+   "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": 0.1111,
+   "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/8b34ede97f8c98df087e/record.json b/records/sd1/sweep/8b34ede97f8c98df087e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..90d8e2ba92cfcbad174e8ab52605144e86b9de23
--- /dev/null
+++ b/records/sd1/sweep/8b34ede97f8c98df087e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b34ede97f8c98df087e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:04Z",
+ "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": "input_blocks.7.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "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/0cdf4f02e730809d18ed260763224e3632f448484927c5e4dd8eddc7fbbf604c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0cdf4f02e730809d18ed260763224e3632f448484927c5e4dd8eddc7fbbf604c",
+  "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": "1042100bf96b5d591700"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b34ede97f8c98df087e/workflow.json b/records/sd1/sweep/8b34ede97f8c98df087e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd7771eb6e9dc246ae3b86d70aca7f15e3e53846
--- /dev/null
+++ b/records/sd1/sweep/8b34ede97f8c98df087e/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 8b34ede97f8c98df087e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.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/8b3718ae0335b791f48c/measurements.json b/records/sd1/sweep/8b3718ae0335b791f48c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..205d1fb29da34997aa13aacca83a994840906084
--- /dev/null
+++ b/records/sd1/sweep/8b3718ae0335b791f48c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b3718ae0335b791f48c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.414151,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.436993,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.267483,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.12561,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.302,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00391,
+   "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": 48.31,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8874,
+   "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": 0.2973,
+   "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/8b3718ae0335b791f48c/record.json b/records/sd1/sweep/8b3718ae0335b791f48c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d5a44973e0883da2db363840eddfb380b6aa59c
--- /dev/null
+++ b/records/sd1/sweep/8b3718ae0335b791f48c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b3718ae0335b791f48c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:45Z",
+ "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": "input_blocks.5.1.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/7b08022f54ab81c780f9bfaaed0c60773d5506512c6b8883bdb17d29b595c5ec.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7b08022f54ab81c780f9bfaaed0c60773d5506512c6b8883bdb17d29b595c5ec",
+  "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": "7cdb9fbed88636982b3d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b3718ae0335b791f48c/workflow.json b/records/sd1/sweep/8b3718ae0335b791f48c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..db15e1689124ca895f992e222ee47ac4cf899dac
--- /dev/null
+++ b/records/sd1/sweep/8b3718ae0335b791f48c/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 8b3718ae0335b791f48c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.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/8b3763a4a897d3f1c013/measurements.json b/records/sd1/sweep/8b3763a4a897d3f1c013/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..201197c062e2867352b2b25dd1fb99be4ac76d54
--- /dev/null
+++ b/records/sd1/sweep/8b3763a4a897d3f1c013/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b3763a4a897d3f1c013",
+ "values": {
+  "cosine_distance": {
+   "value": 0.187193,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.305568,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.43235,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.284668,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.501,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00323,
+   "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": 22.803,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0121,
+   "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": 0.3778,
+   "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/8b3763a4a897d3f1c013/record.json b/records/sd1/sweep/8b3763a4a897d3f1c013/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb0ebf76ea5effb50479427bbbec205ccd1b06f3
--- /dev/null
+++ b/records/sd1/sweep/8b3763a4a897d3f1c013/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b3763a4a897d3f1c013",
+ "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.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.mid",
+   "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/9095b4c544fc3c488d0dec8d342516a78a4d33be3de3a8f5e6e3c37faa8a1de6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9095b4c544fc3c488d0dec8d342516a78a4d33be3de3a8f5e6e3c37faa8a1de6",
+  "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": "c2358c55e441dcfeb3a2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b3763a4a897d3f1c013/workflow.json b/records/sd1/sweep/8b3763a4a897d3f1c013/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cea7d0293b5c8c6e7f421807925495864a6aed8f
--- /dev/null
+++ b/records/sd1/sweep/8b3763a4a897d3f1c013/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 8b3763a4a897d3f1c013)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/8b3c9a4b4f7aa550f1db/measurements.json b/records/sd1/sweep/8b3c9a4b4f7aa550f1db/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4ae3f2e17a8a5a79533a73d01aeee70c65ac69f
--- /dev/null
+++ b/records/sd1/sweep/8b3c9a4b4f7aa550f1db/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b3c9a4b4f7aa550f1db",
+ "values": {
+  "cosine_distance": {
+   "value": 0.493764,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.716557,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.819635,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.378813,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.426,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00586,
+   "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": 49.048,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9278,
+   "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": 0.8019,
+   "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/8b3c9a4b4f7aa550f1db/record.json b/records/sd1/sweep/8b3c9a4b4f7aa550f1db/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..333beae7c28d811230fb09b9620559d6b96ac01c
--- /dev/null
+++ b/records/sd1/sweep/8b3c9a4b4f7aa550f1db/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b3c9a4b4f7aa550f1db",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:23Z",
+ "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": "output_blocks.4.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.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/3eab3c7f5d6029b5b34e32a8c76f831099e88a1aa094e32c7782db7e27039f39.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3eab3c7f5d6029b5b34e32a8c76f831099e88a1aa094e32c7782db7e27039f39",
+  "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": "acdcefce209924711a03"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b3c9a4b4f7aa550f1db/workflow.json b/records/sd1/sweep/8b3c9a4b4f7aa550f1db/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8b28137072e5e1f53a0c94721680a8cbaaf1957
--- /dev/null
+++ b/records/sd1/sweep/8b3c9a4b4f7aa550f1db/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 8b3c9a4b4f7aa550f1db)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.transformer_blocks.0.ff.net.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/8b44d78af85addd92704/measurements.json b/records/sd1/sweep/8b44d78af85addd92704/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ef45daf0fe7d2dba1b2c2d8055d6dda389d996e
--- /dev/null
+++ b/records/sd1/sweep/8b44d78af85addd92704/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b44d78af85addd92704",
+ "values": {
+  "cosine_distance": {
+   "value": 0.065503,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.08874,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.024522,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.017264,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.874,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00631,
+   "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": 10.49,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.972,
+   "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": 0.316,
+   "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/8b44d78af85addd92704/record.json b/records/sd1/sweep/8b44d78af85addd92704/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0f6505108256840e57f256ef747d30ec25f6c0e
--- /dev/null
+++ b/records/sd1/sweep/8b44d78af85addd92704/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b44d78af85addd92704",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:03Z",
+ "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": "output_blocks.7.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "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/d8ff582d22e8c0c4d684c78493f25c14719a0fe8648de7848a18006356bf3e60.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d8ff582d22e8c0c4d684c78493f25c14719a0fe8648de7848a18006356bf3e60",
+  "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": "3b82f90cf68a01468646"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b44d78af85addd92704/workflow.json b/records/sd1/sweep/8b44d78af85addd92704/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d478e3d1ad953f07aeeb0a7ac8b079fbec28784
--- /dev/null
+++ b/records/sd1/sweep/8b44d78af85addd92704/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 8b44d78af85addd92704)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.attn1.to_k\",\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/8b4b01450c4083f56cf7/interpretations.jsonl b/records/sd1/sweep/8b4b01450c4083f56cf7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..8d49c4d0ed8d67d3e1ce3f51bdb6a5418ca35c5c
--- /dev/null
+++ b/records/sd1/sweep/8b4b01450c4083f56cf7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b4b01450c4083f56cf7"}, "kind": "caption", "value": "Rough expressive oil portrait with blue eyes, dark curls and patchy grey-ochre background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "89736d24fcd51cbc"}
+{"target": {"record": "8b4b01450c4083f56cf7"}, "kind": "change", "value": "The smooth illustration becomes an expressive, blotchy oil painting, with dark tangled hair, vivid blue eyes, a realistic face and a patchy grey and ochre background that breaks up the edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3406d56769a2fb5f"}
+{"target": {"record": "8b4b01450c4083f56cf7"}, "kind": "keywords", "value": ["expressive oil", "blotchy brushwork", "blue eyes", "patchy background", "rough portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5523c9c9ef366d44"}
+{"target": {"record": "8b4b01450c4083f56cf7"}, "kind": "effect_tags", "value": ["painterly", "texture", "style_change", "subject_kept", "fragmented"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1a2f1179b3e5102d"}
diff --git a/records/sd1/sweep/8b4b01450c4083f56cf7/measurements.json b/records/sd1/sweep/8b4b01450c4083f56cf7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ee1cb346061afb34d8c1cb20d2e28f824ff0753
--- /dev/null
+++ b/records/sd1/sweep/8b4b01450c4083f56cf7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b4b01450c4083f56cf7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.266953,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.392322,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.264922,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.089204,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.859,
+   "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": [],
+   "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": 30.362,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9879,
+   "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": 0.0359,
+   "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/8b4b01450c4083f56cf7/record.json b/records/sd1/sweep/8b4b01450c4083f56cf7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0480fe3ff3f76d3b6365447df1a9d5614b97952
--- /dev/null
+++ b/records/sd1/sweep/8b4b01450c4083f56cf7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b4b01450c4083f56cf7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:18Z",
+ "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": "output_blocks.8.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "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/b970b3ecf48fc2272ad287e15fe1f1c07c3655aee3ef8fef8e7ff9523b4e56b2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b970b3ecf48fc2272ad287e15fe1f1c07c3655aee3ef8fef8e7ff9523b4e56b2",
+  "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": "a836762a82be8ed36c7c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b4b01450c4083f56cf7/workflow.json b/records/sd1/sweep/8b4b01450c4083f56cf7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cff23e68d7cc00044bfd6ed78815d7ac9ddc61d7
--- /dev/null
+++ b/records/sd1/sweep/8b4b01450c4083f56cf7/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 8b4b01450c4083f56cf7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.out_layers.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/8b4dbac932044c2ec433/interpretations.jsonl b/records/sd1/sweep/8b4dbac932044c2ec433/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..63fe2de5de85336c6d1d633e80ceb4198ff29a80
--- /dev/null
+++ b/records/sd1/sweep/8b4dbac932044c2ec433/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b4dbac932044c2ec433"}, "kind": "caption", "value": "Blurry distorted face with dark smeared features framed by fuzzy fur-like hood", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cd60182fe24a0e19"}
+{"target": {"record": "8b4dbac932044c2ec433"}, "kind": "change", "value": "The portrait melts into a smudged, blurry form with a lopsided face and dark blotches inside a hazy hood-like surround.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "635d0a88306ecad3"}
+{"target": {"record": "8b4dbac932044c2ec433"}, "kind": "keywords", "value": ["smeared face", "blurry distortion", "hood", "melted portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5c99e5ace2d1002c"}
+{"target": {"record": "8b4dbac932044c2ec433"}, "kind": "effect_tags", "value": ["blur", "distorted", "surreal", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d50073949db7fccd"}
diff --git a/records/sd1/sweep/8b4dbac932044c2ec433/measurements.json b/records/sd1/sweep/8b4dbac932044c2ec433/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..32369cfebe4b5bf2350a8835f6863f65da6592cd
--- /dev/null
+++ b/records/sd1/sweep/8b4dbac932044c2ec433/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b4dbac932044c2ec433",
+ "values": {
+  "cosine_distance": {
+   "value": 0.364232,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.659807,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.528304,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.295427,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.465,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00206,
+   "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": 37.503,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0394,
+   "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": 0.5162,
+   "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/8b4dbac932044c2ec433/record.json b/records/sd1/sweep/8b4dbac932044c2ec433/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..91781256cd98e57dfdfc19f7c83cdf714910040d
--- /dev/null
+++ b/records/sd1/sweep/8b4dbac932044c2ec433/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b4dbac932044c2ec433",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:09Z",
+ "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": 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.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.mid",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "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/f0400a778aa016d3e4b84673f93c3cadd0bd88bc7bb0f3fe49457f2a0161af6d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "f0400a778aa016d3e4b84673f93c3cadd0bd88bc7bb0f3fe49457f2a0161af6d",
+  "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",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ef1d5cd8c19421a56406"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b4dbac932044c2ec433/workflow.json b/records/sd1/sweep/8b4dbac932044c2ec433/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..30b1c1bcc39072422fbdaee8aa017b4572e72529
--- /dev/null
+++ b/records/sd1/sweep/8b4dbac932044c2ec433/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 8b4dbac932044c2ec433)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.skip_connection\",\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/8b53d8cd99b3e3d45290/interpretations.jsonl b/records/sd1/sweep/8b53d8cd99b3e3d45290/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5b389091940e6f8e4a2df457d3714785edbeef91
--- /dev/null
+++ b/records/sd1/sweep/8b53d8cd99b3e3d45290/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b53d8cd99b3e3d45290"}, "kind": "caption", "value": "Pale ghostly old man with white beard blending into streaky grey waves, dark frame.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b30080a2c1470203"}
+{"target": {"record": "8b53d8cd99b3e3d45290"}, "kind": "change", "value": "The portrait becomes a washed-out, ghostly figure overlaid with horizontal wave streaks and an old-photo border.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1711126a03311362"}
+{"target": {"record": "8b53d8cd99b3e3d45290"}, "kind": "keywords", "value": ["ghostly portrait", "wave overlay", "vintage frame", "faded old man"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c68fb74e4567e47c"}
+{"target": {"record": "8b53d8cd99b3e3d45290"}, "kind": "effect_tags", "value": ["surreal", "washed_out", "desaturated", "texture", "framing", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cdb7db75d987b34b"}
diff --git a/records/sd1/sweep/8b53d8cd99b3e3d45290/measurements.json b/records/sd1/sweep/8b53d8cd99b3e3d45290/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b37c6c433c7416f9801f5fc58d02ae7f61ee5046
--- /dev/null
+++ b/records/sd1/sweep/8b53d8cd99b3e3d45290/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b53d8cd99b3e3d45290",
+ "values": {
+  "std": {
+   "value": 40.314,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01089,
+   "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": 34.961,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.381335,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.403232,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.180464,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.1659,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0082,
+   "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": 0.0297,
+   "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/8b53d8cd99b3e3d45290/record.json b/records/sd1/sweep/8b53d8cd99b3e3d45290/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..166f2e70c937971f7501bdf16081a9721f6002b3
--- /dev/null
+++ b/records/sd1/sweep/8b53d8cd99b3e3d45290/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8b53d8cd99b3e3d45290",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:21:53Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "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": "2976534844fe80be26b4",
+  "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": "8e63d8ce117edb95b920"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b53d8cd99b3e3d45290/workflow.json b/records/sd1/sweep/8b53d8cd99b3e3d45290/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..56008d56661b847f5557d94f858f74cf4fbc8cce
--- /dev/null
+++ b/records/sd1/sweep/8b53d8cd99b3e3d45290/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 8b53d8cd99b3e3d45290)"
+  },
+  "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": "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/8b560c4c35eeb7d303a3/measurements.json b/records/sd1/sweep/8b560c4c35eeb7d303a3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..61898e4381b2909ab9ffec5eefa8342a29f8559d
--- /dev/null
+++ b/records/sd1/sweep/8b560c4c35eeb7d303a3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b560c4c35eeb7d303a3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.241895,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.340312,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.211977,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.15946,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.468,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00332,
+   "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": 28.928,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9645,
+   "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": 0.0111,
+   "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/8b560c4c35eeb7d303a3/record.json b/records/sd1/sweep/8b560c4c35eeb7d303a3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..86a4b2a5be42ec2eec168e058456977505e946e6
--- /dev/null
+++ b/records/sd1/sweep/8b560c4c35eeb7d303a3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b560c4c35eeb7d303a3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:29Z",
+ "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": 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.transformer_blocks.0.attn1.to_out.0",
+   "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/997fab789a3ffe5851c86a118f2c974385db7deccb43e7a53f3c0e97a846c7fa.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "997fab789a3ffe5851c86a118f2c974385db7deccb43e7a53f3c0e97a846c7fa",
+  "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": "63b44ef450a18beecd60"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b560c4c35eeb7d303a3/workflow.json b/records/sd1/sweep/8b560c4c35eeb7d303a3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9dfc977321f87f463aa80e4a23a57d614068019
--- /dev/null
+++ b/records/sd1/sweep/8b560c4c35eeb7d303a3/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 8b560c4c35eeb7d303a3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.transformer_blocks.0.attn1.to_out.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/8b6120e49ed59bcf2dce/interpretations.jsonl b/records/sd1/sweep/8b6120e49ed59bcf2dce/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c605dc6bd73005b27d90017da7e4121463e614e9
--- /dev/null
+++ b/records/sd1/sweep/8b6120e49ed59bcf2dce/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b6120e49ed59bcf2dce"}, "kind": "caption", "value": "Abstract swirl of teal and ochre brushy shapes like crumpled leaves or fabric", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d6b10d9258662607"}
+{"target": {"record": "8b6120e49ed59bcf2dce"}, "kind": "change", "value": "The portrait disappears entirely into an abstract field of teal and mustard painterly folds with no recognisable figure.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "06e0b5f2b1da2fec"}
+{"target": {"record": "8b6120e49ed59bcf2dce"}, "kind": "keywords", "value": ["abstract", "teal and ochre", "brush strokes", "no subject", "organic folds"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4a27381d24a50aa5"}
+{"target": {"record": "8b6120e49ed59bcf2dce"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "painterly", "colour_shift", "texture"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e8fdaa8c88968cc4"}
diff --git a/records/sd1/sweep/8b6120e49ed59bcf2dce/measurements.json b/records/sd1/sweep/8b6120e49ed59bcf2dce/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d15c2b57f3f262266be21296b20d62e9347c64b
--- /dev/null
+++ b/records/sd1/sweep/8b6120e49ed59bcf2dce/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b6120e49ed59bcf2dce",
+ "values": {
+  "cosine_distance": {
+   "value": 0.584014,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.740224,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.887682,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.380949,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 30.447,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00866,
+   "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": 56.452,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.849,
+   "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/8b6120e49ed59bcf2dce/record.json b/records/sd1/sweep/8b6120e49ed59bcf2dce/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b67c0f2ba0d39281412444782932e9a72e29f31
--- /dev/null
+++ b/records/sd1/sweep/8b6120e49ed59bcf2dce/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b6120e49ed59bcf2dce",
+ "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.out_layers.3",
+   "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/ab70f0d73ef8e913f8780284a27178184302428795cd18e00ce73510fa04ff4d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ab70f0d73ef8e913f8780284a27178184302428795cd18e00ce73510fa04ff4d",
+  "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": "960d2f91ff28eef90ce4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b6120e49ed59bcf2dce/workflow.json b/records/sd1/sweep/8b6120e49ed59bcf2dce/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9d8d351c15e73f364c3781452ba522280b8b0c5
--- /dev/null
+++ b/records/sd1/sweep/8b6120e49ed59bcf2dce/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 8b6120e49ed59bcf2dce)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.out_layers.3\",\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/8b6915a7a4caf90a3c76/measurements.json b/records/sd1/sweep/8b6915a7a4caf90a3c76/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08a8f4f13cb418010251fda724a43a93dda4e7fa
--- /dev/null
+++ b/records/sd1/sweep/8b6915a7a4caf90a3c76/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b6915a7a4caf90a3c76",
+ "values": {
+  "std": {
+   "value": 40.679,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01459,
+   "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": 36.045,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.415458,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.555995,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.40262,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.077799,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9893,
+   "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"
+  },
+  "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/8b6915a7a4caf90a3c76/record.json b/records/sd1/sweep/8b6915a7a4caf90a3c76/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..25b1e3fc2c48e03e5ef296152998628ae68868e2
--- /dev/null
+++ b/records/sd1/sweep/8b6915a7a4caf90a3c76/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8b6915a7a4caf90a3c76",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:03Z",
+ "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": 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": 30.0
+   },
+   "group": "out.lo",
+   "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": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "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": "aa775ca861f4867c98cb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b6915a7a4caf90a3c76/workflow.json b/records/sd1/sweep/8b6915a7a4caf90a3c76/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..671550a6a014f7178e753a5ab6392083d91cd3f1
--- /dev/null
+++ b/records/sd1/sweep/8b6915a7a4caf90a3c76/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 8b6915a7a4caf90a3c76)"
+  },
+  "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\": 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 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/8b6d655baafb0e7b3fdf/interpretations.jsonl b/records/sd1/sweep/8b6d655baafb0e7b3fdf/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..efd69323832b978f71df728cfafc25915978eefe
--- /dev/null
+++ b/records/sd1/sweep/8b6d655baafb0e7b3fdf/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8b6d655baafb0e7b3fdf"}, "kind": "caption", "value": "Frontal portrait of a young woman with an auburn bob and dark sweater on a cream background", "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": "8770f6592d72cfed"}
+{"target": {"record": "8b6d655baafb0e7b3fdf"}, "kind": "change", "value": "The pose becomes frontal with a longer reddish-brown bob and a knit sweater; the grey framed backdrop turns into a plain warm cream background with softer, warmer light.", "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": "a199897d3a46626d"}
+{"target": {"record": "8b6d655baafb0e7b3fdf"}, "kind": "keywords", "value": ["frontal portrait", "auburn bob", "cream background", "knit sweater", "warm tones"], "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": "05fefcd0af19cc4c"}
+{"target": {"record": "8b6d655baafb0e7b3fdf"}, "kind": "effect_tags", "value": ["subject_kept", "new_composition", "colour_shift"], "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": "58a3b55398013b73"}
diff --git a/records/sd1/sweep/8b6d655baafb0e7b3fdf/measurements.json b/records/sd1/sweep/8b6d655baafb0e7b3fdf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b42052cee6d4712ed805d7fb2b3a8709b268031d
--- /dev/null
+++ b/records/sd1/sweep/8b6d655baafb0e7b3fdf/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b6d655baafb0e7b3fdf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.41098,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.43657,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.403369,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.137423,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.219,
+   "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": 45.368,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0326,
