diff --git a/records/sd1/sweep/5a56177963d9a0d23ae0/output.webp b/records/sd1/sweep/5a56177963d9a0d23ae0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0a7de98a6812984f17e62c6660a06b80f9d9a152
--- /dev/null
+++ b/records/sd1/sweep/5a56177963d9a0d23ae0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44d61a33adbfc5955f33d76230606d3f831d5290cbd93cf63760fcc30c25cd04
+size 22374
diff --git a/records/sd1/sweep/5a562c8d64551e48234e/output.webp b/records/sd1/sweep/5a562c8d64551e48234e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9cb9de5e3ba68c539847aa83dea2dd670ba5e094
--- /dev/null
+++ b/records/sd1/sweep/5a562c8d64551e48234e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a96dc393bba78626c4733e08f5dc4447de444e33b630b30252d5288a5b1b8712
+size 45044
diff --git a/records/sd1/sweep/5a76f8e003071e8a2c40/output.webp b/records/sd1/sweep/5a76f8e003071e8a2c40/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c6e68936f6dcb3b9a011b9d8da7bc520bcf7969e
--- /dev/null
+++ b/records/sd1/sweep/5a76f8e003071e8a2c40/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02ca276cfa1c8ff00b2568d81978f0a514fc026bbda03c965baa821726ee3911
+size 38540
diff --git a/records/sd1/sweep/5a884ad3f652b44fe203/output.webp b/records/sd1/sweep/5a884ad3f652b44fe203/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..da9b30fdd6067b68273731a7478878933cad5619
--- /dev/null
+++ b/records/sd1/sweep/5a884ad3f652b44fe203/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f07438dcf0becad02ac2b515712fdefba14b16f183dfe5832fedfed95a2783da
+size 25592
diff --git a/records/sd1/sweep/5a93d8d408cccb519193/output.webp b/records/sd1/sweep/5a93d8d408cccb519193/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0be274b2a057e8ff6a250953290e4648b325af27
--- /dev/null
+++ b/records/sd1/sweep/5a93d8d408cccb519193/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2d5f72c8a51ad2a246fd18f39ede9e17a4369f79871a16d1506942e57a361315
+size 58540
diff --git a/records/sd1/sweep/5a969db1afbcba3cf5e1/output.webp b/records/sd1/sweep/5a969db1afbcba3cf5e1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3cfd341312f6d5ee3eb08ff43bafa43bbdb9499e
--- /dev/null
+++ b/records/sd1/sweep/5a969db1afbcba3cf5e1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cceaed3ed6215eb5befd0918efed63390ad1305ba533d92f2e2cf76b4620f02
+size 27244
diff --git a/records/sd1/sweep/5aa40c4bceab56bdbe13/output.webp b/records/sd1/sweep/5aa40c4bceab56bdbe13/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..87f98e5fd4178fe6c5c77e18cac19fed310f0f6c
--- /dev/null
+++ b/records/sd1/sweep/5aa40c4bceab56bdbe13/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94460c4efd1d35e93de4e1bb9eaaae76c0a715c8fe104a00ff2d718005cea5a7
+size 35260
diff --git a/records/sd1/sweep/5aa78e4473a7c3b90d86/output.webp b/records/sd1/sweep/5aa78e4473a7c3b90d86/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dfc8765da578f3ab9221afc152bd9c90c71ebed4
--- /dev/null
+++ b/records/sd1/sweep/5aa78e4473a7c3b90d86/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:35d582863a1614225e3d34a7af937cb76708c3d6f5d043941bac0c4184dece34
+size 45336
diff --git a/records/sd1/sweep/5ab1d2eb10c89f182306/output.webp b/records/sd1/sweep/5ab1d2eb10c89f182306/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a39c36f2542ee50eb76c9288f45e395a0c3b94fd
--- /dev/null
+++ b/records/sd1/sweep/5ab1d2eb10c89f182306/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d29fece6619f14263b5c1e6176b4afd24e80c6579892bb196fe4aa66d37dbfc