+   "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": 0.2556,
+   "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/8b6d655baafb0e7b3fdf/record.json b/records/sd1/sweep/8b6d655baafb0e7b3fdf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef31aea3b2cfbc79e7ddfcffd58a1284a69cb562
--- /dev/null
+++ b/records/sd1/sweep/8b6d655baafb0e7b3fdf/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b6d655baafb0e7b3fdf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:45Z",
+ "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": "input_blocks.8.0.in_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "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/32a49bd46f6a6272db223362900b980444cc3146c41cde8606c0ba984f6d7f90.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "32a49bd46f6a6272db223362900b980444cc3146c41cde8606c0ba984f6d7f90",
+  "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": "c174e0b75792028bcce9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b6d655baafb0e7b3fdf/workflow.json b/records/sd1/sweep/8b6d655baafb0e7b3fdf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ce68d5a12df8ed631c2078fa28646f1d50a7d25
--- /dev/null
+++ b/records/sd1/sweep/8b6d655baafb0e7b3fdf/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 8b6d655baafb0e7b3fdf)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.in_layers.0\",\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/8b77cb3da74d7f539e8e/measurements.json b/records/sd1/sweep/8b77cb3da74d7f539e8e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e94c741f05ac50cb11fd1cde8f6994c9ec9742d
--- /dev/null
+++ b/records/sd1/sweep/8b77cb3da74d7f539e8e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8b77cb3da74d7f539e8e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.399369,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.497492,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.306293,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.237947,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.833,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00462,
+   "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": 46.711,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0582,
+   "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": 0.1393,
+   "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/8b77cb3da74d7f539e8e/record.json b/records/sd1/sweep/8b77cb3da74d7f539e8e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b66583baa6e1d6323d8d4638622157de449385a1
--- /dev/null
+++ b/records/sd1/sweep/8b77cb3da74d7f539e8e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8b77cb3da74d7f539e8e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:36Z",
+ "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.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "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/0cd5a2c56b58158e5236aca8760f9384a873d08344ac32155fb37ab5049a5464.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0cd5a2c56b58158e5236aca8760f9384a873d08344ac32155fb37ab5049a5464",
+  "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": "d4743f7817a6187a9997"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b77cb3da74d7f539e8e/workflow.json b/records/sd1/sweep/8b77cb3da74d7f539e8e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e7a6b44ca2a74424defba2f473216a0dd81dfae
--- /dev/null
+++ b/records/sd1/sweep/8b77cb3da74d7f539e8e/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 8b77cb3da74d7f539e8e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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": [
+    "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/8b7fbbc94ba85c9ea2fa/measurements.json b/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..90c7827626339f7adb4791e06b38375c03d1b045
--- /dev/null
+++ b/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b7fbbc94ba85c9ea2fa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.418889,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.636742,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.509402,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.259193,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 31.869,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00242,
+   "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": 38.926,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0518,
+   "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": 0.497,
+   "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/8b7fbbc94ba85c9ea2fa/record.json b/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..68218bf806b7273e274023990a9bb512a0d96381
--- /dev/null
+++ b/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b7fbbc94ba85c9ea2fa",
+ "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": 180
+   },
+   "group": "out.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/bfd85ab7cda362c18497d0cc741c24738c10f1c1efb347852cc74bf59eb6fdef.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "bfd85ab7cda362c18497d0cc741c24738c10f1c1efb347852cc74bf59eb6fdef",
+  "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": "0b5404f13081b901d3c3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/workflow.json b/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f4f2d117dd96390f5a6cfce9b2f5fefac1b91c3
--- /dev/null
+++ b/records/sd1/sweep/8b7fbbc94ba85c9ea2fa/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 8b7fbbc94ba85c9ea2fa)"
+  },
+  "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\": 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/8b86b89281aa8edb192c/measurements.json b/records/sd1/sweep/8b86b89281aa8edb192c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a92ab224de4eb4688ce75c09c2827c88150b1b2
--- /dev/null
+++ b/records/sd1/sweep/8b86b89281aa8edb192c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8b86b89281aa8edb192c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.067099,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.110651,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.03181,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.024668,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 73.568,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00626,
+   "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": 19.726,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9991,
+   "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": 0.1458,
+   "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/8b86b89281aa8edb192c/record.json b/records/sd1/sweep/8b86b89281aa8edb192c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..514b0fe261af1d84a8566c418bcd8677b9fb4516
--- /dev/null
+++ b/records/sd1/sweep/8b86b89281aa8edb192c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8b86b89281aa8edb192c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:01Z",
+ "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.7.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "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/7c203003f71a11ed171c3914f1ed3e8865044c6f417ad14e077eccc03bee9254.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7c203003f71a11ed171c3914f1ed3e8865044c6f417ad14e077eccc03bee9254",
+  "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": "9367f13ea8d1020d7f59"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8b86b89281aa8edb192c/workflow.json b/records/sd1/sweep/8b86b89281aa8edb192c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2485c0b89ab7133a1e89780874b8fb2e9e18432c
--- /dev/null
+++ b/records/sd1/sweep/8b86b89281aa8edb192c/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 8b86b89281aa8edb192c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.proj_in\",\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/8babda2add6eb643d142/interpretations.jsonl b/records/sd1/sweep/8babda2add6eb643d142/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2d4ba1c7780733946e72348dc58907dde41b324f
--- /dev/null
+++ b/records/sd1/sweep/8babda2add6eb643d142/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8babda2add6eb643d142"}, "kind": "caption", "value": "Painted portrait of a dark-haired woman in a black turtleneck, nearly identical to the original", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "93eb7a2888aac6e6"}
+{"target": {"record": "8babda2add6eb643d142"}, "kind": "change", "value": "Almost no visible change. The hair is slightly darker and the face marginally different in shape.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d3ba8c1d63b409af"}
+{"target": {"record": "8babda2add6eb643d142"}, "kind": "keywords", "value": ["near identical", "portrait", "minor variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "df5250f9ad7f620a"}
+{"target": {"record": "8babda2add6eb643d142"}, "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": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "82026dd0fea7f54a"}
diff --git a/records/sd1/sweep/8babda2add6eb643d142/measurements.json b/records/sd1/sweep/8babda2add6eb643d142/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad465717a60b90e065e16f8acc4caea4568cb81d
--- /dev/null
+++ b/records/sd1/sweep/8babda2add6eb643d142/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8babda2add6eb643d142",
+ "values": {
+  "cosine_distance": {
+   "value": 0.035567,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.065572,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.01694,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008309,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.188,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00721,
+   "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": 9.906,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9838,
+   "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": 0.1437,
+   "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/8babda2add6eb643d142/record.json b/records/sd1/sweep/8babda2add6eb643d142/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb57e2aa97b4d3540439b6e1b26944ced0402f73
--- /dev/null
+++ b/records/sd1/sweep/8babda2add6eb643d142/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8babda2add6eb643d142",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:02Z",
+ "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": 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.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "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/d3a376a0b8d14de9e4fb57b613828d9ec3d4880b4adda446f90a0d7a5c86d156.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d3a376a0b8d14de9e4fb57b613828d9ec3d4880b4adda446f90a0d7a5c86d156",
+  "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": "5c0a0dca99cabb1eff0d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8babda2add6eb643d142/workflow.json b/records/sd1/sweep/8babda2add6eb643d142/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d396c4684d3a3c495764bdd965e9ca11a1f9139
--- /dev/null
+++ b/records/sd1/sweep/8babda2add6eb643d142/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 8babda2add6eb643d142)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.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/8bad19c498a95a67bf78/measurements.json b/records/sd1/sweep/8bad19c498a95a67bf78/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..13715f652b68846e6d78acd70c3151b4d6503b9d
--- /dev/null
+++ b/records/sd1/sweep/8bad19c498a95a67bf78/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8bad19c498a95a67bf78",
+ "values": {
+  "cosine_distance": {
+   "value": 0.506522,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.599034,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.695136,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.332123,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.138,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.05299,
+   "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": 61.296,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0524,
+   "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": 0.7481,
+   "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/8bad19c498a95a67bf78/record.json b/records/sd1/sweep/8bad19c498a95a67bf78/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3acc5dc7a9bc141eed7213cae69fff1ec31b760f
--- /dev/null
+++ b/records/sd1/sweep/8bad19c498a95a67bf78/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8bad19c498a95a67bf78",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:34Z",
+ "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": 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.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/38f541ba943d93a38267f15345684d43770d3b37727cf39df02c53594df6e879.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "38f541ba943d93a38267f15345684d43770d3b37727cf39df02c53594df6e879",
+  "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": "b789a77ea8ad79ab5972"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bad19c498a95a67bf78/workflow.json b/records/sd1/sweep/8bad19c498a95a67bf78/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca4479187e873e82f06bb35bbc40d8ac6dbe9558
--- /dev/null
+++ b/records/sd1/sweep/8bad19c498a95a67bf78/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 8bad19c498a95a67bf78)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\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/8bad6d5377a0c9621a3e/interpretations.jsonl b/records/sd1/sweep/8bad6d5377a0c9621a3e/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..28f66bd49757464df0d1f09d333167eefd7c1cb6
--- /dev/null
+++ b/records/sd1/sweep/8bad6d5377a0c9621a3e/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8bad6d5377a0c9621a3e"}, "kind": "caption", "value": "Pale monochrome portrait of person with messy dark hair inside a circular vignette on dotted 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": "22abf47a03f6b480"}
+{"target": {"record": "8bad6d5377a0c9621a3e"}, "kind": "change", "value": "Very similar to F: a bright, high-contrast greyscale androgynous portrait in a circular vignette on textured paper replaces the sepia photo.", "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": "88b714df2d9a4426"}
+{"target": {"record": "8bad6d5377a0c9621a3e"}, "kind": "keywords", "value": ["monochrome portrait", "circular vignette", "dotted paper", "androgynous face", "pale tones"], "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": "57bde7969e7508cb"}
+{"target": {"record": "8bad6d5377a0c9621a3e"}, "kind": "effect_tags", "value": ["desaturated", "high_contrast", "framing", "subject_kept", "style_change", "texture"], "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": "2123c025f5135421"}
diff --git a/records/sd1/sweep/8bad6d5377a0c9621a3e/measurements.json b/records/sd1/sweep/8bad6d5377a0c9621a3e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a01259c37b8d4be042d4cda7c289fa2aa4af244
--- /dev/null
+++ b/records/sd1/sweep/8bad6d5377a0c9621a3e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8bad6d5377a0c9621a3e",
+ "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/8bad6d5377a0c9621a3e/record.json b/records/sd1/sweep/8bad6d5377a0c9621a3e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b3d575055b22329a938b92e2b69f7d12c2c4673
--- /dev/null
+++ b/records/sd1/sweep/8bad6d5377a0c9621a3e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8bad6d5377a0c9621a3e",
+ "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.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/09011bd7b94ee62e46c6ccf6897cb5ecdccf1a903d9ee519e4c4e085237f5289.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "09011bd7b94ee62e46c6ccf6897cb5ecdccf1a903d9ee519e4c4e085237f5289",
+  "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": "350a812d9e776e59be6d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bad6d5377a0c9621a3e/workflow.json b/records/sd1/sweep/8bad6d5377a0c9621a3e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f05e257f206e32eb38782c8a6d49be0056101183
--- /dev/null
+++ b/records/sd1/sweep/8bad6d5377a0c9621a3e/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 8bad6d5377a0c9621a3e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.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": [
+    "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/8bb184543a95e412bfa2/measurements.json b/records/sd1/sweep/8bb184543a95e412bfa2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b490a8d27bcbffbf5edec16851406e92894cfdf7
--- /dev/null
+++ b/records/sd1/sweep/8bb184543a95e412bfa2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8bb184543a95e412bfa2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.064434,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.089204,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.045938,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.031971,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.187,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.004,
+   "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": 12.84,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.017,
+   "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": 0.2188,
+   "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/8bb184543a95e412bfa2/record.json b/records/sd1/sweep/8bb184543a95e412bfa2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..90cf03a4f68593600a5bfb6ca4d2e0126695f081
--- /dev/null
+++ b/records/sd1/sweep/8bb184543a95e412bfa2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8bb184543a95e412bfa2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:51Z",
+ "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": "output_blocks.9.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "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/af73296c70c709bbebded7506a268c0660f429477cd38dcf75c16c9a6aa15a3e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "af73296c70c709bbebded7506a268c0660f429477cd38dcf75c16c9a6aa15a3e",
+  "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"
+    ],
+    "previous_id": "ddf9df253828b87d60b2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bb184543a95e412bfa2/workflow.json b/records/sd1/sweep/8bb184543a95e412bfa2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1bd5eae70a623a2425429c47a7a23e5a2d14284c
--- /dev/null
+++ b/records/sd1/sweep/8bb184543a95e412bfa2/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 8bb184543a95e412bfa2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.transformer_blocks.0.attn2.to_q\",\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/8bb28e3432a9743c3a00/measurements.json b/records/sd1/sweep/8bb28e3432a9743c3a00/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c56631d0474138a23c2d0fd7e9e499d30eafcd2b
--- /dev/null
+++ b/records/sd1/sweep/8bb28e3432a9743c3a00/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8bb28e3432a9743c3a00",
+ "values": {
+  "std": {
+   "value": 94.039,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00305,
+   "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": 82.138,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.725149,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.710073,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.408631,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.26041,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8408,
+   "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": 0.3855,
+   "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/8bb28e3432a9743c3a00/record.json b/records/sd1/sweep/8bb28e3432a9743c3a00/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..875b000a4107e4dd5f7b56022fb83177ff9eeac4
--- /dev/null
+++ b/records/sd1/sweep/8bb28e3432a9743c3a00/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8bb28e3432a9743c3a00",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:48:20Z",
+ "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": 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.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": "27f5baccf7c0d4f1de23",
+  "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": "0580072b88d14c34b3b9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bb28e3432a9743c3a00/workflow.json b/records/sd1/sweep/8bb28e3432a9743c3a00/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2fb5ca60e7dd265fd3a07003960f1beb68f26ee
--- /dev/null
+++ b/records/sd1/sweep/8bb28e3432a9743c3a00/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 8bb28e3432a9743c3a00)"
+  },
+  "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": "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/8bc6e7665d4a56a851c7/measurements.json b/records/sd1/sweep/8bc6e7665d4a56a851c7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..eab0cb6562c6a0b346dce4644e19a2b4ebb9f64a
--- /dev/null
+++ b/records/sd1/sweep/8bc6e7665d4a56a851c7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8bc6e7665d4a56a851c7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.127266,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.184128,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.136798,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.1887,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.866,
+   "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": 16.203,
+   "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": 0.4469,
+   "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/8bc6e7665d4a56a851c7/record.json b/records/sd1/sweep/8bc6e7665d4a56a851c7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..364e1bf9e0d02db14bf6e95012b8a41dec3103a8
--- /dev/null
+++ b/records/sd1/sweep/8bc6e7665d4a56a851c7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8bc6e7665d4a56a851c7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:13Z",
+ "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": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/274aff489585a46c104889dfb31512e6a20135b962879fff16f55a51f8914458.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "274aff489585a46c104889dfb31512e6a20135b962879fff16f55a51f8914458",
+  "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": "d4e331469b141b69f29a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bc6e7665d4a56a851c7/workflow.json b/records/sd1/sweep/8bc6e7665d4a56a851c7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e03cc0f026ece6c4b412ec9e3c6ec76c71bd5c5
--- /dev/null
+++ b/records/sd1/sweep/8bc6e7665d4a56a851c7/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 8bc6e7665d4a56a851c7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.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": [
+    "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/8bca1a1bbda111592518/measurements.json b/records/sd1/sweep/8bca1a1bbda111592518/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d622de02207a585a38c97c2c25a2bc1a40c4a5dd
--- /dev/null
+++ b/records/sd1/sweep/8bca1a1bbda111592518/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8bca1a1bbda111592518",
+ "values": {
+  "cosine_distance": {
+   "value": 0.254696,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.408015,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.215821,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.115498,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 36.566,
+   "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": [],
+   "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": 31.973,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9798,
+   "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/8bca1a1bbda111592518/record.json b/records/sd1/sweep/8bca1a1bbda111592518/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..02e1622d485b4da9c1a298254949faabcb485cae
--- /dev/null
+++ b/records/sd1/sweep/8bca1a1bbda111592518/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8bca1a1bbda111592518",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:18Z",
+ "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": "output_blocks.11.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "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/92cef4f400d15a1c4ee03187af523660d2245c01582fd7d2b06299d6dbb1aad1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "92cef4f400d15a1c4ee03187af523660d2245c01582fd7d2b06299d6dbb1aad1",
+  "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": "75385294a36c27b5f9e3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bca1a1bbda111592518/workflow.json b/records/sd1/sweep/8bca1a1bbda111592518/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a61936fe44b8dd0612cac14c89e643db2fe194f
--- /dev/null
+++ b/records/sd1/sweep/8bca1a1bbda111592518/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 8bca1a1bbda111592518)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/8bd7853328c0a731c1a7/measurements.json b/records/sd1/sweep/8bd7853328c0a731c1a7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0504004a5507ce22ec2e6ea2c2eff3ab67b46b96
--- /dev/null
+++ b/records/sd1/sweep/8bd7853328c0a731c1a7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8bd7853328c0a731c1a7",
+ "values": {
+  "std": {
+   "value": 43.856,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01128,
+   "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": 10.627,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.050046,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.170424,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.044299,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.021054,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9748,
+   "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": 0.078,
+   "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/8bd7853328c0a731c1a7/record.json b/records/sd1/sweep/8bd7853328c0a731c1a7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..186c6100637f76f29bfac4176b0e1a8bde6fec2d
--- /dev/null
+++ b/records/sd1/sweep/8bd7853328c0a731c1a7/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8bd7853328c0a731c1a7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:47:08Z",
+ "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.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "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": [
+   "in.hi",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.5,
+    "applied": 0.5,
+    "reason": "outside safe range [0.6, 1.6] 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": "210878fb7a5ec97aa009"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bd7853328c0a731c1a7/workflow.json b/records/sd1/sweep/8bd7853328c0a731c1a7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a36bc2f9f99458636441afb3fa92dfab551e5e15
--- /dev/null
+++ b/records/sd1/sweep/8bd7853328c0a731c1a7/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 8bd7853328c0a731c1a7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\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/8bd7cf2bdcd8cbad6b36/measurements.json b/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9120f9170391be5ec8b18080ffac45f1e66c0377
--- /dev/null
+++ b/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8bd7cf2bdcd8cbad6b36",
+ "values": {
+  "cosine_distance": {
+   "value": 0.361931,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.617942,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.711069,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.388469,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 37.202,
+   "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": [],
+   "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": 41.696,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8724,
+   "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": 0.376,
+   "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/8bd7cf2bdcd8cbad6b36/record.json b/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..57a8939a946fa0b73a9ff5292efa15faeeb95706
--- /dev/null
+++ b/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8bd7cf2bdcd8cbad6b36",
+ "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": 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/2ea64ee8285a31d9b3ad5d56bd13ac1f0af264b03bd76a4ecb02dc52553f7908.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2ea64ee8285a31d9b3ad5d56bd13ac1f0af264b03bd76a4ecb02dc52553f7908",
+  "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": "e78614b5240432cc4137"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/workflow.json b/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f932191adf020385d736761bb0c05241788d407
--- /dev/null
+++ b/records/sd1/sweep/8bd7cf2bdcd8cbad6b36/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 8bd7cf2bdcd8cbad6b36)"
+  },
+  "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\": 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/8be3df023662ff94906d/measurements.json b/records/sd1/sweep/8be3df023662ff94906d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c19f6056112259a0ee5fbeecb911a39178040969
--- /dev/null
+++ b/records/sd1/sweep/8be3df023662ff94906d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8be3df023662ff94906d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.203089,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.380821,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.304083,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.159792,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.682,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00281,
+   "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": 21.044,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.951,
+   "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": 0.5605,
+   "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/8be3df023662ff94906d/record.json b/records/sd1/sweep/8be3df023662ff94906d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bfb0ece0bae58d35ee1f297839d628f7190c9ab
--- /dev/null
+++ b/records/sd1/sweep/8be3df023662ff94906d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8be3df023662ff94906d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:02Z",
+ "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.11.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.hi",
+   "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/7774cf775465fbbdb32114b7c196a3a9eed8e05b93fcb7cba18fea7547e5044f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7774cf775465fbbdb32114b7c196a3a9eed8e05b93fcb7cba18fea7547e5044f",
+  "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": "1ad60046d20fefd7a80e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8be3df023662ff94906d/workflow.json b/records/sd1/sweep/8be3df023662ff94906d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..94a31fc537fb4b37d9ecaef07564cfb2306a07dd
--- /dev/null
+++ b/records/sd1/sweep/8be3df023662ff94906d/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 8be3df023662ff94906d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.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/8be6c01c6f5329a69788/measurements.json b/records/sd1/sweep/8be6c01c6f5329a69788/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a88cc8c8622f2feb2d3a243bd5b6ce08c522702
--- /dev/null
+++ b/records/sd1/sweep/8be6c01c6f5329a69788/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8be6c01c6f5329a69788",
+ "values": {
+  "cosine_distance": {
+   "value": 0.209414,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.282065,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.105722,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.065143,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 41.736,
+   "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": 29.286,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9286,
+   "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": 0.2131,
+   "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/8be6c01c6f5329a69788/record.json b/records/sd1/sweep/8be6c01c6f5329a69788/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..246c76f60f7191d6f3556bf73943958863789700
--- /dev/null
+++ b/records/sd1/sweep/8be6c01c6f5329a69788/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8be6c01c6f5329a69788",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:59Z",
+ "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": "output_blocks.5.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "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/773bb321ec842b2de1d1b2d6d285ccbd1247d05215020e6018799ab21260af48.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "773bb321ec842b2de1d1b2d6d285ccbd1247d05215020e6018799ab21260af48",
+  "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": "670a17300a835ee30293"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8be6c01c6f5329a69788/workflow.json b/records/sd1/sweep/8be6c01c6f5329a69788/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b7d06b5f5badccd72863221748eb1a87bea9e5f
--- /dev/null
+++ b/records/sd1/sweep/8be6c01c6f5329a69788/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 8be6c01c6f5329a69788)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.norm\",\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/8befbd9c844aaf275e69/measurements.json b/records/sd1/sweep/8befbd9c844aaf275e69/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb4f092d06c7f44de5457f997052479b418b4c4e
--- /dev/null
+++ b/records/sd1/sweep/8befbd9c844aaf275e69/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8befbd9c844aaf275e69",
+ "values": {
+  "cosine_distance": {
+   "value": 0.013824,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.028918,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.021954,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004412,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.911,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00529,
+   "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": 4.761,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0118,
+   "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": 0.122,
+   "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/8befbd9c844aaf275e69/record.json b/records/sd1/sweep/8befbd9c844aaf275e69/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..779f50a51ab62b8e9ec699a711ae874127a0cfbe
--- /dev/null
+++ b/records/sd1/sweep/8befbd9c844aaf275e69/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8befbd9c844aaf275e69",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:34Z",
+ "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": 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.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.hi",
+   "kind": "cross_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/6bf5d1f27f79499caa5ac0a3287baa56b34f0ca03f3ae598c4d301fe11cafd5a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6bf5d1f27f79499caa5ac0a3287baa56b34f0ca03f3ae598c4d301fe11cafd5a",
+  "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": "cbe0cee305c50cb89043"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8befbd9c844aaf275e69/workflow.json b/records/sd1/sweep/8befbd9c844aaf275e69/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2acbcaf6fc3edc483015d1c8ef7cf759e79fb26e
--- /dev/null
+++ b/records/sd1/sweep/8befbd9c844aaf275e69/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 8befbd9c844aaf275e69)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.transformer_blocks.0.attn2.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/8bf1f75cf81aed84288b/measurements.json b/records/sd1/sweep/8bf1f75cf81aed84288b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b5041d7b2547cc9c656977c15dc187015479d15
--- /dev/null
+++ b/records/sd1/sweep/8bf1f75cf81aed84288b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8bf1f75cf81aed84288b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.16397,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.150908,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.196469,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.167734,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.058,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00567,
+   "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": 15.831,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9376,
+   "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": 0.5501,
+   "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/8bf1f75cf81aed84288b/record.json b/records/sd1/sweep/8bf1f75cf81aed84288b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..662ac2b549a1d0e159d3b9ab6ef165f8bf1280ab
--- /dev/null
+++ b/records/sd1/sweep/8bf1f75cf81aed84288b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8bf1f75cf81aed84288b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:40Z",
+ "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.2.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "mid",
+   "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/122b9172bd205908250123e5339a8bb76d274ac64c592022fe301ec37d940e5b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "122b9172bd205908250123e5339a8bb76d274ac64c592022fe301ec37d940e5b",
+  "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"
+    ],
+    "previous_id": "134f37d8d01b3d8c48dc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bf1f75cf81aed84288b/workflow.json b/records/sd1/sweep/8bf1f75cf81aed84288b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0b6002f8f6aaf1f2a882a2e8c6f547d5cc1fd7c
--- /dev/null
+++ b/records/sd1/sweep/8bf1f75cf81aed84288b/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 8bf1f75cf81aed84288b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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": [
+    "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/8bf43fbd1f3e099634d4/measurements.json b/records/sd1/sweep/8bf43fbd1f3e099634d4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6606a92c0b062debd07c3c5a07403a1025240dc
--- /dev/null
+++ b/records/sd1/sweep/8bf43fbd1f3e099634d4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8bf43fbd1f3e099634d4",
+ "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/8bf43fbd1f3e099634d4/record.json b/records/sd1/sweep/8bf43fbd1f3e099634d4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb623cc5106d9eaf0d7ce0de7fe6d5ad35feaf51
--- /dev/null
+++ b/records/sd1/sweep/8bf43fbd1f3e099634d4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8bf43fbd1f3e099634d4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:29Z",
+ "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": 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.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "mid",
+   "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/56573fbaa7414024e4cb41f895b4482f49e808adeb12e54e1ad84298202bf82c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "56573fbaa7414024e4cb41f895b4482f49e808adeb12e54e1ad84298202bf82c",
+  "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": "95a93fdea521780f5084"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bf43fbd1f3e099634d4/workflow.json b/records/sd1/sweep/8bf43fbd1f3e099634d4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..da5d5b6240d29d3983fc89808a890a6cfebd87ee
--- /dev/null
+++ b/records/sd1/sweep/8bf43fbd1f3e099634d4/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 8bf43fbd1f3e099634d4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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/8bf68287f173c213a536/measurements.json b/records/sd1/sweep/8bf68287f173c213a536/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..07668c61b670d54f7693cfdd8e35499955749b87
--- /dev/null
+++ b/records/sd1/sweep/8bf68287f173c213a536/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8bf68287f173c213a536",
+ "values": {
+  "cosine_distance": {
+   "value": 0.134415,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.200493,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.066883,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.04729,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.443,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01512,
+   "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": 19.22,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0015,
+   "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": 0.0601,
+   "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/8bf68287f173c213a536/record.json b/records/sd1/sweep/8bf68287f173c213a536/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..72a4f265b8b67fda3298c52a656373342cc6f24b
--- /dev/null
+++ b/records/sd1/sweep/8bf68287f173c213a536/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8bf68287f173c213a536",
+ "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": 1.5
+   },
+   "group": "out.hi",
+   "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/3ac16a80ee8801f55c119a83fb8cafd7e7168007b7783160d1b320ebd7730020.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3ac16a80ee8801f55c119a83fb8cafd7e7168007b7783160d1b320ebd7730020",
+  "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": "7158e8199e2d4ae49223"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8bf68287f173c213a536/workflow.json b/records/sd1/sweep/8bf68287f173c213a536/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..62c44bf52d41e7abdaee641b6365c86520c7ee50
--- /dev/null
+++ b/records/sd1/sweep/8bf68287f173c213a536/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 8bf68287f173c213a536)"
+  },
+  "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\": 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/8c0d11cf31ed8bd7257a/measurements.json b/records/sd1/sweep/8c0d11cf31ed8bd7257a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..62c59ff5b364f2939cf102772f7ec8be7a3171cc
--- /dev/null
+++ b/records/sd1/sweep/8c0d11cf31ed8bd7257a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c0d11cf31ed8bd7257a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.105978,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.106086,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.037355,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.067167,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.336,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00737,
+   "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": 14.43,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9955,
+   "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": 0.498,
+   "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/8c0d11cf31ed8bd7257a/record.json b/records/sd1/sweep/8c0d11cf31ed8bd7257a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfe720463a68f12e0e19ab31e99d912fdedbca56
--- /dev/null
+++ b/records/sd1/sweep/8c0d11cf31ed8bd7257a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8c0d11cf31ed8bd7257a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:57Z",
+ "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.10.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "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/a274d83402c3453cd30ce6f69d8faefd56272b6095f8c13aa2cd46a59ff04546.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a274d83402c3453cd30ce6f69d8faefd56272b6095f8c13aa2cd46a59ff04546",
+  "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": "5579905e0720643aa699"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c0d11cf31ed8bd7257a/workflow.json b/records/sd1/sweep/8c0d11cf31ed8bd7257a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..218366d9809702ba6d8ec60edf98fac189fa8775
--- /dev/null
+++ b/records/sd1/sweep/8c0d11cf31ed8bd7257a/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 8c0d11cf31ed8bd7257a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.transformer_blocks.0.norm2\",\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/8c0ded19e4335ff17e8e/measurements.json b/records/sd1/sweep/8c0ded19e4335ff17e8e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebce9f6151d9282dbbfbb51feedf96ab84061c10
--- /dev/null
+++ b/records/sd1/sweep/8c0ded19e4335ff17e8e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c0ded19e4335ff17e8e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.132725,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.17882,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.059888,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.03217,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.905,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00321,
+   "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": 21.317,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9751,
+   "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": 0.2293,
+   "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/8c0ded19e4335ff17e8e/record.json b/records/sd1/sweep/8c0ded19e4335ff17e8e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..caca99dd55e05db71d1fa5f6f2866839413fa6ab
--- /dev/null
+++ b/records/sd1/sweep/8c0ded19e4335ff17e8e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c0ded19e4335ff17e8e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:10Z",
+ "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.5.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "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/679fc7e87390494d236358803072708e9b2efb4dc2ab85c0540bd74435300514.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "679fc7e87390494d236358803072708e9b2efb4dc2ab85c0540bd74435300514",
+  "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": "fd48b98fd6fed8085726"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c0ded19e4335ff17e8e/workflow.json b/records/sd1/sweep/8c0ded19e4335ff17e8e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..add925f9ad0020cd6a824609aff0300cdfec1159
--- /dev/null
+++ b/records/sd1/sweep/8c0ded19e4335ff17e8e/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 8c0ded19e4335ff17e8e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.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/8c108ddeffeefa20047d/measurements.json b/records/sd1/sweep/8c108ddeffeefa20047d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fa2a3210831da5a3c26c67e5dada04ab7883dc6
--- /dev/null
+++ b/records/sd1/sweep/8c108ddeffeefa20047d/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "8c108ddeffeefa20047d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.723132,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.839516,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.946303,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.546755,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 91.039,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02493,
+   "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": 101.675,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.7801,
+   "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": 0.9868,
+   "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/8c108ddeffeefa20047d/record.json b/records/sd1/sweep/8c108ddeffeefa20047d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..63439277c9136765cf0317953849d3f89df044ef
--- /dev/null
+++ b/records/sd1/sweep/8c108ddeffeefa20047d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8c108ddeffeefa20047d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:03Z",
+ "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": "out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.conv",
+   "kind": "out_conv",
+   "module_type": "SiLU",
+   "container": "Sequential",
+   "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/f0c422e1ce096d8234f8a29291ae15e22fb97fa06bd0032d569992b5987faf95.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f0c422e1ce096d8234f8a29291ae15e22fb97fa06bd0032d569992b5987faf95",
+  "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": "ec5be5fbc9b09d4492d9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c108ddeffeefa20047d/workflow.json b/records/sd1/sweep/8c108ddeffeefa20047d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..55b51b1894e90ca01c379b8868a135e4fb7c9d6a
--- /dev/null
+++ b/records/sd1/sweep/8c108ddeffeefa20047d/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 8c108ddeffeefa20047d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"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": [
+    "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/8c1fa1fe404c8fb58ae7/measurements.json b/records/sd1/sweep/8c1fa1fe404c8fb58ae7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..71b0714d7dd5c9efc8cc6e94e230a6ee0a4ac026
--- /dev/null
+++ b/records/sd1/sweep/8c1fa1fe404c8fb58ae7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8c1fa1fe404c8fb58ae7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.492346,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.556053,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.364409,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.161189,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.569,
+   "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": 51.265,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9624,
+   "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": 0.3227,
+   "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/8c1fa1fe404c8fb58ae7/record.json b/records/sd1/sweep/8c1fa1fe404c8fb58ae7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ca0f9115d7f20828f234e69f83b2bffa89af6cd
--- /dev/null
+++ b/records/sd1/sweep/8c1fa1fe404c8fb58ae7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8c1fa1fe404c8fb58ae7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:01Z",
+ "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.1.0.in_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "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/ce6dc71274b803724bb04d4dd48f058976543b72695da5493192fbba77e31ea7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ce6dc71274b803724bb04d4dd48f058976543b72695da5493192fbba77e31ea7",
+  "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": "d550c212b4a10a0125d9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c1fa1fe404c8fb58ae7/workflow.json b/records/sd1/sweep/8c1fa1fe404c8fb58ae7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb5f6f147a5f9a84a6079d76a9818eb9f1fc0dff
--- /dev/null
+++ b/records/sd1/sweep/8c1fa1fe404c8fb58ae7/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 8c1fa1fe404c8fb58ae7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.in_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/8c204d5a719d3c35c9f9/measurements.json b/records/sd1/sweep/8c204d5a719d3c35c9f9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ec15dee3f8b7e464d87d79a878b78cec90b6bc3
--- /dev/null
+++ b/records/sd1/sweep/8c204d5a719d3c35c9f9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c204d5a719d3c35c9f9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.284997,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.4041,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.399848,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.188543,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.245,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00779,
+   "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": 32.148,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0073,
+   "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": 0.113,
+   "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/8c204d5a719d3c35c9f9/record.json b/records/sd1/sweep/8c204d5a719d3c35c9f9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0819c3daf5242f81f4ab999433a295ec129fa2c3
--- /dev/null
+++ b/records/sd1/sweep/8c204d5a719d3c35c9f9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8c204d5a719d3c35c9f9",
+ "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.3",
+   "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/27094896f1fa156a4824ec3942188ed6b8533223c9c8366577b7420b48303a89.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "27094896f1fa156a4824ec3942188ed6b8533223c9c8366577b7420b48303a89",
+  "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": "ba9ad42fddcfabd63fb5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c204d5a719d3c35c9f9/workflow.json b/records/sd1/sweep/8c204d5a719d3c35c9f9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f60da0acc8cfcfc827908945f6e77e9ef953730
--- /dev/null
+++ b/records/sd1/sweep/8c204d5a719d3c35c9f9/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 8c204d5a719d3c35c9f9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.3\",\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/8c27c779e4030102aa1d/measurements.json b/records/sd1/sweep/8c27c779e4030102aa1d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d31f40774ee948992cd4ceb63c6d2ff91fea70c7
--- /dev/null
+++ b/records/sd1/sweep/8c27c779e4030102aa1d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c27c779e4030102aa1d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.321104,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.353396,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.158832,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.056826,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.372,
+   "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": 37.56,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9584,
+   "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": 0.106,
+   "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/8c27c779e4030102aa1d/record.json b/records/sd1/sweep/8c27c779e4030102aa1d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f6d6c3b14a336e0972bb476a9937fd55367c294
--- /dev/null
+++ b/records/sd1/sweep/8c27c779e4030102aa1d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c27c779e4030102aa1d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:02Z",
+ "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.4.1.transformer_blocks.0.attn1.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "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/a2056d318c907458382ac8ca6b51f6d55fa74c69cff76b4e6fa601f222372b96.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a2056d318c907458382ac8ca6b51f6d55fa74c69cff76b4e6fa601f222372b96",
+  "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": "c2307fad6bf37faf34ce"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c27c779e4030102aa1d/workflow.json b/records/sd1/sweep/8c27c779e4030102aa1d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ed8cf8b04fd7f7a197033a4dc36c8090069ee51
--- /dev/null
+++ b/records/sd1/sweep/8c27c779e4030102aa1d/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 8c27c779e4030102aa1d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.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/8c2d4024bbe87406b260/measurements.json b/records/sd1/sweep/8c2d4024bbe87406b260/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1adc8cd789e8ccf59bc72bac9af5debbcd14dfb
--- /dev/null
+++ b/records/sd1/sweep/8c2d4024bbe87406b260/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c2d4024bbe87406b260",
+ "values": {
+  "std": {
+   "value": 45.292,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01265,
+   "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": 15.239,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.096017,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.2201,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.102344,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.071526,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0815,
+   "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": 0.0576,
+   "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/8c2d4024bbe87406b260/record.json b/records/sd1/sweep/8c2d4024bbe87406b260/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f166f1aa02c91df2d3a175fadf378dd4db3c8c2b
--- /dev/null
+++ b/records/sd1/sweep/8c2d4024bbe87406b260/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8c2d4024bbe87406b260",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:22:07Z",
+ "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.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.lo",
+   "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": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "4eb62d8b5c080c6bf598"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c2d4024bbe87406b260/workflow.json b/records/sd1/sweep/8c2d4024bbe87406b260/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a27539c2821ee6069abed4fcd96a4a1a84cb3433
--- /dev/null
+++ b/records/sd1/sweep/8c2d4024bbe87406b260/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 8c2d4024bbe87406b260)"
+  },
+  "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\": 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": "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/8c2f31512c7f26110915/measurements.json b/records/sd1/sweep/8c2f31512c7f26110915/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dd611f18503745fdfc049829db4f90d990e4138
--- /dev/null
+++ b/records/sd1/sweep/8c2f31512c7f26110915/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c2f31512c7f26110915",
+ "values": {
+  "std": {
+   "value": 41.102,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01748,
+   "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": 25.655,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.139443,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.224754,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.033223,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.022429,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9959,
+   "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": 0.0285,
+   "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/8c2f31512c7f26110915/record.json b/records/sd1/sweep/8c2f31512c7f26110915/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5608c2cfe2ce414573f959447f3f219e2312b75e
--- /dev/null
+++ b/records/sd1/sweep/8c2f31512c7f26110915/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8c2f31512c7f26110915",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:15:03Z",
+ "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": 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": 0.333,
+   "window": "early",
+   "bucket": "damp",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "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,
+   "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": "3f2dd45de45390b38d1b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c2f31512c7f26110915/workflow.json b/records/sd1/sweep/8c2f31512c7f26110915/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..05140bc5a8c0cfd8907e7db05aab7d4caef19846
--- /dev/null
+++ b/records/sd1/sweep/8c2f31512c7f26110915/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 8c2f31512c7f26110915)"
+  },
+  "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\": 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": 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/8c38e2fe7d2bc8e18a00/measurements.json b/records/sd1/sweep/8c38e2fe7d2bc8e18a00/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ac0a152dc8b310f17d4414fb8b258f82dab692b
--- /dev/null
+++ b/records/sd1/sweep/8c38e2fe7d2bc8e18a00/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c38e2fe7d2bc8e18a00",
+ "values": {
+  "cosine_distance": {
+   "value": 0.610687,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.478831,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.33065,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.133389,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.28,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00661,
+   "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": 52.303,
+   "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": 0.1258,
+   "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/8c38e2fe7d2bc8e18a00/record.json b/records/sd1/sweep/8c38e2fe7d2bc8e18a00/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1355a2da4007ffea76a9356eeed7fd5f8babf465
--- /dev/null
+++ b/records/sd1/sweep/8c38e2fe7d2bc8e18a00/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8c38e2fe7d2bc8e18a00",
+ "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.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.mid",