+size 24776
diff --git a/records/sd1/sweep/5ab4790cddeda73b254b/output.webp b/records/sd1/sweep/5ab4790cddeda73b254b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..87ec63d5f09b102a818e8d94f1ee81a3ffac7284
--- /dev/null
+++ b/records/sd1/sweep/5ab4790cddeda73b254b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:39ec43fe4a8eef81bd25c4d2c5e64cb141469ce89a863ba2f691612fa1da3dc5
+size 29096
diff --git a/records/sd1/sweep/5ac4cea6386546bf4d75/output.webp b/records/sd1/sweep/5ac4cea6386546bf4d75/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9cdcdd621551e6e9d7af1566bf2f89da99a17d01
--- /dev/null
+++ b/records/sd1/sweep/5ac4cea6386546bf4d75/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db28d4bfe0af0c731d0ebf461a6539091ffe0b676da9e9436b13d4b237b4317c
+size 49760
diff --git a/records/sd1/sweep/5acfc91c20d8bcc09cff/output.webp b/records/sd1/sweep/5acfc91c20d8bcc09cff/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ce43f7c95a021b3d93eabaa2f032e652d81bd11c
--- /dev/null
+++ b/records/sd1/sweep/5acfc91c20d8bcc09cff/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa719cdb4a177afd34ee01c0c1d8acbf9d6d50bda2eb5e55a7359665adf69558
+size 36000
diff --git a/records/sd1/sweep/5ad89f6ce85ca4097b33/output.webp b/records/sd1/sweep/5ad89f6ce85ca4097b33/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..51131077d6108b3e2ea952b83d832fa527542e5e
--- /dev/null
+++ b/records/sd1/sweep/5ad89f6ce85ca4097b33/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03b251589db7ac47a9165b008d606181598fcaa76af5c69ee524ff5fac8d52da
+size 48568
diff --git a/records/sd1/sweep/5ae815e390ac07c92e64/output.webp b/records/sd1/sweep/5ae815e390ac07c92e64/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8cc645443d9c02c59adb25763e0fd34c415e818b
--- /dev/null
+++ b/records/sd1/sweep/5ae815e390ac07c92e64/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:985c67745c139b689b4de958c2cddf8ea8c254c1bb9c9c2cdd7d779439216bd6
+size 51980
diff --git a/records/sd1/sweep/5aeadc4cc0ed3f1956cf/output.webp b/records/sd1/sweep/5aeadc4cc0ed3f1956cf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..38bb4533be84dc192823c5b31d5952ea9f7ad367
--- /dev/null
+++ b/records/sd1/sweep/5aeadc4cc0ed3f1956cf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09e50abd807abd56a418cb7ec588aa5d49e32c996b0253e7a772414e56e39d60
+size 44490
diff --git a/records/sd1/sweep/5aed2922a55c0dca5291/output.webp b/records/sd1/sweep/5aed2922a55c0dca5291/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f56f14827b8fcde9e6dd854f42af03d45701a78b
--- /dev/null
+++ b/records/sd1/sweep/5aed2922a55c0dca5291/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f822ebabeb999b17bbd617380763b6c39f85aafa358af9b56c6f339bbd052227
+size 51400
diff --git a/records/sd1/sweep/5af39537f110e0a64edf/output.webp b/records/sd1/sweep/5af39537f110e0a64edf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e40f7a26abb4bb66f230ab5507e478ad14a3c09e
--- /dev/null
+++ b/records/sd1/sweep/5af39537f110e0a64edf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2fdf00baf31202f24404fdbe31acbf19ccb326cea6a9dce80d3b6511dc3dde0b
+size 37674
diff --git a/records/sd1/sweep/5af76b110b0d3ee0d56f/output.webp b/records/sd1/sweep/5af76b110b0d3ee0d56f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..557b24d04911850019c206cd8c5b3259e9d1a74b
--- /dev/null
+++ b/records/sd1/sweep/5af76b110b0d3ee0d56f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca81fc7985321f079e287afa91726f4f3f9da8304946f8a51199006dd1ee23d0
+size 28038
diff --git a/records/sd1/sweep/5af7c4641ec0e807ebfe/output.webp b/records/sd1/sweep/5af7c4641ec0e807ebfe/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b4e68065b36ee1c55c7377f09d73185ae6b39634
--- /dev/null
+++ b/records/sd1/sweep/5af7c4641ec0e807ebfe/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da3fe50f2a780da616189d5bc6fab01695e617658690dcaa0281d4364c23ec00
+size 24870
diff --git a/records/sd1/sweep/5afa285914439326a001/output.webp b/records/sd1/sweep/5afa285914439326a001/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f4e4781de14e641e1d0cba2d7347958123bc2ef0
--- /dev/null
+++ b/records/sd1/sweep/5afa285914439326a001/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed902acdd34be874d0ca3495d16ec1e46adee72de409aacf93a2453bafdb7816
+size 33538
diff --git a/records/sd1/sweep/5afce56cab85df2663b7/output.webp b/records/sd1/sweep/5afce56cab85df2663b7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e5192a0f4c58f44758761f78a4b5281d9175435f
--- /dev/null
+++ b/records/sd1/sweep/5afce56cab85df2663b7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:31f881763092c142c892475127220883dc5ea6bd6a90a23f0d2938bc13cdccec
+size 44854
diff --git a/records/sd1/sweep/5b02501ae2dde7b8db39/output.webp b/records/sd1/sweep/5b02501ae2dde7b8db39/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a3a7e010d9e271e0bd2934f387aee9000daf616b
--- /dev/null
+++ b/records/sd1/sweep/5b02501ae2dde7b8db39/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10f9d82aec0b5d6fe9e10b53ed3262f82cfd0f814edc4a858934499b2ba3f03d
+size 25746
diff --git a/records/sd1/sweep/5b04acf1be41d576555c/output.webp b/records/sd1/sweep/5b04acf1be41d576555c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..839c046f74da0e5af628f5f51f80bdb1f2c73a03
--- /dev/null
+++ b/records/sd1/sweep/5b04acf1be41d576555c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba0c3c1e5529564ceb5412d8470ed6ea5faa32aecad6f05221d98cc7e4caa1b0
+size 57978
diff --git a/records/sd1/sweep/5b07e795915e50e36aa9/output.webp b/records/sd1/sweep/5b07e795915e50e36aa9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c19508a248eeb329df2dfe8ddf0248d10b8eae80
--- /dev/null
+++ b/records/sd1/sweep/5b07e795915e50e36aa9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:93298fe75222257e97a4a54e6c775d1114fc34fb84c4c05c0adae594a25268a2
+size 45046
diff --git a/records/sd1/sweep/5b2b723e6dd101b21814/output.webp b/records/sd1/sweep/5b2b723e6dd101b21814/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3a46c9dc879ffafbbddc361cff674b80ac83678e
--- /dev/null
+++ b/records/sd1/sweep/5b2b723e6dd101b21814/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca9b81e4c5412e92a9f8a726471b92a440e8cf9554adb5ae2fb601479e60aa54
+size 62604
diff --git a/records/sd1/sweep/5b34fbdbeefd4cb30023/output.webp b/records/sd1/sweep/5b34fbdbeefd4cb30023/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f0d68718f86f1c088ff40e6b29c6ce8a8366343a
--- /dev/null
+++ b/records/sd1/sweep/5b34fbdbeefd4cb30023/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fa91cfbf32feb504821ff74bb34497c28938d5807bd52d6f3acd1935127cb50
+size 50538
diff --git a/records/sd1/sweep/5b3b65b334dffefbeae2/output.webp b/records/sd1/sweep/5b3b65b334dffefbeae2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7b5e31fa156314a746b67961b4aad542b14f395b
--- /dev/null