+   "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/f4d5abc4d6a951036dc877fcc3dbfa3ca4e628e05d168bc2fdebe4fd79782047.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f4d5abc4d6a951036dc877fcc3dbfa3ca4e628e05d168bc2fdebe4fd79782047",
+  "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": "f0645af0a4879594c020"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c38e2fe7d2bc8e18a00/workflow.json b/records/sd1/sweep/8c38e2fe7d2bc8e18a00/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a993684e870de42925382b198c752f52faccc5c
--- /dev/null
+++ b/records/sd1/sweep/8c38e2fe7d2bc8e18a00/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 8c38e2fe7d2bc8e18a00)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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/8c4e22e82b8eeb778d0f/measurements.json b/records/sd1/sweep/8c4e22e82b8eeb778d0f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d9f43b9419b37cae4750fd5c7218f78b17b8098e
--- /dev/null
+++ b/records/sd1/sweep/8c4e22e82b8eeb778d0f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c4e22e82b8eeb778d0f",
+ "values": {
+  "std": {
+   "value": 38.526,
+   "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"
+  },
+  "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": 18.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.156444,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.474232,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.895858,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.307653,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8169,
+   "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": 0.1928,
+   "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/8c4e22e82b8eeb778d0f/record.json b/records/sd1/sweep/8c4e22e82b8eeb778d0f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b733f695c7c262b56d1baa86abbc975dbf10b9cf
--- /dev/null
+++ b/records/sd1/sweep/8c4e22e82b8eeb778d0f/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8c4e22e82b8eeb778d0f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:01:15Z",
+ "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": 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": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "slight",
+   "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": [
+   "in.hi",
+   "rotate",
+   30.0,
+   "late"
+  ],
+  "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": "06c5045b81c4e4392d32"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c4e22e82b8eeb778d0f/workflow.json b/records/sd1/sweep/8c4e22e82b8eeb778d0f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..74039d80418b255b18f0c35b6c1daa3d7e940f33
--- /dev/null
+++ b/records/sd1/sweep/8c4e22e82b8eeb778d0f/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 8c4e22e82b8eeb778d0f)"
+  },
+  "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\": 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/8c500cd34848eb700ac4/measurements.json b/records/sd1/sweep/8c500cd34848eb700ac4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8aa6924b6666bc34f2b6750b4b57618a57e46423
--- /dev/null
+++ b/records/sd1/sweep/8c500cd34848eb700ac4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8c500cd34848eb700ac4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.16204,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.243988,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.265183,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.231536,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.743,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00431,
+   "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": 19.746,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8903,
+   "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": 0.7148,
+   "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/8c500cd34848eb700ac4/record.json b/records/sd1/sweep/8c500cd34848eb700ac4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..84c48bbddec1b1cf370b552f1379f5c797408e73
--- /dev/null
+++ b/records/sd1/sweep/8c500cd34848eb700ac4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8c500cd34848eb700ac4",
+ "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.6.0.in_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/332e8ea99ed2de5ec3cba4bce9862d9191a8c0a6a5a2db67211876ba9367818d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "332e8ea99ed2de5ec3cba4bce9862d9191a8c0a6a5a2db67211876ba9367818d",
+  "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": "bd8322933c826e142f51"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c500cd34848eb700ac4/workflow.json b/records/sd1/sweep/8c500cd34848eb700ac4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fee3c6f3cbbef1468c3a5b57b74fa17ca72dae89
--- /dev/null
+++ b/records/sd1/sweep/8c500cd34848eb700ac4/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 8c500cd34848eb700ac4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.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": [
+    "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/8c58fbaed59244876fd7/interpretations.jsonl b/records/sd1/sweep/8c58fbaed59244876fd7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e0dfd9954ecba5561c96a273134fa5c0e69a2ee7
--- /dev/null
+++ b/records/sd1/sweep/8c58fbaed59244876fd7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8c58fbaed59244876fd7"}, "kind": "caption", "value": "Dark, moody painted portrait of a woman with black hair against a near-black background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "23f577dd606339fe"}
+{"target": {"record": "8c58fbaed59244876fd7"}, "kind": "change", "value": "The image goes much darker and moodier. Hair turns black, the background deepens to near-black, the face becomes more gaunt and solemn, and the paper border is gone.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1caef10913fbc670"}
+{"target": {"record": "8c58fbaed59244876fd7"}, "kind": "keywords", "value": ["dark portrait", "moody", "low key", "black hair", "chiaroscuro"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d66bf84ede502820"}
+{"target": {"record": "8c58fbaed59244876fd7"}, "kind": "effect_tags", "value": ["dark", "subject_kept", "desaturated", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "77a8988a387f96a9"}
diff --git a/records/sd1/sweep/8c58fbaed59244876fd7/measurements.json b/records/sd1/sweep/8c58fbaed59244876fd7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d38b1478611bd97d3e022409a6e7a6359293ada0
--- /dev/null
+++ b/records/sd1/sweep/8c58fbaed59244876fd7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c58fbaed59244876fd7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.420211,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.436245,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.175746,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.066914,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 35.271,
+   "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": [],
+   "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": 58.215,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9772,
+   "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": 0.0917,
+   "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/8c58fbaed59244876fd7/record.json b/records/sd1/sweep/8c58fbaed59244876fd7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..63cb311701f02e233343e012acd9ca464d3cd0c0
--- /dev/null
+++ b/records/sd1/sweep/8c58fbaed59244876fd7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c58fbaed59244876fd7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:36Z",
+ "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": "input_blocks.1.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.hi",
+   "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/c67643df6f2e040ef5cca59123bf4df0cd0f5d536dfe5960b716a29438b1baaa.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c67643df6f2e040ef5cca59123bf4df0cd0f5d536dfe5960b716a29438b1baaa",
+  "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": "dbcdf2158894b47204ad"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c58fbaed59244876fd7/workflow.json b/records/sd1/sweep/8c58fbaed59244876fd7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bf50b19a1ec88cc957ad1aa58d0fa8a6ebf6788
--- /dev/null
+++ b/records/sd1/sweep/8c58fbaed59244876fd7/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 8c58fbaed59244876fd7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.emb_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/8c5c3c1c3f1b881b12fd/interpretations.jsonl b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..adc97cc211ed7b69cfcc9be513ddc1cafa05f085
--- /dev/null
+++ b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8c5c3c1c3f1b881b12fd"}, "kind": "caption", "value": "Portrait of a young East Asian woman with sleek black bob on grey board", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "49e046759afcda4d"}
+{"target": {"record": "8c5c3c1c3f1b881b12fd"}, "kind": "change", "value": "The subject is redrawn as a different young woman with a sleek, straight black bob, softer features and a more realistic rendering. The framing stays similar.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d663c7f486782c26"}
+{"target": {"record": "8c5c3c1c3f1b881b12fd"}, "kind": "keywords", "value": ["portrait", "black bob", "different face", "sleek hair", "realistic"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8928f02b5c838ee8"}
+{"target": {"record": "8c5c3c1c3f1b881b12fd"}, "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": "2026-10-05T21:16:16Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6e025c417eaad518"}
diff --git a/records/sd1/sweep/8c5c3c1c3f1b881b12fd/measurements.json b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..37e746dfc1ef7422e3540baa3ae644d9b1646f31
--- /dev/null
+++ b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8c5c3c1c3f1b881b12fd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.200469,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.243725,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.108616,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.094954,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.762,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00518,
+   "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": 21.129,
+   "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": 0.0522,
+   "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/8c5c3c1c3f1b881b12fd/record.json b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..73bbea54eedb4ac23388a0ff3c3badf584cac1e0
--- /dev/null
+++ b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c5c3c1c3f1b881b12fd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:35Z",
+ "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": "input_blocks.8.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "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/b25135a95ae07100ca988726e5899aeed429a26472da79abacd473f4256fe2b8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b25135a95ae07100ca988726e5899aeed429a26472da79abacd473f4256fe2b8",
+  "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": "bc9f0e17876922257269"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c5c3c1c3f1b881b12fd/workflow.json b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c26e8514a1adcdb71e9e6e5dfac0cd7e5a3818ac
--- /dev/null
+++ b/records/sd1/sweep/8c5c3c1c3f1b881b12fd/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 8c5c3c1c3f1b881b12fd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.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/8c66cb23b8faee203914/measurements.json b/records/sd1/sweep/8c66cb23b8faee203914/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd69f8732b8376f1eab93fdbed95118923f712f4
--- /dev/null
+++ b/records/sd1/sweep/8c66cb23b8faee203914/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c66cb23b8faee203914",
+ "values": {
+  "cosine_distance": {
+   "value": 0.277513,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.5476,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.367026,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.225645,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 36.799,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00516,
+   "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": 33.855,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0084,
+   "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": 0.4081,
+   "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/8c66cb23b8faee203914/record.json b/records/sd1/sweep/8c66cb23b8faee203914/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eccb81e36656df3347f7fd232aa26220f09a137e
--- /dev/null
+++ b/records/sd1/sweep/8c66cb23b8faee203914/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c66cb23b8faee203914",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:07Z",
+ "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.6.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "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/06d52d5d450b5c3bfc515e6aa93e7c85f075fa8126665291dc0a4eccaf25c76a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "06d52d5d450b5c3bfc515e6aa93e7c85f075fa8126665291dc0a4eccaf25c76a",
+  "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": "d04fa899bdf5524559e5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c66cb23b8faee203914/workflow.json b/records/sd1/sweep/8c66cb23b8faee203914/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f561b25444eb08007aeae062a4c65c8365fdcb4
--- /dev/null
+++ b/records/sd1/sweep/8c66cb23b8faee203914/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 8c66cb23b8faee203914)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.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/8c671ec24f820184ac80/measurements.json b/records/sd1/sweep/8c671ec24f820184ac80/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..36f25e61815f1f80d9f7e98c433e552a108078c6
--- /dev/null
+++ b/records/sd1/sweep/8c671ec24f820184ac80/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c671ec24f820184ac80",
+ "values": {
+  "std": {
+   "value": 28.571,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03335,
+   "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": 44.148,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.466619,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.565571,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.456409,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.251475,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9175,
+   "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": 0.0339,
+   "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/8c671ec24f820184ac80/record.json b/records/sd1/sweep/8c671ec24f820184ac80/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c27ea455f225b44da292061e97f7afd1de7d0b9
--- /dev/null
+++ b/records/sd1/sweep/8c671ec24f820184ac80/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "8c671ec24f820184ac80",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:55:32Z",
+ "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": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "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",
+   "rotate",
+   180.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 180.0,
+    "applied": 180.0,
+    "reason": "outside safe range [10.0, 90.0] for out.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": "29128b000e90773a04d9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c671ec24f820184ac80/workflow.json b/records/sd1/sweep/8c671ec24f820184ac80/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..165f5a7f25523c0befd71cf23d4f8362df5f6635
--- /dev/null
+++ b/records/sd1/sweep/8c671ec24f820184ac80/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 8c671ec24f820184ac80)"
+  },
+  "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\": 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": 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/8c80f2f9acc65460a62d/interpretations.jsonl b/records/sd1/sweep/8c80f2f9acc65460a62d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a0e127ac212d61e03452131db353eb58718959c3
--- /dev/null
+++ b/records/sd1/sweep/8c80f2f9acc65460a62d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8c80f2f9acc65460a62d"}, "kind": "caption", "value": "Painted portrait of a dark-haired young woman in a black turtleneck against a grey-blue backdrop", "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": "c06b8abe07d30042"}
+{"target": {"record": "8c80f2f9acc65460a62d"}, "kind": "change", "value": "The face gets a little more angular and the hair parts more to one side and spreads wider, while the palette and framing stay nearly the same.", "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": "f89b3abf4221345b"}
+{"target": {"record": "8c80f2f9acc65460a62d"}, "kind": "keywords", "value": ["portrait", "face variation", "hair change", "painted portrait", "subject kept"], "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": "39a4ff65531365f8"}
+{"target": {"record": "8c80f2f9acc65460a62d"}, "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": "2026-10-05T21:24:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7fc4385a0c9b0d22"}
diff --git a/records/sd1/sweep/8c80f2f9acc65460a62d/measurements.json b/records/sd1/sweep/8c80f2f9acc65460a62d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..446be88f74ca1ac3b80e8a8e88220f9b083bff1b
--- /dev/null
+++ b/records/sd1/sweep/8c80f2f9acc65460a62d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c80f2f9acc65460a62d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.070222,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.112331,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.039322,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010734,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.649,
+   "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": 13.561,
+   "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": 0.1357,
+   "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/8c80f2f9acc65460a62d/record.json b/records/sd1/sweep/8c80f2f9acc65460a62d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e960286cc2b7b049b839cec6c67b7d4bab6c0dc
--- /dev/null
+++ b/records/sd1/sweep/8c80f2f9acc65460a62d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c80f2f9acc65460a62d",
+ "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.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "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/dd8b89f98860e39dfe232c6e00549808199bbaa31d45e75d914708c564e7a057.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "dd8b89f98860e39dfe232c6e00549808199bbaa31d45e75d914708c564e7a057",
+  "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": "63dedcf55f4267777ac9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c80f2f9acc65460a62d/workflow.json b/records/sd1/sweep/8c80f2f9acc65460a62d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fb2523bb680cb74fe190e90550ee3d6028cd9b7
--- /dev/null
+++ b/records/sd1/sweep/8c80f2f9acc65460a62d/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 8c80f2f9acc65460a62d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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/8c87210a822235993fb7/measurements.json b/records/sd1/sweep/8c87210a822235993fb7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..96d6063a49be8a6745154ba09c13f4afc52d96c1
--- /dev/null
+++ b/records/sd1/sweep/8c87210a822235993fb7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8c87210a822235993fb7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.236907,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.249996,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.161259,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027238,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.921,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00714,
+   "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": 33.267,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9371,
+   "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": 0.3025,
+   "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/8c87210a822235993fb7/record.json b/records/sd1/sweep/8c87210a822235993fb7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0293701ab03aa5fa6eecaed4c086a0d633e06f0
--- /dev/null
+++ b/records/sd1/sweep/8c87210a822235993fb7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c87210a822235993fb7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:01Z",
+ "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.2.0.out_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/ea2a812ddfdfc780e2ac0c405530be931e5a225813b5b06ed046e81b5db24e3f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ea2a812ddfdfc780e2ac0c405530be931e5a225813b5b06ed046e81b5db24e3f",
+  "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": "42cce1835ec42477e770"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c87210a822235993fb7/workflow.json b/records/sd1/sweep/8c87210a822235993fb7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0816689830d9931c456c0f8c32ba06a4d434d76
--- /dev/null
+++ b/records/sd1/sweep/8c87210a822235993fb7/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 8c87210a822235993fb7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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/8c8ac815b9ffc06591f2/measurements.json b/records/sd1/sweep/8c8ac815b9ffc06591f2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..45c04e843fe99dadc024270b49dc03b36e3526dc
--- /dev/null
+++ b/records/sd1/sweep/8c8ac815b9ffc06591f2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8c8ac815b9ffc06591f2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.145892,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.182916,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.136701,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.070408,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.155,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00357,
+   "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": 17.662,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.937,
+   "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": 0.0579,
+   "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/8c8ac815b9ffc06591f2/record.json b/records/sd1/sweep/8c8ac815b9ffc06591f2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..471589d64ce5870c8ec4970d00ce1e865f2d9f58
--- /dev/null
+++ b/records/sd1/sweep/8c8ac815b9ffc06591f2/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c8ac815b9ffc06591f2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:46Z",
+ "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": "output_blocks.0.0.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.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/8cf7b71ddc0bacaf68fea653c345804910617d79758f60ffa13c5514f9452720.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8cf7b71ddc0bacaf68fea653c345804910617d79758f60ffa13c5514f9452720",
+  "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": "a34628f7e0729c6445e7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c8ac815b9ffc06591f2/workflow.json b/records/sd1/sweep/8c8ac815b9ffc06591f2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..96d3a3adc3409cf18507577d4170435a60e454da
--- /dev/null
+++ b/records/sd1/sweep/8c8ac815b9ffc06591f2/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 8c8ac815b9ffc06591f2)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.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/8c8ee2fd385074b4b509/measurements.json b/records/sd1/sweep/8c8ee2fd385074b4b509/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c517c9936d6e5d8425986b4f055e66928b7abc3a
--- /dev/null
+++ b/records/sd1/sweep/8c8ee2fd385074b4b509/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8c8ee2fd385074b4b509",
+ "values": {
+  "cosine_distance": {
+   "value": 0.020141,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.035889,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.020122,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.006351,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.799,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0048,
+   "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": 5.708,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9888,
+   "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": 0.1589,
+   "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/8c8ee2fd385074b4b509/record.json b/records/sd1/sweep/8c8ee2fd385074b4b509/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bd506ce6b60f582085c9d5d1fc07eba5a7c40eeb
--- /dev/null
+++ b/records/sd1/sweep/8c8ee2fd385074b4b509/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c8ee2fd385074b4b509",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:12Z",
+ "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": 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_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.mid",
+   "kind": "cross_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/993f79b8fa4862bc560aef5cd5587baf0535809f5c15c161a9f94d9818f299c6.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "993f79b8fa4862bc560aef5cd5587baf0535809f5c15c161a9f94d9818f299c6",
+  "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": "b3d4415dff690d7df4ea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c8ee2fd385074b4b509/workflow.json b/records/sd1/sweep/8c8ee2fd385074b4b509/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..05e12e7b1343269e171c0e780c0264754c860a62
--- /dev/null
+++ b/records/sd1/sweep/8c8ee2fd385074b4b509/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 8c8ee2fd385074b4b509)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.attn2.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/8c988792fa3e42114a78/measurements.json b/records/sd1/sweep/8c988792fa3e42114a78/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae7950d0fcf34aaaedb753bf6229e4dd6a2af40c
--- /dev/null
+++ b/records/sd1/sweep/8c988792fa3e42114a78/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8c988792fa3e42114a78",
+ "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/8c988792fa3e42114a78/record.json b/records/sd1/sweep/8c988792fa3e42114a78/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..79b4288a6692dc82f424d81e939636590d400d91
--- /dev/null
+++ b/records/sd1/sweep/8c988792fa3e42114a78/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8c988792fa3e42114a78",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:53Z",
+ "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": 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.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "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/86665ba71f3b770e7feab5e95556766ea150fc24f6861778814c5d33643db703.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "86665ba71f3b770e7feab5e95556766ea150fc24f6861778814c5d33643db703",
+  "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": "d27c87440976f53ec11a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8c988792fa3e42114a78/workflow.json b/records/sd1/sweep/8c988792fa3e42114a78/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a50cbbb96c8c89ddad25a191557c496599838ae1
--- /dev/null
+++ b/records/sd1/sweep/8c988792fa3e42114a78/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 8c988792fa3e42114a78)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.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": [
+    "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/8ca6bc238cdb0cb5f019/interpretations.jsonl b/records/sd1/sweep/8ca6bc238cdb0cb5f019/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..9b753ef18443066fd1a13e55d298457883fc778a
--- /dev/null
+++ b/records/sd1/sweep/8ca6bc238cdb0cb5f019/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ca6bc238cdb0cb5f019"}, "kind": "caption", "value": "Dark blue neon street at night with glowing signs, crowds and wet cobblestones", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1ace659660386c55"}
+{"target": {"record": "8ca6bc238cdb0cb5f019"}, "kind": "change", "value": "Recomposed neon street with darker, bluer lighting, new signs, a cobbled reflective road and a deeper vanishing point.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f46012ac67b53d72"}
+{"target": {"record": "8ca6bc238cdb0cb5f019"}, "kind": "keywords", "value": ["neon street", "blue night", "cobblestones", "crowd", "cyberpunk"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "99f950f02bf59c15"}
+{"target": {"record": "8ca6bc238cdb0cb5f019"}, "kind": "effect_tags", "value": ["subject_kept", "new_composition", "dark", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:44:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2b141cec84588401"}
diff --git a/records/sd1/sweep/8ca6bc238cdb0cb5f019/measurements.json b/records/sd1/sweep/8ca6bc238cdb0cb5f019/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f4f34f9b96f399fba2939b27f81228488a11a96
--- /dev/null
+++ b/records/sd1/sweep/8ca6bc238cdb0cb5f019/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ca6bc238cdb0cb5f019",
+ "values": {
+  "std": {
+   "value": 33.231,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03473,
+   "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": 40.296,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.363789,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.534969,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.192029,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.043869,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0265,
+   "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": 0.3822,
+   "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/8ca6bc238cdb0cb5f019/record.json b/records/sd1/sweep/8ca6bc238cdb0cb5f019/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4de513e817a602a0c3110d17a7c7e5344db65d26