+++ b/records/sd1/sweep/5b3b65b334dffefbeae2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f098f4e9db71ed4f435f617ce052036fbcafc8ab0ebd9117078d091bedb1062b
+size 15426
diff --git a/records/sd1/sweep/5b4ee0342f8231d30fb2/output.webp b/records/sd1/sweep/5b4ee0342f8231d30fb2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3bc14f8012707bcb8c13f0d8bc553f5076195691
--- /dev/null
+++ b/records/sd1/sweep/5b4ee0342f8231d30fb2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4dd5543bcead7e1d9383e097190f7c3e6e1672e2c3cdefeec6e45e68de5676b2
+size 58260
diff --git a/records/sd1/sweep/5b4ef92923bdc53693c4/output.webp b/records/sd1/sweep/5b4ef92923bdc53693c4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..56fd6c211ff8b7a4995bcc5a7320bac83dfaaf7f
--- /dev/null
+++ b/records/sd1/sweep/5b4ef92923bdc53693c4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:367cec1d272ddc5cf48898eea381bd67e97c43d1bb3d276870cf13f83384ed15
+size 33096
diff --git a/records/sd1/sweep/5b58f430db02cfc68321/output.webp b/records/sd1/sweep/5b58f430db02cfc68321/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e2ce4b0d4663450a815e5e7fd5af5a4d34607fc6
--- /dev/null
+++ b/records/sd1/sweep/5b58f430db02cfc68321/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7857fecd87e39ff49028b805926d1cbc172c7dc1a95f03b436aba061269b3db
+size 15312
diff --git a/records/sd1/sweep/5b6bd86d1fb0b2152d9c/output.webp b/records/sd1/sweep/5b6bd86d1fb0b2152d9c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b8d93efdc97b1754b222f0c38474a5d7c3a13376
--- /dev/null
+++ b/records/sd1/sweep/5b6bd86d1fb0b2152d9c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e664e29a16007e91a394d9d79ae8d2e34299a8370d0d6ec241a49e820b6af57
+size 30370
diff --git a/records/sd1/sweep/5b6ed6dc039c41473889/output.webp b/records/sd1/sweep/5b6ed6dc039c41473889/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c6d49ec11e365406e114bb2463089080843e14ac
--- /dev/null
+++ b/records/sd1/sweep/5b6ed6dc039c41473889/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf249a8519ff74fac8ad234a4ab59cbbdca710d6aee86ba0ecedf30797f4a942
+size 31400
diff --git a/records/sd1/sweep/5b71eef766e63dce04f9/output.webp b/records/sd1/sweep/5b71eef766e63dce04f9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d8bd85995f95dccce6f550b02e53cf9d20d6704f
--- /dev/null
+++ b/records/sd1/sweep/5b71eef766e63dce04f9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a19e61b147c3898c1742e31cf6eac56b55f311a30a56c29118d1176492dadf07
+size 35208
diff --git a/records/sd1/sweep/5b77104bf3fdce5a17ae/output.webp b/records/sd1/sweep/5b77104bf3fdce5a17ae/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..86371be0a452dd9dccbf985e4fc7d1a2d38516ad
--- /dev/null
+++ b/records/sd1/sweep/5b77104bf3fdce5a17ae/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d01b1249455f6af4be9a8fbba9628d230ccfc25708cf79a13d0fd187167da086
+size 45030
diff --git a/records/sd1/sweep/5b7a48a1a58cf908f901/output.webp b/records/sd1/sweep/5b7a48a1a58cf908f901/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..37962b43b724ea01065342338a307c8e31279015
--- /dev/null
+++ b/records/sd1/sweep/5b7a48a1a58cf908f901/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6dfb525c5b149725a98181378fdaa4f7e5a53ed279beeb600c9e66c1ebb4b893
+size 31930
diff --git a/records/sd1/sweep/5b7a5f60442090d82691/output.webp b/records/sd1/sweep/5b7a5f60442090d82691/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..93e9612c04b3eba30ffa17da1c793bda7b97b79e
--- /dev/null
+++ b/records/sd1/sweep/5b7a5f60442090d82691/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06fa3118ae7e08208fab40c7de4de25c0f1f94e3fe50e6a90b4ca82533ab098e
+size 21452
diff --git a/records/sd1/sweep/5b9ca649a724d2e42f73/output.webp b/records/sd1/sweep/5b9ca649a724d2e42f73/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2a9c7ebe4a6d06386e1300e5796bb449d21cb7e1
--- /dev/null
+++ b/records/sd1/sweep/5b9ca649a724d2e42f73/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33df22ae460393a0b8d7d875914177dd33ed73440429b7f63000d310c1b158ae
+size 21064
diff --git a/records/sd1/sweep/5ba5332734e13dbd658a/output.webp b/records/sd1/sweep/5ba5332734e13dbd658a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bb99cc9164aad55dae00f059c4afbb8c8141bfe8
--- /dev/null
+++ b/records/sd1/sweep/5ba5332734e13dbd658a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82a5e9784727eb382e88c954b8804b55aaa58993c98f55dca3e117c1bdfede7e
+size 26536
diff --git a/records/sd1/sweep/5bad0595ecce3dbf9e5e/output.webp b/records/sd1/sweep/5bad0595ecce3dbf9e5e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d8ba91171bc92b387c68afb9f131db5f82d10dba
--- /dev/null
+++ b/records/sd1/sweep/5bad0595ecce3dbf9e5e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b3d59d97595cd49617f09eb3681960aa3e440efc4beae72db9812ef81e129e9
+size 17188
diff --git a/records/sd1/sweep/5bb705e5c4e07182d148/output.webp b/records/sd1/sweep/5bb705e5c4e07182d148/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..29c07ff4bf3af29f4890a44e5cf57df3d6a3eb88
--- /dev/null
+++ b/records/sd1/sweep/5bb705e5c4e07182d148/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc85aa9834b1ee5862f5564f1e88d47f9c5a8db05341e7d430f4ceeefef5a083
+size 17426
diff --git a/records/sd1/sweep/5bbf9d74f2fad820761a/output.webp b/records/sd1/sweep/5bbf9d74f2fad820761a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c92c4986b23ec3384123c936890088889e5b1d99
--- /dev/null
+++ b/records/sd1/sweep/5bbf9d74f2fad820761a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a277ffb33b6ff8b6b6845219edfb09e136eacfdbcbc21ca065fc3e48b800233d
+size 17102
diff --git a/records/sd1/sweep/5bc22e89c6226ac37d1b/output.webp b/records/sd1/sweep/5bc22e89c6226ac37d1b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b8cb5e1552d655cc8dbbc00733c3408a686ac9e5
--- /dev/null
+++ b/records/sd1/sweep/5bc22e89c6226ac37d1b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c95218f12afd3347b2181e37201d65abb6e5cb9f21cf41e1b09c4a009450d130
+size 26744
diff --git a/records/sd1/sweep/5bc75bf1ccf8826eba6f/output.webp b/records/sd1/sweep/5bc75bf1ccf8826eba6f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d416080c2a8c4246bc6b0482c7497273cc7aa1cc
--- /dev/null
+++ b/records/sd1/sweep/5bc75bf1ccf8826eba6f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11c7956b0aa06286e70dc8cc76831dce0130566a65fef5fc57e9630e541cfcf0
+size 36162
diff --git a/records/sd1/sweep/5bd08da592924f80439a/output.webp b/records/sd1/sweep/5bd08da592924f80439a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b1c0a656ecd96d6c66938bdd2e4ff9dfb7aad85f
--- /dev/null
+++ b/records/sd1/sweep/5bd08da592924f80439a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11b487aca92e03e075cd2e09abe17038d4bf39e14ece5649454e6773fa57ab83
+size 28758