--- /dev/null
+++ b/records/sd1/sweep/8ca6bc238cdb0cb5f019/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8ca6bc238cdb0cb5f019",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:12:30Z",
+ "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.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "zero",
+   "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.hi",
+   "multiply",
+   0.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.6, 1.6] 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": "32dc15844988c6f74151"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ca6bc238cdb0cb5f019/workflow.json b/records/sd1/sweep/8ca6bc238cdb0cb5f019/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5a89dbdde062445c4edbf0be21b6cf539020ae2
--- /dev/null
+++ b/records/sd1/sweep/8ca6bc238cdb0cb5f019/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 8ca6bc238cdb0cb5f019)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\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": 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/8cb872fa9623f545d7a0/measurements.json b/records/sd1/sweep/8cb872fa9623f545d7a0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..897f7a8859af895bfc963ca319c2287e66587077
--- /dev/null
+++ b/records/sd1/sweep/8cb872fa9623f545d7a0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8cb872fa9623f545d7a0",
+ "values": {
+  "std": {
+   "value": 51.146,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01197,
+   "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": 31.62,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.243902,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.304853,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.080191,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.060617,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.003,
+   "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": 0.1923,
+   "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/8cb872fa9623f545d7a0/record.json b/records/sd1/sweep/8cb872fa9623f545d7a0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..58ae902fdda6d4ad3e68b153232613d5efbaf99f
--- /dev/null
+++ b/records/sd1/sweep/8cb872fa9623f545d7a0/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8cb872fa9623f545d7a0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:40:11Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "half",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "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",
+   "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": "93e6f804a39ee415bdb3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cb872fa9623f545d7a0/workflow.json b/records/sd1/sweep/8cb872fa9623f545d7a0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f094d3bf2846011e2d4ce449d829a56473314bc
--- /dev/null
+++ b/records/sd1/sweep/8cb872fa9623f545d7a0/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 8cb872fa9623f545d7a0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\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": "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/8cc22ffd3405fe614484/measurements.json b/records/sd1/sweep/8cc22ffd3405fe614484/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..de1251740291f7b14969dd398ab84e08f1135872
--- /dev/null
+++ b/records/sd1/sweep/8cc22ffd3405fe614484/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8cc22ffd3405fe614484",
+ "values": {
+  "cosine_distance": {
+   "value": 0.384417,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.420851,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.281247,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.116036,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.115,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0086,
+   "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": 42.881,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9797,
+   "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": 0.348,
+   "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/8cc22ffd3405fe614484/record.json b/records/sd1/sweep/8cc22ffd3405fe614484/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..59aa23755ed88e620c377b1efe361fbf066835e9
--- /dev/null
+++ b/records/sd1/sweep/8cc22ffd3405fe614484/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8cc22ffd3405fe614484",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:15Z",
+ "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": "input_blocks.4.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/efc9d3f1c92bfcd8f7b0852b851a131ef2426d3933f6798ef60cf18066542e77.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "efc9d3f1c92bfcd8f7b0852b851a131ef2426d3933f6798ef60cf18066542e77",
+  "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": "add8a08bfbdff49dcbd1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cc22ffd3405fe614484/workflow.json b/records/sd1/sweep/8cc22ffd3405fe614484/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0be1fd6ca21a655b7039b4b96cd52694e11cb59b
--- /dev/null
+++ b/records/sd1/sweep/8cc22ffd3405fe614484/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 8cc22ffd3405fe614484)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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/8ccc91c1b22474384303/interpretations.jsonl b/records/sd1/sweep/8ccc91c1b22474384303/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..551412003cd3cc2ebc585fcabab5a91c78ff8bae
--- /dev/null
+++ b/records/sd1/sweep/8ccc91c1b22474384303/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ccc91c1b22474384303"}, "kind": "caption", "value": "Portrait of woman with windswept wavy hair on pale grey-green brushed background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "00334cf40243a954"}
+{"target": {"record": "8ccc91c1b22474384303"}, "kind": "change", "value": "The face stays similar but the hair becomes airier and windswept, and the background lightens to a pale brushed grey-green with the frame fading out.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f00a6f32aeabe464"}
+{"target": {"record": "8ccc91c1b22474384303"}, "kind": "keywords", "value": ["windswept hair", "lighter background", "airy", "painterly portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "113a0991144fb494"}
+{"target": {"record": "8ccc91c1b22474384303"}, "kind": "effect_tags", "value": ["subject_kept", "bright", "painterly", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3a88134aa7712ee8"}
diff --git a/records/sd1/sweep/8ccc91c1b22474384303/measurements.json b/records/sd1/sweep/8ccc91c1b22474384303/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cd4e4f6e068a14df0f9c01bfb6dd23c53b689f8
--- /dev/null
+++ b/records/sd1/sweep/8ccc91c1b22474384303/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ccc91c1b22474384303",
+ "values": {
+  "cosine_distance": {
+   "value": 0.170666,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.21839,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.081901,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.033075,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.741,
+   "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": [],
+   "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": 21.97,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9703,
+   "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": 0.1655,
+   "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/8ccc91c1b22474384303/record.json b/records/sd1/sweep/8ccc91c1b22474384303/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c0e5037592cb9799adcf5e1c82bd5d43f6026a3
--- /dev/null
+++ b/records/sd1/sweep/8ccc91c1b22474384303/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ccc91c1b22474384303",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:13Z",
+ "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": "middle_block.1.transformer_blocks.0.ff.net.0.proj",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "ff",
+   "module_type": "Linear",
+   "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/2b1997312f7720d77f38edec1491bbe36a8c0dcb2360af6897f861c5a036ba1c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2b1997312f7720d77f38edec1491bbe36a8c0dcb2360af6897f861c5a036ba1c",
+  "type_path": "UNetModel.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": "e49667c57c3149d2d28e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ccc91c1b22474384303/workflow.json b/records/sd1/sweep/8ccc91c1b22474384303/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..82720fb457cd4f275c418c415a4e92d287a71877
--- /dev/null
+++ b/records/sd1/sweep/8ccc91c1b22474384303/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 8ccc91c1b22474384303)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.ff.net.0.proj\",\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/8ccfa61f65571835c7c1/measurements.json b/records/sd1/sweep/8ccfa61f65571835c7c1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..029ecca8bbbef276cb9ca86bd3717a567bcddf36
--- /dev/null
+++ b/records/sd1/sweep/8ccfa61f65571835c7c1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ccfa61f65571835c7c1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.240976,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.29823,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.224121,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.207817,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.42,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00294,
+   "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": 25.859,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9747,
+   "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": 0.0673,
+   "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/8ccfa61f65571835c7c1/record.json b/records/sd1/sweep/8ccfa61f65571835c7c1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7802d0c8384747ffe68ec1072223839b064e9992
--- /dev/null
+++ b/records/sd1/sweep/8ccfa61f65571835c7c1/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8ccfa61f65571835c7c1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:12Z",
+ "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.5.0.out_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.lo",
+   "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/2e04b9de198c0f713036ea938c609488dee81e2310d8ceb368059dc333a2b31e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2e04b9de198c0f713036ea938c609488dee81e2310d8ceb368059dc333a2b31e",
+  "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": "c5f71cc272b313cfee8f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ccfa61f65571835c7c1/workflow.json b/records/sd1/sweep/8ccfa61f65571835c7c1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..36fb1a78f3cade3bd55a0f636f5aebcb39b106ac
--- /dev/null
+++ b/records/sd1/sweep/8ccfa61f65571835c7c1/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 8ccfa61f65571835c7c1)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.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/8cd548030f5993ce052a/measurements.json b/records/sd1/sweep/8cd548030f5993ce052a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8a20c1d546c793de1f6086a5f755003887fd69e
--- /dev/null
+++ b/records/sd1/sweep/8cd548030f5993ce052a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8cd548030f5993ce052a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.320212,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.38619,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.163636,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.113696,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.447,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0045,
+   "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": 35.529,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.974,
+   "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": 0.3122,
+   "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/8cd548030f5993ce052a/record.json b/records/sd1/sweep/8cd548030f5993ce052a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1ed7de4e553e0848c5e255bd2fc6a86086e12fe
--- /dev/null
+++ b/records/sd1/sweep/8cd548030f5993ce052a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8cd548030f5993ce052a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:52Z",
+ "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.11.0.out_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "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/911143b74fb7b336c13ce65cb18c7fe5288c70b203d3b031eb8c6fcd367a61b4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "911143b74fb7b336c13ce65cb18c7fe5288c70b203d3b031eb8c6fcd367a61b4",
+  "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": "f8a633be25ef2ca2ce90"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cd548030f5993ce052a/workflow.json b/records/sd1/sweep/8cd548030f5993ce052a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1b2a62e24666c18af9ef6a1c984d2ef74d4502d
--- /dev/null
+++ b/records/sd1/sweep/8cd548030f5993ce052a/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 8cd548030f5993ce052a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.out_layers.2\",\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/8cd93b9c47222e922271/measurements.json b/records/sd1/sweep/8cd93b9c47222e922271/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4fd1d974750e6778e214e1eac31b66f15b603bf8
--- /dev/null
+++ b/records/sd1/sweep/8cd93b9c47222e922271/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8cd93b9c47222e922271",
+ "values": {
+  "cosine_distance": {
+   "value": 0.583342,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.583304,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.459127,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.19107,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.191,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00427,
+   "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": 58.198,
+   "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": 0.0423,
+   "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/8cd93b9c47222e922271/record.json b/records/sd1/sweep/8cd93b9c47222e922271/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e4737e4f2c0777635b5de15b97f78ece5e300de
--- /dev/null
+++ b/records/sd1/sweep/8cd93b9c47222e922271/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8cd93b9c47222e922271",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:23Z",
+ "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.7.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.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/ab2d1f0e648d3f91db14484a96de9b9db719c6906e96120fa822115c3eed34bb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ab2d1f0e648d3f91db14484a96de9b9db719c6906e96120fa822115c3eed34bb",
+  "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": "5006bf4c1f452be731d2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cd93b9c47222e922271/workflow.json b/records/sd1/sweep/8cd93b9c47222e922271/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc81cf82ad3410699fecea4680d09176821dc7ce
--- /dev/null
+++ b/records/sd1/sweep/8cd93b9c47222e922271/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 8cd93b9c47222e922271)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.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/8ce05dfaa17ce968baa8/measurements.json b/records/sd1/sweep/8ce05dfaa17ce968baa8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..81fea492c97b7d6c16b3fbcafe359e2148d76b15
--- /dev/null
+++ b/records/sd1/sweep/8ce05dfaa17ce968baa8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ce05dfaa17ce968baa8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.546023,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.631533,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.896101,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.448375,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.215,
+   "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"
+  },
+  "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": 48.918,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8687,
+   "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": 0.6374,
+   "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/8ce05dfaa17ce968baa8/record.json b/records/sd1/sweep/8ce05dfaa17ce968baa8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6636aa1d3eb52490e3af1d2055af16f48a05443c
--- /dev/null
+++ b/records/sd1/sweep/8ce05dfaa17ce968baa8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8ce05dfaa17ce968baa8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:17Z",
+ "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.3.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "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/afbc379cf7daccaa3306aefc9d825e0e6813d12d97d1dda77ddbc0c0ac3ec86b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "afbc379cf7daccaa3306aefc9d825e0e6813d12d97d1dda77ddbc0c0ac3ec86b",
+  "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"
+    ],
+    "previous_id": "9ae8efacda7af314e4fc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ce05dfaa17ce968baa8/workflow.json b/records/sd1/sweep/8ce05dfaa17ce968baa8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7390d8d720c6a414e7718568fd0413aee176edf5
--- /dev/null
+++ b/records/sd1/sweep/8ce05dfaa17ce968baa8/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 8ce05dfaa17ce968baa8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.transformer_blocks.0.attn1.to_out.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/8ce20642ab502e0849fb/measurements.json b/records/sd1/sweep/8ce20642ab502e0849fb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1765252ec11c572366a472212512e0897a60739
--- /dev/null
+++ b/records/sd1/sweep/8ce20642ab502e0849fb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ce20642ab502e0849fb",
+ "values": {
+  "std": {
+   "value": 48.241,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01316,
+   "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": 5.759,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.022063,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.032181,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004349,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004557,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9961,
+   "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": 0.2252,
+   "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/8ce20642ab502e0849fb/record.json b/records/sd1/sweep/8ce20642ab502e0849fb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..35bf4e33e7617a8c1b23f86fed932e1f58c5fd7f
--- /dev/null
+++ b/records/sd1/sweep/8ce20642ab502e0849fb/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8ce20642ab502e0849fb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:29:51Z",
+ "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": 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_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": "2976534844fe80be26b4",
+  "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": "d82c5d41ece9464e243f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ce20642ab502e0849fb/workflow.json b/records/sd1/sweep/8ce20642ab502e0849fb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d95868ccddbee2f3aa5bf91ceb9ae860136d8969
--- /dev/null
+++ b/records/sd1/sweep/8ce20642ab502e0849fb/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 8ce20642ab502e0849fb)"
+  },
+  "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": "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/8ce5e3774137828813a8/interpretations.jsonl b/records/sd1/sweep/8ce5e3774137828813a8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e0c6a9bd0547bd6de4b06ee076a38b3ff8979bcb
--- /dev/null
+++ b/records/sd1/sweep/8ce5e3774137828813a8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ce5e3774137828813a8"}, "kind": "caption", "value": "Smeared painterly face with streaky black and white hair on a teal-grey background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "67421e18ebbcabd1"}
+{"target": {"record": "8ce5e3774137828813a8"}, "kind": "change", "value": "The face melts into rough painterly smears with a blank, distorted face and streaky two-tone hair, though the framed backdrop stays.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bb8d7d16f176c328"}
+{"target": {"record": "8ce5e3774137828813a8"}, "kind": "keywords", "value": ["smeared face", "painterly distortion", "streaky hair", "melted portrait", "teal background"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "30b52b7e5f5ce4ba"}
+{"target": {"record": "8ce5e3774137828813a8"}, "kind": "effect_tags", "value": ["painterly", "distorted", "blur", "surreal", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bd83f7cc1789595f"}
diff --git a/records/sd1/sweep/8ce5e3774137828813a8/measurements.json b/records/sd1/sweep/8ce5e3774137828813a8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d253b47cd962121ceb17207e4f3d60ef3347157e
--- /dev/null
+++ b/records/sd1/sweep/8ce5e3774137828813a8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8ce5e3774137828813a8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.395544,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.498664,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.376286,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.233241,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.564,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00794,
+   "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": 46.955,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0651,
+   "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": 0.519,
+   "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/8ce5e3774137828813a8/record.json b/records/sd1/sweep/8ce5e3774137828813a8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f795712141df6971f8238ecd2dfb0315b08bf5e
--- /dev/null
+++ b/records/sd1/sweep/8ce5e3774137828813a8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ce5e3774137828813a8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:20Z",
+ "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": "input_blocks.5.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.mid",
+   "kind": "skip",
+   "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/3f03dd8d57becc884d5b9cc04cf2daf10e5f380dee97df62656ce470d1f3826a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3f03dd8d57becc884d5b9cc04cf2daf10e5f380dee97df62656ce470d1f3826a",
+  "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": "a6a136d1fb5b912f9526"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ce5e3774137828813a8/workflow.json b/records/sd1/sweep/8ce5e3774137828813a8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee73c35b940282478762470ada2639ce72140100
--- /dev/null
+++ b/records/sd1/sweep/8ce5e3774137828813a8/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 8ce5e3774137828813a8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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": [
+    "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/8ce60bf1a0c7b08d0feb/interpretations.jsonl b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..fef916c9c37daf300ae79aaf9b3853061f3a0208
--- /dev/null
+++ b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ce60bf1a0c7b08d0feb"}, "kind": "caption", "value": "Dense bright multicoloured confetti noise filling the frame, no recognisable figure", "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": "6f42f88d040161b5"}
+{"target": {"record": "8ce60bf1a0c7b08d0feb"}, "kind": "change", "value": "The painted portrait is completely replaced by evenly spread rainbow static.", "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": "359fbc460fffbec9"}
+{"target": {"record": "8ce60bf1a0c7b08d0feb"}, "kind": "keywords", "value": ["rainbow static", "confetti noise", "broken output", "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": "d124e76954af2de3"}
+{"target": {"record": "8ce60bf1a0c7b08d0feb"}, "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": "e8aad59bbf92d3e1"}
diff --git a/records/sd1/sweep/8ce60bf1a0c7b08d0feb/measurements.json b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ba63ea217436560490e2a75c163938d55307f06
--- /dev/null
+++ b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ce60bf1a0c7b08d0feb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.972009,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.773393,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.918063,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.45723,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 90.601,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02553,
+   "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": 103.0,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8942,
+   "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": 0.9867,
+   "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/8ce60bf1a0c7b08d0feb/record.json b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f3511979756acf55250f02c9fae44577d1be863b
--- /dev/null
+++ b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ce60bf1a0c7b08d0feb",
+ "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": 90
+   },
+   "group": "out.conv",
+   "kind": "out_conv",
+   "module_type": "Conv2d",
+   "container": "Sequential",
+   "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/eb77822bf4195ee6d60ced8927ceb7fb3c61be422d568ac3b004908497e71842.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "eb77822bf4195ee6d60ced8927ceb7fb3c61be422d568ac3b004908497e71842",
+  "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": "214a816e0f421e260b0d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ce60bf1a0c7b08d0feb/workflow.json b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef980fe07932332720b060203123e52f0fad112f
--- /dev/null
+++ b/records/sd1/sweep/8ce60bf1a0c7b08d0feb/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 8ce60bf1a0c7b08d0feb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"out.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/8ce96e78791f5d8129b0/measurements.json b/records/sd1/sweep/8ce96e78791f5d8129b0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1fcdf8c816d84ab79eab0ae8b0b58da4d28330b
--- /dev/null
+++ b/records/sd1/sweep/8ce96e78791f5d8129b0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ce96e78791f5d8129b0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.374238,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.448904,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.26219,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.21245,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 87.148,
+   "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": 45.484,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8869,
+   "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": 0.1887,
+   "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/8ce96e78791f5d8129b0/record.json b/records/sd1/sweep/8ce96e78791f5d8129b0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ebaba492d197cd7795e6d9af4f59a830c6d217f
--- /dev/null
+++ b/records/sd1/sweep/8ce96e78791f5d8129b0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8ce96e78791f5d8129b0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:50Z",
+ "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.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "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/e02047adfd17d9c91e81e32fd4eea2811c0e1f42e20f03974cc601cf8dd82c56.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e02047adfd17d9c91e81e32fd4eea2811c0e1f42e20f03974cc601cf8dd82c56",
+  "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"
+    ],
+    "previous_id": "8fffa1d7cded02ae77ea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ce96e78791f5d8129b0/workflow.json b/records/sd1/sweep/8ce96e78791f5d8129b0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8930fde24f6634e15da152624b03f1244f458926
--- /dev/null
+++ b/records/sd1/sweep/8ce96e78791f5d8129b0/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 8ce96e78791f5d8129b0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.attn2.to_k\",\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/8cf06763019b38492baa/measurements.json b/records/sd1/sweep/8cf06763019b38492baa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..784819ad960c3d4b9947f70df9f31bad2558c37f
--- /dev/null
+++ b/records/sd1/sweep/8cf06763019b38492baa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8cf06763019b38492baa",
+ "values": {
+  "std": {
+   "value": 40.188,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01229,
+   "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": 11.243,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.043612,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.16436,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.050841,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.022281,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9584,
+   "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": 0.0263,
+   "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/8cf06763019b38492baa/record.json b/records/sd1/sweep/8cf06763019b38492baa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..373e8c298b265cceeca73586bdb1b50550b2e098
--- /dev/null
+++ b/records/sd1/sweep/8cf06763019b38492baa/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8cf06763019b38492baa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:01:05Z",
+ "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": 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": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "boost",
+   "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": [
+   "in.hi",
+   "multiply",
+   2.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 2.0,
+    "applied": 2.0,
+    "reason": "outside safe range [0.6, 1.6] 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": "f064b8d06be5f8e8f397"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cf06763019b38492baa/workflow.json b/records/sd1/sweep/8cf06763019b38492baa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4d9d3c1678146fca1212806f3dce7a5008b13e5
--- /dev/null
+++ b/records/sd1/sweep/8cf06763019b38492baa/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 8cf06763019b38492baa)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\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 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/8cf4b0418270c51ba92c/measurements.json b/records/sd1/sweep/8cf4b0418270c51ba92c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4729d1fd4ab4c96783ff26375a854ca1e670fd65
--- /dev/null
+++ b/records/sd1/sweep/8cf4b0418270c51ba92c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8cf4b0418270c51ba92c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.387853,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.387237,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.359013,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.245669,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.889,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00619,
+   "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": 45.663,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0318,
+   "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": 0.1777,
+   "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/8cf4b0418270c51ba92c/record.json b/records/sd1/sweep/8cf4b0418270c51ba92c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..74c0e5b44b7ca7fc326a524d6cc74f76df08f3db
--- /dev/null
+++ b/records/sd1/sweep/8cf4b0418270c51ba92c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8cf4b0418270c51ba92c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:36Z",
+ "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.proj_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "in.hi",
+   "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/115b0d1c657ac45cfe8900ef82a63a29e59732559eddf02cfbf6890b3f138afb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "115b0d1c657ac45cfe8900ef82a63a29e59732559eddf02cfbf6890b3f138afb",
+  "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": "ecc6adab771be042dbb1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cf4b0418270c51ba92c/workflow.json b/records/sd1/sweep/8cf4b0418270c51ba92c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b14ac5b9e7af31272323e038752592892b7be5d8
--- /dev/null
+++ b/records/sd1/sweep/8cf4b0418270c51ba92c/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 8cf4b0418270c51ba92c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.proj_out\",\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/8cfbfed40d94bee6f383/interpretations.jsonl b/records/sd1/sweep/8cfbfed40d94bee6f383/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..b1e3c2770b2e246ad2207539fc2aa3849a7ed7b5
--- /dev/null
+++ b/records/sd1/sweep/8cfbfed40d94bee6f383/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8cfbfed40d94bee6f383"}, "kind": "caption", "value": "Blurry mosaic-like face made of brownish and blue tiles emerging from a murky background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bbc8301a0274d7d3"}
+{"target": {"record": "8cfbfed40d94bee6f383"}, "kind": "change", "value": "The crisp illustration dissolves into a blurry, tiled mosaic of muddy browns and blues, and the face survives only as a vague ghost.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "01132f7d2415ba0e"}
+{"target": {"record": "8cfbfed40d94bee6f383"}, "kind": "keywords", "value": ["mosaic", "tiled blur", "ghost face", "murky", "dissolved portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b278dfb05d1a6f90"}
+{"target": {"record": "8cfbfed40d94bee6f383"}, "kind": "effect_tags", "value": ["blur", "fragmented", "texture", "abstract", "distorted", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8c042a2fded03d3e"}
diff --git a/records/sd1/sweep/8cfbfed40d94bee6f383/measurements.json b/records/sd1/sweep/8cfbfed40d94bee6f383/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b56c12bbac5ade6888ffb6e1fc61a97bba406817
--- /dev/null
+++ b/records/sd1/sweep/8cfbfed40d94bee6f383/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8cfbfed40d94bee6f383",
+ "values": {
+  "cosine_distance": {
+   "value": 0.555154,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.569177,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.572312,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.227006,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 31.36,
+   "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": [],
+   "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": 41.085,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0619,
+   "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": 0.3572,
+   "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/8cfbfed40d94bee6f383/record.json b/records/sd1/sweep/8cfbfed40d94bee6f383/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3d61c0f89ce620b934cf74791fb3ac4727350f0
--- /dev/null
+++ b/records/sd1/sweep/8cfbfed40d94bee6f383/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8cfbfed40d94bee6f383",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:40Z",
+ "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": 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": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.conv",
+   "kind": "out_conv",
+   "module_type": "Conv2d",
+   "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/9bac1463b83552dbf0aa405a61c00688881945992212bcf2cc747976bb23f680.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9bac1463b83552dbf0aa405a61c00688881945992212bcf2cc747976bb23f680",
+  "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": "4dc8318b7b8a4cd3ab21"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8cfbfed40d94bee6f383/workflow.json b/records/sd1/sweep/8cfbfed40d94bee6f383/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d1248d00ea5cc4f3cb7c546e672fad8ecad67b7a
--- /dev/null
+++ b/records/sd1/sweep/8cfbfed40d94bee6f383/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 8cfbfed40d94bee6f383)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"out.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/8d0397206cd464ac90c2/measurements.json b/records/sd1/sweep/8d0397206cd464ac90c2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d93c9938e4ba2e3da96203cacc220ccedd7b7abe
--- /dev/null
+++ b/records/sd1/sweep/8d0397206cd464ac90c2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d0397206cd464ac90c2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.205457,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.279685,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.122345,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.039297,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.185,
+   "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": 27.869,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9794,
+   "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": 0.1285,
+   "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/8d0397206cd464ac90c2/record.json b/records/sd1/sweep/8d0397206cd464ac90c2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..039760386808e572b3534eda4d527beb922b1cf8
--- /dev/null
+++ b/records/sd1/sweep/8d0397206cd464ac90c2/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d0397206cd464ac90c2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:48Z",
+ "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.10.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.lo",
+   "kind": "skip",
+   "module_type": "Identity",
+   "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/ff1773c251b751af4bd5a0d4b99cf9b339d0a5caab4b28db9569dff954bc7e15.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ff1773c251b751af4bd5a0d4b99cf9b339d0a5caab4b28db9569dff954bc7e15",
+  "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": "801e946206919f7b8335"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d0397206cd464ac90c2/workflow.json b/records/sd1/sweep/8d0397206cd464ac90c2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce49064c1d7acbba350b8390c75682f915be3853
--- /dev/null
+++ b/records/sd1/sweep/8d0397206cd464ac90c2/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 8d0397206cd464ac90c2)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.skip_connection\",\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/8d03a0feedad6cf323ab/interpretations.jsonl b/records/sd1/sweep/8d03a0feedad6cf323ab/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d0151eaa006d12283d06a5d9153812b3182d6c22
--- /dev/null
+++ b/records/sd1/sweep/8d03a0feedad6cf323ab/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d03a0feedad6cf323ab"}, "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": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d327877c13025fa8"}
+{"target": {"record": "8d03a0feedad6cf323ab"}, "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": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ff4ada58ccfbdb78"}
+{"target": {"record": "8d03a0feedad6cf323ab"}, "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": "2026-10-05T21:23:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "165ba27c75511b05"}
+{"target": {"record": "8d03a0feedad6cf323ab"}, "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": "c4b7e60570b29160"}
diff --git a/records/sd1/sweep/8d03a0feedad6cf323ab/measurements.json b/records/sd1/sweep/8d03a0feedad6cf323ab/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..65572d0f8bfc415f2fc2d424b2194eaa31ae161c
--- /dev/null
+++ b/records/sd1/sweep/8d03a0feedad6cf323ab/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d03a0feedad6cf323ab",
+ "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/8d03a0feedad6cf323ab/record.json b/records/sd1/sweep/8d03a0feedad6cf323ab/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7641258bc30f1da5722ba1fb2f2bf5504f5c56fe
--- /dev/null
+++ b/records/sd1/sweep/8d03a0feedad6cf323ab/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d03a0feedad6cf323ab",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:06Z",
+ "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": "middle_block.2.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "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/08c0bed43557756c2d19c8ce94a148517760cea91f794079154a7e7d08944c47.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "08c0bed43557756c2d19c8ce94a148517760cea91f794079154a7e7d08944c47",
+  "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": "0d00f14f221c601ce882"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d03a0feedad6cf323ab/workflow.json b/records/sd1/sweep/8d03a0feedad6cf323ab/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3c3c5cf851bca850de06413d5251bba44f5ee2a
--- /dev/null
+++ b/records/sd1/sweep/8d03a0feedad6cf323ab/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 8d03a0feedad6cf323ab)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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/8d12aa7d7baef3371b5a/measurements.json b/records/sd1/sweep/8d12aa7d7baef3371b5a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..30302dbe6443c0dd9aa5f7ba756f2e88902adc42
--- /dev/null
+++ b/records/sd1/sweep/8d12aa7d7baef3371b5a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d12aa7d7baef3371b5a",
+ "values": {
+  "std": {
+   "value": 37.532,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01549,
+   "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": 2.793,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.00189,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.011217,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.002573,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001314,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9989,
+   "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": 0.0143,
+   "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/8d12aa7d7baef3371b5a/record.json b/records/sd1/sweep/8d12aa7d7baef3371b5a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e8ff83e3bfc5742cbae5ecadf6eb5095ec1ceb7
--- /dev/null
+++ b/records/sd1/sweep/8d12aa7d7baef3371b5a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8d12aa7d7baef3371b5a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:16:43Z",
+ "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": 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": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "mid",
+   "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.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": "11f87c01ed1a1c9b5451"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d12aa7d7baef3371b5a/workflow.json b/records/sd1/sweep/8d12aa7d7baef3371b5a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b980b247d088d56b40daa220b0debc6c7ac9c7c
--- /dev/null
+++ b/records/sd1/sweep/8d12aa7d7baef3371b5a/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 8d12aa7d7baef3371b5a)"
+  },
+  "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 \"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/8d12da897bf8ebcfda2e/measurements.json b/records/sd1/sweep/8d12da897bf8ebcfda2e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dce39926ffc176cc9c500df529c56535ebc62f71
--- /dev/null
+++ b/records/sd1/sweep/8d12da897bf8ebcfda2e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d12da897bf8ebcfda2e",
+ "values": {
+  "std": {
+   "value": 32.306,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01934,
+   "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": 6.809,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013137,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.051647,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.035356,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007377,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.994,
+   "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": 0.1961,
+   "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/8d12da897bf8ebcfda2e/record.json b/records/sd1/sweep/8d12da897bf8ebcfda2e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..71a02f98e6fcd9c50183e959e3ef8f0ac8a96784
--- /dev/null
+++ b/records/sd1/sweep/8d12da897bf8ebcfda2e/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8d12da897bf8ebcfda2e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:18:16Z",
+ "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 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": 0.667,
+   "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": [
+   "in.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": "8ad65cad26090d26d202"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d12da897bf8ebcfda2e/workflow.json b/records/sd1/sweep/8d12da897bf8ebcfda2e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f33b0120030092e7185ab22a7d144809dfbcc5b
--- /dev/null
+++ b/records/sd1/sweep/8d12da897bf8ebcfda2e/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 8d12da897bf8ebcfda2e)"
+  },
+  "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\": 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/8d242640959b52d4a8a5/interpretations.jsonl b/records/sd1/sweep/8d242640959b52d4a8a5/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..b74e6e499404a5208db82c6523325ee3ed022445
--- /dev/null
+++ b/records/sd1/sweep/8d242640959b52d4a8a5/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d242640959b52d4a8a5"}, "kind": "caption", "value": "Framed portrait where the face is a blank brown smear under swept dark hair", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a075e86f0be0def6"}
+{"target": {"record": "8d242640959b52d4a8a5"}, "kind": "change", "value": "The face is erased into a featureless smear while the hair, dark shirt and grey framed backdrop remain.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2661b4f2789af66d"}
+{"target": {"record": "8d242640959b52d4a8a5"}, "kind": "keywords", "value": ["faceless", "erased face", "smeared features", "framed portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "42a9713e4940b855"}
+{"target": {"record": "8d242640959b52d4a8a5"}, "kind": "effect_tags", "value": ["distorted", "blur", "surreal", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e1ecb555b93bb44e"}
diff --git a/records/sd1/sweep/8d242640959b52d4a8a5/measurements.json b/records/sd1/sweep/8d242640959b52d4a8a5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2accf884b2f2cf34a26741abc472164801fbe4d3
--- /dev/null
+++ b/records/sd1/sweep/8d242640959b52d4a8a5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d242640959b52d4a8a5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.169329,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.274429,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.371271,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.254795,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.893,
+   "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": [
+    "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": 23.261,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0057,
+   "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": 0.7549,
+   "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/8d242640959b52d4a8a5/record.json b/records/sd1/sweep/8d242640959b52d4a8a5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3069016d3112016b2acbb476e5dcaaf22763938d
--- /dev/null
+++ b/records/sd1/sweep/8d242640959b52d4a8a5/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d242640959b52d4a8a5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:06Z",
+ "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.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "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/9c1b4665d140d95f267b8ab5a434d4f0fd970bd595fa1fc5a31ec5ea6641faef.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9c1b4665d140d95f267b8ab5a434d4f0fd970bd595fa1fc5a31ec5ea6641faef",
+  "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": "d8f028f2701938af0567"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d242640959b52d4a8a5/workflow.json b/records/sd1/sweep/8d242640959b52d4a8a5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e10bf84761f5b23ce89255157debad7a84a9f387
--- /dev/null
+++ b/records/sd1/sweep/8d242640959b52d4a8a5/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 8d242640959b52d4a8a5)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_layers.0\",\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/8d25632f6caf7ad2bc00/measurements.json b/records/sd1/sweep/8d25632f6caf7ad2bc00/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7782935aa73c2cfdbebacadb27213a7e75168d82
--- /dev/null
+++ b/records/sd1/sweep/8d25632f6caf7ad2bc00/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d25632f6caf7ad2bc00",
+ "values": {
+  "cosine_distance": {
+   "value": 0.112272,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.127212,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.140001,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.058057,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.564,
+   "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": 13.82,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9844,
+   "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": 0.3817,
+   "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/8d25632f6caf7ad2bc00/record.json b/records/sd1/sweep/8d25632f6caf7ad2bc00/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b5ac38508eec514babfa45d180982dcad6faebc
--- /dev/null
+++ b/records/sd1/sweep/8d25632f6caf7ad2bc00/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d25632f6caf7ad2bc00",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:59Z",
+ "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": "output_blocks.7.0.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.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/cd0834dea79b9d6fa56863d23db9e980af28526a699dc2c04036db0a16695ed4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "cd0834dea79b9d6fa56863d23db9e980af28526a699dc2c04036db0a16695ed4",
+  "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": "0ba91344971836d99bcf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d25632f6caf7ad2bc00/workflow.json b/records/sd1/sweep/8d25632f6caf7ad2bc00/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f063bf8e20c25316bc27846a9c8f8c3a51875f0e
--- /dev/null
+++ b/records/sd1/sweep/8d25632f6caf7ad2bc00/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 8d25632f6caf7ad2bc00)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.out_layers.3\",\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/8d351775e0ebc61fe326/measurements.json b/records/sd1/sweep/8d351775e0ebc61fe326/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ba01200b00fe3d81fb0a121d31cf6e9e18f0290
--- /dev/null
+++ b/records/sd1/sweep/8d351775e0ebc61fe326/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d351775e0ebc61fe326",
+ "values": {
+  "cosine_distance": {
+   "value": 0.247316,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.33212,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.182482,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.06218,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.03,
+   "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": 29.886,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9737,
+   "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/8d351775e0ebc61fe326/record.json b/records/sd1/sweep/8d351775e0ebc61fe326/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..98ff62215bf5ff273b4d30dcd56409c6b57199a8
--- /dev/null
+++ b/records/sd1/sweep/8d351775e0ebc61fe326/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d351775e0ebc61fe326",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:15Z",
+ "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": "input_blocks.5.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.mid",
+   "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/7fdbcb3b93d73265421e1aabe9cddd44ee453b4b1460491742bdd40e17f2b083.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7fdbcb3b93d73265421e1aabe9cddd44ee453b4b1460491742bdd40e17f2b083",
+  "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": "e7a3524d6450845926bf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d351775e0ebc61fe326/workflow.json b/records/sd1/sweep/8d351775e0ebc61fe326/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..08bffbd9d5248730791e759980d11493acbd238b
--- /dev/null
+++ b/records/sd1/sweep/8d351775e0ebc61fe326/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 8d351775e0ebc61fe326)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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/8d3ec8916904d1162a03/interpretations.jsonl b/records/sd1/sweep/8d3ec8916904d1162a03/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5063acaa2d441731d60ee788d686121a1d34fcd3
--- /dev/null
+++ b/records/sd1/sweep/8d3ec8916904d1162a03/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d3ec8916904d1162a03"}, "kind": "caption", "value": "Smeared, painterly seascape with faint lighthouse under a muted sunset sky", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e875decbae28fc19"}
+{"target": {"record": "8d3ec8916904d1162a03"}, "kind": "change", "value": "The layout stays but everything is blurred into muddy, mottled brushstrokes; the colours turn dull and the lighthouse almost disappears.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8990b9de92c33a87"}
+{"target": {"record": "8d3ec8916904d1162a03"}, "kind": "keywords", "value": ["smeared seascape", "faded lighthouse", "muddy painterly", "mottled sunset", "washed out coast"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "560df918f98e69da"}
+{"target": {"record": "8d3ec8916904d1162a03"}, "kind": "effect_tags", "value": ["painterly", "blur", "washed_out", "texture", "subject_kept", "desaturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ef7431c022b4dff1"}
diff --git a/records/sd1/sweep/8d3ec8916904d1162a03/measurements.json b/records/sd1/sweep/8d3ec8916904d1162a03/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..365ee0882d5f7cc0e6fbade57addd34b642d44d0
--- /dev/null
+++ b/records/sd1/sweep/8d3ec8916904d1162a03/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d3ec8916904d1162a03",
+ "values": {
+  "std": {
+   "value": 39.293,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00317,
+   "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": 20.95,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.187383,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.552347,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.869041,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.281263,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9248,
+   "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": 0.1636,
+   "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/8d3ec8916904d1162a03/record.json b/records/sd1/sweep/8d3ec8916904d1162a03/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d3eed383ff438b960eb9dbfcb68f8250dd54a199
--- /dev/null
+++ b/records/sd1/sweep/8d3ec8916904d1162a03/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8d3ec8916904d1162a03",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:59:48Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.hi",
+   "kind": "attn_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.hi",
+   "scale",
+   0.7,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.7,
+    "applied": 0.7,
+    "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"
+    ],
+    "previous_id": "7972abb21bf2f4f49368"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d3ec8916904d1162a03/workflow.json b/records/sd1/sweep/8d3ec8916904d1162a03/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5de22fc923fd27ccdc4d0a3c9731b7ac28d0985
--- /dev/null
+++ b/records/sd1/sweep/8d3ec8916904d1162a03/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 8d3ec8916904d1162a03)"
+  },
+  "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\": 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/8d450366f7e6733c4ac6/measurements.json b/records/sd1/sweep/8d450366f7e6733c4ac6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..502c8053a037b9b202a16620ffc18d341b4d6f6d
--- /dev/null
+++ b/records/sd1/sweep/8d450366f7e6733c4ac6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d450366f7e6733c4ac6",
+ "values": {
+  "std": {
+   "value": 26.214,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02629,
+   "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": 21.235,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.203003,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.386411,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131077,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.05617,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8957,
+   "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": 0.4974,
+   "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/8d450366f7e6733c4ac6/record.json b/records/sd1/sweep/8d450366f7e6733c4ac6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..15195bd331450c86ccd121b9d7b4e7f7e7c662fe
--- /dev/null
+++ b/records/sd1/sweep/8d450366f7e6733c4ac6/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8d450366f7e6733c4ac6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:00:43Z",
+ "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": 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": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "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",
+   "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": "cfbef607d1725ab9a7f6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d450366f7e6733c4ac6/workflow.json b/records/sd1/sweep/8d450366f7e6733c4ac6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cbc9c4d60ec101ac6e127292c10450b4264657c
--- /dev/null
+++ b/records/sd1/sweep/8d450366f7e6733c4ac6/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 8d450366f7e6733c4ac6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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": 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/8d4983188ca3cef04567/measurements.json b/records/sd1/sweep/8d4983188ca3cef04567/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..29ae019b709cf28c15e259d5c7c3bca814f1a01d
--- /dev/null
+++ b/records/sd1/sweep/8d4983188ca3cef04567/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d4983188ca3cef04567",
+ "values": {
+  "std": {
+   "value": 29.812,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01673,
+   "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": 38.094,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.433541,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.631235,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.734725,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.290705,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8797,
+   "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": 0.045,
+   "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/8d4983188ca3cef04567/record.json b/records/sd1/sweep/8d4983188ca3cef04567/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4b7c4f383b8e72c32813a5d3c1b47c17146bcf5
--- /dev/null
+++ b/records/sd1/sweep/8d4983188ca3cef04567/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8d4983188ca3cef04567",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:45:50Z",
+ "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": 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": 180.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "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",
+   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": "752fba4d836f3b83f4d4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d4983188ca3cef04567/workflow.json b/records/sd1/sweep/8d4983188ca3cef04567/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ca3f95c392932d117090c8b67860c5d265047a1
--- /dev/null
+++ b/records/sd1/sweep/8d4983188ca3cef04567/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 8d4983188ca3cef04567)"
+  },
+  "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 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/8d57d68adcbb3490b3de/measurements.json b/records/sd1/sweep/8d57d68adcbb3490b3de/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ff0ce1299cab65e1fc7eb25821696f5fa6a1171
--- /dev/null
+++ b/records/sd1/sweep/8d57d68adcbb3490b3de/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d57d68adcbb3490b3de",
+ "values": {
+  "cosine_distance": {
+   "value": 0.234009,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.455062,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.533917,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.34588,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.017,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00412,
+   "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": 29.666,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.988,
+   "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": 0.8598,
+   "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/8d57d68adcbb3490b3de/record.json b/records/sd1/sweep/8d57d68adcbb3490b3de/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d5c13ec92c2ee712fb17d0e6b5892f5b10c5919
--- /dev/null
+++ b/records/sd1/sweep/8d57d68adcbb3490b3de/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d57d68adcbb3490b3de",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:06Z",
+ "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.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.mid",
+   "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/1a88f1a57ae9ef9261bbdb994ba81a94d2a5834bc76cbeb910de4e9dd0a7b01a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1a88f1a57ae9ef9261bbdb994ba81a94d2a5834bc76cbeb910de4e9dd0a7b01a",
+  "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": "c1c6b848f44b8f5057b6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d57d68adcbb3490b3de/workflow.json b/records/sd1/sweep/8d57d68adcbb3490b3de/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..aefd2d932d765c5f157698aa370a73fbdc989a6a
--- /dev/null
+++ b/records/sd1/sweep/8d57d68adcbb3490b3de/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 8d57d68adcbb3490b3de)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_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": [
+    "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/8d5c7617513210b03224/measurements.json b/records/sd1/sweep/8d5c7617513210b03224/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6495d6d3d67edee3da46a258d6ecb80b7d99b6e
--- /dev/null
+++ b/records/sd1/sweep/8d5c7617513210b03224/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d5c7617513210b03224",
+ "values": {
+  "std": {
+   "value": 17.681,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03479,
+   "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": 33.877,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.315677,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.52409,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.696375,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.107149,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9554,
+   "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": 0.3215,
+   "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/8d5c7617513210b03224/record.json b/records/sd1/sweep/8d5c7617513210b03224/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a6190c069f325b090aa6007e73ca6065846ffddc
--- /dev/null
+++ b/records/sd1/sweep/8d5c7617513210b03224/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8d5c7617513210b03224",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:58:16Z",
+ "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": 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.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "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.lo",
+   "scale",
+   0.7,
+   "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": "9e0c85eff528e1d9b89b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d5c7617513210b03224/workflow.json b/records/sd1/sweep/8d5c7617513210b03224/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..61fc409f6fb8fdd67024adb5070ca68f6cddac05
--- /dev/null
+++ b/records/sd1/sweep/8d5c7617513210b03224/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 8d5c7617513210b03224)"
+  },
+  "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\": 0.7\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": 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/8d5ea030c011eee9cea7/interpretations.jsonl b/records/sd1/sweep/8d5ea030c011eee9cea7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..864f786e738c27a1979a12d29fa0aa498fd840e7
--- /dev/null
+++ b/records/sd1/sweep/8d5ea030c011eee9cea7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d5ea030c011eee9cea7"}, "kind": "caption", "value": "Abstract multicoloured smeared shapes vaguely suggesting a face, in teal, green, orange", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0ac066f6d1583fd1"}
+{"target": {"record": "8d5ea030c011eee9cea7"}, "kind": "change", "value": "The portrait breaks into a vivid abstract smear of teal, lime, orange and black strokes, with only a faint hint of a face remaining.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b363173c4aa3a3f2"}
+{"target": {"record": "8d5ea030c011eee9cea7"}, "kind": "keywords", "value": ["abstract face", "colour smear", "vivid strokes", "teal and orange", "dissolved portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "964b09ea582b2e3e"}
+{"target": {"record": "8d5ea030c011eee9cea7"}, "kind": "effect_tags", "value": ["abstract", "saturated", "colour_shift", "painterly", "fragmented", "subject_lost"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f96542316e8f59b4"}
diff --git a/records/sd1/sweep/8d5ea030c011eee9cea7/measurements.json b/records/sd1/sweep/8d5ea030c011eee9cea7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..06a21d2c3d1f4fcdcc640e4faf60351904266148
--- /dev/null
+++ b/records/sd1/sweep/8d5ea030c011eee9cea7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d5ea030c011eee9cea7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.504471,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.704014,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.671133,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.354446,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.103,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00322,
+   "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": 53.133,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.7923,
+   "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/8d5ea030c011eee9cea7/record.json b/records/sd1/sweep/8d5ea030c011eee9cea7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f9d8ae4b7cdb168c46288b5dd78e3a8bf78ed83
--- /dev/null
+++ b/records/sd1/sweep/8d5ea030c011eee9cea7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d5ea030c011eee9cea7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:14Z",
+ "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.8.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "skip",
+   "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/e2659930bd8f4646864c321521afbcf71efdbff172119d59f0b476064bb2ed35.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e2659930bd8f4646864c321521afbcf71efdbff172119d59f0b476064bb2ed35",
+  "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": "84eeb02b8de10d8a937f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d5ea030c011eee9cea7/workflow.json b/records/sd1/sweep/8d5ea030c011eee9cea7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..64499d8d93f187f5c39d2f8c1b1f167abfd84226
--- /dev/null
+++ b/records/sd1/sweep/8d5ea030c011eee9cea7/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 8d5ea030c011eee9cea7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.skip_connection\",\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/8d6a1fcd6bafc729b2f4/measurements.json b/records/sd1/sweep/8d6a1fcd6bafc729b2f4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d5a61d928c14510bebffc1aa477b42120e7e790
--- /dev/null
+++ b/records/sd1/sweep/8d6a1fcd6bafc729b2f4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d6a1fcd6bafc729b2f4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.096649,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.150627,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.11074,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.094759,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.011,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0035,
+   "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": 14.496,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0101,
+   "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": 0.4278,
+   "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/8d6a1fcd6bafc729b2f4/record.json b/records/sd1/sweep/8d6a1fcd6bafc729b2f4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4df18f7bcf3f43eef3878f3d5072613df5d5d80
--- /dev/null
+++ b/records/sd1/sweep/8d6a1fcd6bafc729b2f4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d6a1fcd6bafc729b2f4",
+ "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.out_layers.3",
+   "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/2e3476a221b9abe116eaad749ffdc5c35329487a8e1cb8c2cd32523b4cc251bf.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2e3476a221b9abe116eaad749ffdc5c35329487a8e1cb8c2cd32523b4cc251bf",
+  "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": "ca8d46eb50d02daf59b1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d6a1fcd6bafc729b2f4/workflow.json b/records/sd1/sweep/8d6a1fcd6bafc729b2f4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..eebdefe986acf83b0ef50e94a86dbf447195a41b
--- /dev/null
+++ b/records/sd1/sweep/8d6a1fcd6bafc729b2f4/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 8d6a1fcd6bafc729b2f4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.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": [
+    "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/8d6b125656dcc9267eef/measurements.json b/records/sd1/sweep/8d6b125656dcc9267eef/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8114224cbd97ca1f82812dc106e42ebfae291631
--- /dev/null
+++ b/records/sd1/sweep/8d6b125656dcc9267eef/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d6b125656dcc9267eef",
+ "values": {
+  "std": {
+   "value": 42.235,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01823,
+   "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": 34.581,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.395617,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.500423,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.269566,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.092727,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0174,
+   "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/8d6b125656dcc9267eef/record.json b/records/sd1/sweep/8d6b125656dcc9267eef/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..30c74cebfe8e0c6803455a72291f31d4a116493d
--- /dev/null
+++ b/records/sd1/sweep/8d6b125656dcc9267eef/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8d6b125656dcc9267eef",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:51:25Z",
+ "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": 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.8.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "in.lo",
+   "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": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "b371514439a11804ea77"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d6b125656dcc9267eef/workflow.json b/records/sd1/sweep/8d6b125656dcc9267eef/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5ffefb3150f7df1ebf6478e83c6bbbb49c611164
--- /dev/null
+++ b/records/sd1/sweep/8d6b125656dcc9267eef/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 8d6b125656dcc9267eef)"
+  },
+  "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\": 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 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/8d6bb316232e2b9571f8/interpretations.jsonl b/records/sd1/sweep/8d6bb316232e2b9571f8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7363b0b553d902e48ba6a8398d9c83caef9ed7ca
--- /dev/null
+++ b/records/sd1/sweep/8d6bb316232e2b9571f8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d6bb316232e2b9571f8"}, "kind": "caption", "value": "Stylised woman with slicked hair against a vivid teal crosshatched grid background, illustration look.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2262c90d6cd85d11"}
+{"target": {"record": "8d6bb316232e2b9571f8"}, "kind": "change", "value": "Nearly the same as G: the grey photo turns into a bright teal, grid-textured, illustration-style portrait with smoothed, graphic facial features.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2d8c6b6b9d0f74a6"}
+{"target": {"record": "8d6bb316232e2b9571f8"}, "kind": "keywords", "value": ["teal grid", "illustration style", "crosshatch texture", "vivid portrait", "poster look"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "33f8c0c46176b13a"}
+{"target": {"record": "8d6bb316232e2b9571f8"}, "kind": "effect_tags", "value": ["style_change", "saturated", "colour_shift", "poster", "pattern", "texture", "subject_kept", "bright"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:27:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "912cc175f191d7da"}
diff --git a/records/sd1/sweep/8d6bb316232e2b9571f8/measurements.json b/records/sd1/sweep/8d6bb316232e2b9571f8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..998120144349e4d4b234ee2256e6af4c780fb99d
--- /dev/null
+++ b/records/sd1/sweep/8d6bb316232e2b9571f8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d6bb316232e2b9571f8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.398148,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.588814,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.348807,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.185392,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.994,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03954,
+   "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": 58.734,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9043,
+   "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": 0.6575,
+   "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/8d6bb316232e2b9571f8/record.json b/records/sd1/sweep/8d6bb316232e2b9571f8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd820b5f7bc0d2e2aa70c7e7b8b257d0fc0056d0
--- /dev/null
+++ b/records/sd1/sweep/8d6bb316232e2b9571f8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d6bb316232e2b9571f8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:58Z",
+ "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.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.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/b239f1299d3e5d522f8f3782f4bdebf93a18c3fcde1a98ced0c7aef1e93dc73f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b239f1299d3e5d522f8f3782f4bdebf93a18c3fcde1a98ced0c7aef1e93dc73f",
+  "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": "99fa881899b22ca88906"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d6bb316232e2b9571f8/workflow.json b/records/sd1/sweep/8d6bb316232e2b9571f8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c44329b3b76063d6652bca56f97329316598767
--- /dev/null
+++ b/records/sd1/sweep/8d6bb316232e2b9571f8/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 8d6bb316232e2b9571f8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/8d6c54b57c44161e1369/measurements.json b/records/sd1/sweep/8d6c54b57c44161e1369/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4136ede4e2c363b52506869a6488d40558c5ce5
--- /dev/null
+++ b/records/sd1/sweep/8d6c54b57c44161e1369/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d6c54b57c44161e1369",
+ "values": {
+  "cosine_distance": {
+   "value": 0.358495,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.381516,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.32466,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.131354,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.465,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00674,
+   "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": 37.004,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0095,
+   "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": 0.1445,
+   "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/8d6c54b57c44161e1369/record.json b/records/sd1/sweep/8d6c54b57c44161e1369/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..754606cc9e349ff38bf4c388072d539f1016bcd5
--- /dev/null
+++ b/records/sd1/sweep/8d6c54b57c44161e1369/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d6c54b57c44161e1369",
+ "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": 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/7fb1fd620f099479db4dfd1598e0f027bafef6f0e156fdbec53343bbb795e2db.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7fb1fd620f099479db4dfd1598e0f027bafef6f0e156fdbec53343bbb795e2db",
+  "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": "89591a724d1d3c3c205d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d6c54b57c44161e1369/workflow.json b/records/sd1/sweep/8d6c54b57c44161e1369/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b531c2b03bad9da068643e267180254406f52d88
--- /dev/null
+++ b/records/sd1/sweep/8d6c54b57c44161e1369/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 8d6c54b57c44161e1369)"
+  },
+  "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\": 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/8d75d7b331b2eea88090/measurements.json b/records/sd1/sweep/8d75d7b331b2eea88090/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2bb68d0244d085be57541bb6836c404b9bb1701
--- /dev/null
+++ b/records/sd1/sweep/8d75d7b331b2eea88090/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d75d7b331b2eea88090",
+ "values": {
+  "cosine_distance": {
+   "value": 0.008363,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.015322,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004812,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002136,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.085,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00581,
+   "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": 4.137,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0179,
+   "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": 0.1498,
+   "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/8d75d7b331b2eea88090/record.json b/records/sd1/sweep/8d75d7b331b2eea88090/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ba29570fc7df81acbc5e2e09f7a37da6e97d12a
--- /dev/null
+++ b/records/sd1/sweep/8d75d7b331b2eea88090/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d75d7b331b2eea88090",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:51Z",
+ "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.9.1.transformer_blocks.0.attn2.to_v",
+   "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/18c603366614ad64d14a43801030bee38d69464ba9dfe121cdb181dc030e734e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "18c603366614ad64d14a43801030bee38d69464ba9dfe121cdb181dc030e734e",
+  "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": "3e3cf646be5911f8711c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d75d7b331b2eea88090/workflow.json b/records/sd1/sweep/8d75d7b331b2eea88090/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3eb649a3437dd1d77e40ce78d9e4b006e40a0bcb
--- /dev/null
+++ b/records/sd1/sweep/8d75d7b331b2eea88090/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 8d75d7b331b2eea88090)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.transformer_blocks.0.attn2.to_v\",\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/8d7f50b26c98a502322e/interpretations.jsonl b/records/sd1/sweep/8d7f50b26c98a502322e/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..bd74225341b279cefb9a819b63034ad6fd2cdec2
--- /dev/null
+++ b/records/sd1/sweep/8d7f50b26c98a502322e/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d7f50b26c98a502322e"}, "kind": "caption", "value": "Portrait of woman with lighter highlighted wavy hair and dark top in worn frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d78d91fbfe3fb541"}
+{"target": {"record": "8d7f50b26c98a502322e"}, "kind": "change", "value": "The composition stays while the hair becomes lighter and more golden, the face softer, and the frame border slightly larger.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bba2649dece23652"}
+{"target": {"record": "8d7f50b26c98a502322e"}, "kind": "keywords", "value": ["lighter hair", "golden highlights", "minor variation", "framed portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "44397eec4d65ba88"}
+{"target": {"record": "8d7f50b26c98a502322e"}, "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": "2026-10-05T21:22:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e0f71b500ac1ada0"}
diff --git a/records/sd1/sweep/8d7f50b26c98a502322e/measurements.json b/records/sd1/sweep/8d7f50b26c98a502322e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d3768f2f7fceb4b1e6f179fd8063d59257cf83f
--- /dev/null
+++ b/records/sd1/sweep/8d7f50b26c98a502322e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d7f50b26c98a502322e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.088094,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.120453,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.070585,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.029558,
+   "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": 0.00487,
+   "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": 14.751,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0662,
+   "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": 0.1483,
+   "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/8d7f50b26c98a502322e/record.json b/records/sd1/sweep/8d7f50b26c98a502322e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..36e72049894f5d5e9128f9a7abd71315a07e81b9
--- /dev/null
+++ b/records/sd1/sweep/8d7f50b26c98a502322e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d7f50b26c98a502322e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:14Z",
+ "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": "middle_block.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "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/0cc9359cc042e3a715a0e7756a42cca08db2e2c7852ed2842c97ad2a0f8a4a5d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0cc9359cc042e3a715a0e7756a42cca08db2e2c7852ed2842c97ad2a0f8a4a5d",
+  "type_path": "UNetModel.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": "be1c152cc8488c1ea184"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d7f50b26c98a502322e/workflow.json b/records/sd1/sweep/8d7f50b26c98a502322e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..94a5ffb0028b34dd5227bb123906f8d3efde217c
--- /dev/null
+++ b/records/sd1/sweep/8d7f50b26c98a502322e/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 8d7f50b26c98a502322e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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/8d82b141c7d63d891d52/measurements.json b/records/sd1/sweep/8d82b141c7d63d891d52/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e333690ecf263c403f553503f7e8276c98dd7c1
--- /dev/null
+++ b/records/sd1/sweep/8d82b141c7d63d891d52/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d82b141c7d63d891d52",
+ "values": {
+  "cosine_distance": {
+   "value": 0.147596,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.218504,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.047832,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.028339,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.703,
+   "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": 20.126,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9721,
+   "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": 0.1072,
+   "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/8d82b141c7d63d891d52/record.json b/records/sd1/sweep/8d82b141c7d63d891d52/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1182169bff9bf05ac410c0e3d8e4a622877b6bd
--- /dev/null
+++ b/records/sd1/sweep/8d82b141c7d63d891d52/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d82b141c7d63d891d52",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:57Z",
+ "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.1.1.transformer_blocks.0.attn2.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "cross_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/ff9c09edf7686232cc8287a3ee9b4f8b186ac5243f9e6421fed7e41d314db35c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ff9c09edf7686232cc8287a3ee9b4f8b186ac5243f9e6421fed7e41d314db35c",
+  "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": "9132424540e2bd359cad"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d82b141c7d63d891d52/workflow.json b/records/sd1/sweep/8d82b141c7d63d891d52/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c728a55eac0d50ee3327055854b2bfc225bc661a
--- /dev/null
+++ b/records/sd1/sweep/8d82b141c7d63d891d52/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 8d82b141c7d63d891d52)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.transformer_blocks.0.attn2.to_out.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/8d82ff1deaaf8178f277/measurements.json b/records/sd1/sweep/8d82ff1deaaf8178f277/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4bd0497a4fc171443606a8a064d1863000161b9c
--- /dev/null
+++ b/records/sd1/sweep/8d82ff1deaaf8178f277/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d82ff1deaaf8178f277",
+ "values": {
+  "cosine_distance": {
+   "value": 0.034445,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.060127,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.017196,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010478,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.833,
+   "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": 10.191,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0243,
+   "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": 0.103,
+   "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/8d82ff1deaaf8178f277/record.json b/records/sd1/sweep/8d82ff1deaaf8178f277/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..58fe8a2b50fca12761559361cfcd7dc17f5c4b52
--- /dev/null
+++ b/records/sd1/sweep/8d82ff1deaaf8178f277/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d82ff1deaaf8178f277",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:04Z",
+ "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.5.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "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/bc231e4b915496c78213a7b81441a6d5b81e0b0fef8f301132f01dbdda1e57a9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "bc231e4b915496c78213a7b81441a6d5b81e0b0fef8f301132f01dbdda1e57a9",
+  "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": "d8ce5a5baa961aca7cc1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d82ff1deaaf8178f277/workflow.json b/records/sd1/sweep/8d82ff1deaaf8178f277/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..da17cc4c277afe772243218894e47dcb940c66ef
--- /dev/null
+++ b/records/sd1/sweep/8d82ff1deaaf8178f277/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 8d82ff1deaaf8178f277)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.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/8d841640837f3a31fdeb/measurements.json b/records/sd1/sweep/8d841640837f3a31fdeb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0204774712d58d63c603e500c83e67bdde7f7750
--- /dev/null
+++ b/records/sd1/sweep/8d841640837f3a31fdeb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d841640837f3a31fdeb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.189504,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.511146,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.306684,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.147528,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 40.316,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00286,
+   "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": 27.017,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0429,
+   "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": 0.0785,
+   "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/8d841640837f3a31fdeb/record.json b/records/sd1/sweep/8d841640837f3a31fdeb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..882ce4b69f2db13394cb385794e31e5f177c8729
--- /dev/null
+++ b/records/sd1/sweep/8d841640837f3a31fdeb/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d841640837f3a31fdeb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:47Z",
+ "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": "output_blocks.10.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.hi",
+   "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/d74d3e49f480b1d7a3e013de8419344702acb2d3c4a1646d99f911dd557c558e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d74d3e49f480b1d7a3e013de8419344702acb2d3c4a1646d99f911dd557c558e",
+  "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": "7008ef01a82c31359c3d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d841640837f3a31fdeb/workflow.json b/records/sd1/sweep/8d841640837f3a31fdeb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..37c2442204e942ef66638fe48a508212652ea4f6
--- /dev/null
+++ b/records/sd1/sweep/8d841640837f3a31fdeb/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 8d841640837f3a31fdeb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.in_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": [
+    "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/8d8a0a9a265a3cd93875/measurements.json b/records/sd1/sweep/8d8a0a9a265a3cd93875/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..604e6448f504ebbd8d102c0ab7503a7da6a0c645
--- /dev/null
+++ b/records/sd1/sweep/8d8a0a9a265a3cd93875/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d8a0a9a265a3cd93875",
+ "values": {
+  "cosine_distance": {
+   "value": 0.383779,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.459788,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.246901,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.166669,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 29.338,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00473,
+   "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": 42.729,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0303,
+   "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": 0.4093,
+   "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/8d8a0a9a265a3cd93875/record.json b/records/sd1/sweep/8d8a0a9a265a3cd93875/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a64e5df66bcdf02e69bfb4a164c142f837c07ee
--- /dev/null
+++ b/records/sd1/sweep/8d8a0a9a265a3cd93875/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d8a0a9a265a3cd93875",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:23Z",
+ "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": 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.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.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/816169692f5d05cb8bc9c955792be648c3ea776173707d61f3d096e9f1c2c460.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "816169692f5d05cb8bc9c955792be648c3ea776173707d61f3d096e9f1c2c460",
+  "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": "7ea56d9b7fd60f606ecd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d8a0a9a265a3cd93875/workflow.json b/records/sd1/sweep/8d8a0a9a265a3cd93875/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fc580310cb4d97121bc6dae3e4e8765e12c351c
--- /dev/null
+++ b/records/sd1/sweep/8d8a0a9a265a3cd93875/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 8d8a0a9a265a3cd93875)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.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/8d9a959e5c3e2f7864ac/measurements.json b/records/sd1/sweep/8d9a959e5c3e2f7864ac/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..52220642806cb1dcc9b3ac03a2d0d7c4e5dcd998
--- /dev/null
+++ b/records/sd1/sweep/8d9a959e5c3e2f7864ac/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d9a959e5c3e2f7864ac",
+ "values": {
+  "std": {
+   "value": 29.299,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0257,
+   "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": 5.213,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013001,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.043241,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.035955,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008837,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9788,
+   "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": 0.3001,
+   "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/8d9a959e5c3e2f7864ac/record.json b/records/sd1/sweep/8d9a959e5c3e2f7864ac/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..20b202e38fce1fde7802afe7c732fd839083fb27
--- /dev/null
+++ b/records/sd1/sweep/8d9a959e5c3e2f7864ac/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8d9a959e5c3e2f7864ac",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:59:05Z",
+ "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": 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.8.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "in.lo",
+   "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": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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/8d9a959e5c3e2f7864ac/workflow.json b/records/sd1/sweep/8d9a959e5c3e2f7864ac/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc444c07d59576d1fbd740923437258d009a6b91
--- /dev/null
+++ b/records/sd1/sweep/8d9a959e5c3e2f7864ac/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 8d9a959e5c3e2f7864ac)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\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 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/8d9d7b1684981fee730b/measurements.json b/records/sd1/sweep/8d9d7b1684981fee730b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..293d78adff38f3ce530d9879d06ffd38bac2f121
--- /dev/null
+++ b/records/sd1/sweep/8d9d7b1684981fee730b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d9d7b1684981fee730b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.360731,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.466466,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.294149,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.153584,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.438,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00619,
+   "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": 40.192,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9682,
+   "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": 0.0921,
+   "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/8d9d7b1684981fee730b/record.json b/records/sd1/sweep/8d9d7b1684981fee730b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4cee19ac35f732cc8c726f12bf93c90669b5d9f
--- /dev/null
+++ b/records/sd1/sweep/8d9d7b1684981fee730b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8d9d7b1684981fee730b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:01Z",
+ "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": "middle_block.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "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/f3a2f22ec9205cedf52ae2034f53beaa0030eb0ff28d86f7cd0d728ae333c514.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f3a2f22ec9205cedf52ae2034f53beaa0030eb0ff28d86f7cd0d728ae333c514",
+  "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"
+    ],
+    "previous_id": "ca63da15f565ea9426ed"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d9d7b1684981fee730b/workflow.json b/records/sd1/sweep/8d9d7b1684981fee730b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca1eec3733b060b5b53a5e52d532cdd2c52e198f
--- /dev/null
+++ b/records/sd1/sweep/8d9d7b1684981fee730b/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 8d9d7b1684981fee730b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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": [
+    "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/8d9d99e279a937cd5278/interpretations.jsonl b/records/sd1/sweep/8d9d99e279a937cd5278/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..055ee8087c8eadff0be4342b0bbd3f5ee451ce80
--- /dev/null
+++ b/records/sd1/sweep/8d9d99e279a937cd5278/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8d9d99e279a937cd5278"}, "kind": "caption", "value": "Blurred painted portrait of a wavy-haired woman with darker, muted tones", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5ecb37d4405ee1b6"}
+{"target": {"record": "8d9d99e279a937cd5278"}, "kind": "change", "value": "The portrait is heavily blurred and slightly darkened, keeping the composition but erasing fine detail.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e4351dd0b59448ef"}
+{"target": {"record": "8d9d99e279a937cd5278"}, "kind": "keywords", "value": ["blur", "soft focus", "dark tones", "painted portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2f7366dadc8fe68a"}
+{"target": {"record": "8d9d99e279a937cd5278"}, "kind": "effect_tags", "value": ["blur", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1f4629ad749658bf"}
diff --git a/records/sd1/sweep/8d9d99e279a937cd5278/measurements.json b/records/sd1/sweep/8d9d99e279a937cd5278/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6baaf1e379c89ff6e60e7142feed8eaa20ba9f29
--- /dev/null
+++ b/records/sd1/sweep/8d9d99e279a937cd5278/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8d9d99e279a937cd5278",
+ "values": {
+  "cosine_distance": {
+   "value": 0.187951,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.451443,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.220893,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.134866,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.568,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00031,
+   "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": 22.492,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0159,
+   "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": 0.7641,
+   "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/8d9d99e279a937cd5278/record.json b/records/sd1/sweep/8d9d99e279a937cd5278/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff128953a8981e6b86bd4bcad365055d0743973b
--- /dev/null
+++ b/records/sd1/sweep/8d9d99e279a937cd5278/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8d9d99e279a937cd5278",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:03Z",
+ "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": "output_blocks.11.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "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/ccadcdf20974d74e83b87de2af6be98535a0010fa5a350bc038b3b249f5d3de7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ccadcdf20974d74e83b87de2af6be98535a0010fa5a350bc038b3b249f5d3de7",
+  "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": "dd4b2c5a1a158552d085"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d9d99e279a937cd5278/workflow.json b/records/sd1/sweep/8d9d99e279a937cd5278/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..75d9f31460e0bef102225c27d7de536b3de32aa5
--- /dev/null
+++ b/records/sd1/sweep/8d9d99e279a937cd5278/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 8d9d99e279a937cd5278)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.out_layers.0\",\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/8d9ecd07ed35c1847056/measurements.json b/records/sd1/sweep/8d9ecd07ed35c1847056/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..677863910b07acd62e96eac1aef4408080dca051
--- /dev/null
+++ b/records/sd1/sweep/8d9ecd07ed35c1847056/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8d9ecd07ed35c1847056",
+ "values": {
+  "std": {
+   "value": 29.974,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02944,
+   "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": 2.053,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001444,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.005892,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.00544,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001151,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0009,
+   "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": 0.1708,
+   "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/8d9ecd07ed35c1847056/record.json b/records/sd1/sweep/8d9ecd07ed35c1847056/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..56821c66917d0adc6c9389a787581fec21ab93c5
--- /dev/null
+++ b/records/sd1/sweep/8d9ecd07ed35c1847056/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8d9ecd07ed35c1847056",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:05:25Z",
+ "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": 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.3.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "medium",
+   "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.lo",
+   "add_noise",
+   0.2,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for out.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"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8d9ecd07ed35c1847056/workflow.json b/records/sd1/sweep/8d9ecd07ed35c1847056/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..896e62e8fa96602adecd1bf04bcee5eda233b8fb
--- /dev/null
+++ b/records/sd1/sweep/8d9ecd07ed35c1847056/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 8d9ecd07ed35c1847056)"
+  },
+  "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\": 0.2,\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 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/8da1c4e4d4a521eea027/measurements.json b/records/sd1/sweep/8da1c4e4d4a521eea027/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..088b89e94eba929f86078d39fbec5500d56ed830
--- /dev/null
+++ b/records/sd1/sweep/8da1c4e4d4a521eea027/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8da1c4e4d4a521eea027",
+ "values": {
+  "cosine_distance": {
+   "value": 0.133529,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.183716,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.065496,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.072125,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.538,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00391,
+   "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": 18.254,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0291,
+   "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": 0.3878,
+   "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/8da1c4e4d4a521eea027/record.json b/records/sd1/sweep/8da1c4e4d4a521eea027/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..310777b89fd6e63281e2dd2ecbbec64393ec364b
--- /dev/null
+++ b/records/sd1/sweep/8da1c4e4d4a521eea027/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8da1c4e4d4a521eea027",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:34Z",
+ "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": "output_blocks.6.1.transformer_blocks.0.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "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/76bab7f2749cb59c856f0221bf5216c29b8e89f54057b32526b6c6f22f98ca7e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "76bab7f2749cb59c856f0221bf5216c29b8e89f54057b32526b6c6f22f98ca7e",
+  "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"
+    ],
+    "previous_id": "c989d28413706be7db1a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8da1c4e4d4a521eea027/workflow.json b/records/sd1/sweep/8da1c4e4d4a521eea027/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8401f0340c6e7a88088afc25b42f2f1a6ed3bd19
--- /dev/null
+++ b/records/sd1/sweep/8da1c4e4d4a521eea027/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 8da1c4e4d4a521eea027)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.transformer_blocks.0.attn2.to_v\",\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/8da2e0695595930dbd56/measurements.json b/records/sd1/sweep/8da2e0695595930dbd56/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..15a1ce165b3977c1976872719290db5587b62cc9
--- /dev/null
+++ b/records/sd1/sweep/8da2e0695595930dbd56/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8da2e0695595930dbd56",
+ "values": {
+  "cosine_distance": {
+   "value": 0.549901,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.645891,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.896355,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.333497,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.419,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00168,
+   "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": 52.579,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9149,
+   "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": 0.2218,
+   "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/8da2e0695595930dbd56/record.json b/records/sd1/sweep/8da2e0695595930dbd56/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..624fa2fcb2a7cf4d34523ec374b0fa4bab1c4c37
--- /dev/null
+++ b/records/sd1/sweep/8da2e0695595930dbd56/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8da2e0695595930dbd56",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:58Z",
+ "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": 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": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.hi",
+   "kind": "skip",
+   "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/6c592f35f762604758f1cf72358593ffc6b1b98a06bdde096063d5e570cf3cf5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6c592f35f762604758f1cf72358593ffc6b1b98a06bdde096063d5e570cf3cf5",
+  "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": "64fd21549fe7dae1dce3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8da2e0695595930dbd56/workflow.json b/records/sd1/sweep/8da2e0695595930dbd56/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5d078b4e68399aeed450869e84a0e4ce29d206f
--- /dev/null
+++ b/records/sd1/sweep/8da2e0695595930dbd56/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 8da2e0695595930dbd56)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.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": [
+    "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/8da427c3436ad8371892/measurements.json b/records/sd1/sweep/8da427c3436ad8371892/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..38684ef95e91bcd47f4a50f0f1044c7fb19de8fc
--- /dev/null
+++ b/records/sd1/sweep/8da427c3436ad8371892/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8da427c3436ad8371892",
+ "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/8da427c3436ad8371892/record.json b/records/sd1/sweep/8da427c3436ad8371892/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dc8fbc9c9f4946467b49511b273eb907f3d2e9b
--- /dev/null
+++ b/records/sd1/sweep/8da427c3436ad8371892/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8da427c3436ad8371892",
+ "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": 60
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "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/6239c33d64cb26aee9543b15cd363e65fd28a5a34a9371d49aa25b9d5b8de905.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6239c33d64cb26aee9543b15cd363e65fd28a5a34a9371d49aa25b9d5b8de905",
+  "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": "1dc6ca6c1c64450808f3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8da427c3436ad8371892/workflow.json b/records/sd1/sweep/8da427c3436ad8371892/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..90c4555f6da5509629d570b89dd343d97b03ed98
--- /dev/null
+++ b/records/sd1/sweep/8da427c3436ad8371892/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 8da427c3436ad8371892)"
+  },
+  "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\": 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/8db272a6c551e2ed32e7/measurements.json b/records/sd1/sweep/8db272a6c551e2ed32e7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e10bf1016b1c98c6413440bc788080162e4653b
--- /dev/null
+++ b/records/sd1/sweep/8db272a6c551e2ed32e7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8db272a6c551e2ed32e7",
+ "values": {
+  "std": {
+   "value": 52.031,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0081,
+   "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": 26.551,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.255805,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.568881,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.941072,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.434109,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6525,
+   "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": 0.0656,
+   "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/8db272a6c551e2ed32e7/record.json b/records/sd1/sweep/8db272a6c551e2ed32e7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..688aebb7eaf9d92d19acf917041b4fb1e1b76e57
--- /dev/null
+++ b/records/sd1/sweep/8db272a6c551e2ed32e7/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8db272a6c551e2ed32e7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:31:12Z",
+ "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": 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": 90.0
+   },
+   "group": "out.hi",
+   "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": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "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": "741527d7fb527dbe3a3b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8db272a6c551e2ed32e7/workflow.json b/records/sd1/sweep/8db272a6c551e2ed32e7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f8f84067d9547a1eda1096af0893c84164fcd78
--- /dev/null
+++ b/records/sd1/sweep/8db272a6c551e2ed32e7/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 8db272a6c551e2ed32e7)"
+  },
+  "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\": 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/8dcc07f6f3dc94fa5d5a/measurements.json b/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..000075715a724411b21585d2a40471818c412c82
--- /dev/null
+++ b/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8dcc07f6f3dc94fa5d5a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.486296,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.561739,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.346836,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.201629,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 38.949,
+   "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": [],
+   "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": 44.117,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0355,
+   "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": 0.3271,
+   "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/8dcc07f6f3dc94fa5d5a/record.json b/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1b739b0fc58efb7ef44b17c953c4fc11b467bf0
--- /dev/null
+++ b/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8dcc07f6f3dc94fa5d5a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:37Z",
+ "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.3.0.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/63b357c0c35ce073cd77813fd2a02b7712bda9f06c86adc9db246939a9116829.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "63b357c0c35ce073cd77813fd2a02b7712bda9f06c86adc9db246939a9116829",
+  "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": "65584b1d6263af52be0f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/workflow.json b/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d55fe5ec3244fcad782b8601421893ec3efdc43f
--- /dev/null
+++ b/records/sd1/sweep/8dcc07f6f3dc94fa5d5a/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 8dcc07f6f3dc94fa5d5a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.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/8dd7f5e09aad677753f9/interpretations.jsonl b/records/sd1/sweep/8dd7f5e09aad677753f9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..bb91322c86cc9e9dd30f276bdfb8570c3c723229
--- /dev/null
+++ b/records/sd1/sweep/8dd7f5e09aad677753f9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8dd7f5e09aad677753f9"}, "kind": "caption", "value": "Young person with short tousled hair looking right, dark turtleneck, pale washed backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6b88fabf6637c9f2"}
+{"target": {"record": "8dd7f5e09aad677753f9"}, "kind": "change", "value": "The subject turns slightly to the right with messier hair and a lighter, washed-out skin tone; the composition and muted style mostly stay the same.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "91c023e3a8925873"}
+{"target": {"record": "8dd7f5e09aad677753f9"}, "kind": "keywords", "value": ["portrait", "head turn", "pale tones", "short hair", "soft light"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0c158481ed8b99c4"}
+{"target": {"record": "8dd7f5e09aad677753f9"}, "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": "2026-10-05T08:18:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e4e694af42ddd285"}
diff --git a/records/sd1/sweep/8dd7f5e09aad677753f9/measurements.json b/records/sd1/sweep/8dd7f5e09aad677753f9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c85aa8f38f2a8a0949028cb5db8269e4f9f79047
--- /dev/null
+++ b/records/sd1/sweep/8dd7f5e09aad677753f9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8dd7f5e09aad677753f9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.336202,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.351274,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.195052,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.154702,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.271,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00462,
+   "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": 35.637,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9106,
+   "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": 0.7147,
+   "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/8dd7f5e09aad677753f9/record.json b/records/sd1/sweep/8dd7f5e09aad677753f9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ea842c7cc9a882e8f649bee0c127366c66f5ef4
--- /dev/null
+++ b/records/sd1/sweep/8dd7f5e09aad677753f9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8dd7f5e09aad677753f9",
+ "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.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "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/c4f725d67cc7451d30412fab1f64bdeaa58eb8b988653d8b27e30c18e3b8a6e8.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c4f725d67cc7451d30412fab1f64bdeaa58eb8b988653d8b27e30c18e3b8a6e8",
+  "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": "2f2b4f9731639e42b5a0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8dd7f5e09aad677753f9/workflow.json b/records/sd1/sweep/8dd7f5e09aad677753f9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..75749ca86159ed56c96dc35a1b4f80e9d5fb7a4d
--- /dev/null
+++ b/records/sd1/sweep/8dd7f5e09aad677753f9/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 8dd7f5e09aad677753f9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.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": [
+    "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/8ddba24cc7c5e9d558e0/interpretations.jsonl b/records/sd1/sweep/8ddba24cc7c5e9d558e0/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..3415803bafc506c5b56cee249532d8163e4f9e2b
--- /dev/null
+++ b/records/sd1/sweep/8ddba24cc7c5e9d558e0/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8ddba24cc7c5e9d558e0"}, "kind": "caption", "value": "Heavily blurred portrait of a woman with wavy shoulder-length hair and dark top", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:44:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ee757195fa260a0d"}
+{"target": {"record": "8ddba24cc7c5e9d558e0"}, "kind": "change", "value": "The crisp illustrated girl turns into a soft, out-of-focus photographic blur. Composition, wavy hair and dark top survive, but the facial detail and painted frame are lost.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:44:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "90ae992574126423"}
+{"target": {"record": "8ddba24cc7c5e9d558e0"}, "kind": "keywords", "value": ["blurred portrait", "out of focus", "soft focus", "wavy hair", "lost detail"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:44:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ec26c3f6f82c759c"}
+{"target": {"record": "8ddba24cc7c5e9d558e0"}, "kind": "effect_tags", "value": ["blur", "subject_kept", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:44:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a913a7776afbe848"}
diff --git a/records/sd1/sweep/8ddba24cc7c5e9d558e0/measurements.json b/records/sd1/sweep/8ddba24cc7c5e9d558e0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b6650de82f504fd1a9d402167e123cd9f0dde0a
--- /dev/null
+++ b/records/sd1/sweep/8ddba24cc7c5e9d558e0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8ddba24cc7c5e9d558e0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.266675,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.578541,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.342904,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.192657,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.212,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.004,
+   "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": 27.908,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0032,
+   "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": 0.8352,
+   "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/8ddba24cc7c5e9d558e0/record.json b/records/sd1/sweep/8ddba24cc7c5e9d558e0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd768af6830be231a2c9a6775e1ddcecb5638c0c
--- /dev/null
+++ b/records/sd1/sweep/8ddba24cc7c5e9d558e0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8ddba24cc7c5e9d558e0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:40Z",
+ "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": 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.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/cbe8a0ed5d4c477d7970e8c9eedf3a78b4c09adefc891370f2e6b0124c0d3772.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "cbe8a0ed5d4c477d7970e8c9eedf3a78b4c09adefc891370f2e6b0124c0d3772",
+  "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": "03665700ea7243eb47cc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8ddba24cc7c5e9d558e0/workflow.json b/records/sd1/sweep/8ddba24cc7c5e9d558e0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a982b94c7bc35c46f9b82795e2df54074938c4d
--- /dev/null
+++ b/records/sd1/sweep/8ddba24cc7c5e9d558e0/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 8ddba24cc7c5e9d558e0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.transformer_blocks.0.norm1\",\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/8de30ede7a8381c950b3/measurements.json b/records/sd1/sweep/8de30ede7a8381c950b3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b2f53a276685ffbf1f5a828528cac153e3fc3cc6
--- /dev/null
+++ b/records/sd1/sweep/8de30ede7a8381c950b3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8de30ede7a8381c950b3",
+ "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