diff --git a/records/sd1/sweep/5bd310d8f0cab90b55f2/output.webp b/records/sd1/sweep/5bd310d8f0cab90b55f2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..eccbe8b75b97578c16a224dadf23e90f95b5bae1
--- /dev/null
+++ b/records/sd1/sweep/5bd310d8f0cab90b55f2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7fde3b0c288ca224e4ea8b401d845eebf8702049ca0cb3d8aae3450960190fcc
+size 25260
diff --git a/records/sd1/sweep/5bd789561aa69f29dd1a/output.webp b/records/sd1/sweep/5bd789561aa69f29dd1a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2c50e6e08bbfee80dcb89bf3a191472fc6c2f2dc
--- /dev/null
+++ b/records/sd1/sweep/5bd789561aa69f29dd1a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c4f78f69788bbf0d9f7dbfcd563dbb02dfc183d9f23f4c8e7ef80ddbdb4aa7a
+size 36962
diff --git a/records/sd1/sweep/5bd87332e1fec21f3f59/output.webp b/records/sd1/sweep/5bd87332e1fec21f3f59/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..71f29bc55bcf4a2b6615b2799edbe56d61002c35
--- /dev/null
+++ b/records/sd1/sweep/5bd87332e1fec21f3f59/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6219acbadd56eb95f774a5b55ce120f15333c259255c079334b529293cecc2a
+size 37122
diff --git a/records/sd1/sweep/5bd98396e41d3d1f423c/output.webp b/records/sd1/sweep/5bd98396e41d3d1f423c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..72c0bc8635e747cff5234a7563d4d0f483eba318
--- /dev/null
+++ b/records/sd1/sweep/5bd98396e41d3d1f423c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ca99d4a5a9ab9e9e33d8a25a5f3823fa17abc730c9560e261be0727bdbf797d
+size 20502
diff --git a/records/sd1/sweep/5be2631376ecfc35236b/output.webp b/records/sd1/sweep/5be2631376ecfc35236b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..735e3d1238a26dbe36448f65a12880e99674c91c
--- /dev/null
+++ b/records/sd1/sweep/5be2631376ecfc35236b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec43852ecd32eb102132ca4a8578c3875da80feb2c5a77082419823565319e7b
+size 55824
diff --git a/records/sd1/sweep/5be7eb97e3ab1382c493/output.webp b/records/sd1/sweep/5be7eb97e3ab1382c493/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c641a3e5665100e7be877301315387a49eb570a3
--- /dev/null
+++ b/records/sd1/sweep/5be7eb97e3ab1382c493/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d5dd92a63f8b60f2f4de77f67daf4acffe514b2ac06f582ab36b2dc2c6ffbde
+size 15830
diff --git a/records/sd1/sweep/5beb07762b844322beaf/output.webp b/records/sd1/sweep/5beb07762b844322beaf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cd49365bbf8282bfe61c46d4fcbf5bac21d72055
--- /dev/null
+++ b/records/sd1/sweep/5beb07762b844322beaf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adea43a60f9f1151a6f6bdc4a55ce6578e57e38efcd45d32ada690451ae86a2c
+size 23776
diff --git a/records/sd1/sweep/5bf3cb2fbf858cdc31b3/output.webp b/records/sd1/sweep/5bf3cb2fbf858cdc31b3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9ad616d1e0ef7db8c1cfcf2eaa56ee785da5cdd9
--- /dev/null
+++ b/records/sd1/sweep/5bf3cb2fbf858cdc31b3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b14cc022c6baee43d1437694b94071529e148fe7cfd81e4fb69c95b9134f73e
+size 19932
diff --git a/records/sd1/sweep/5bf8802533e4c3d338d8/output.webp b/records/sd1/sweep/5bf8802533e4c3d338d8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..28e0d7d99980688030fe60df422afb1863a5b98d
--- /dev/null
+++ b/records/sd1/sweep/5bf8802533e4c3d338d8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a980177cf42153048cd4485f2b9ee199d0b827637f3adbcd118c4775a174a91
+size 15094
diff --git a/records/sd1/sweep/5bfcf957ae141a1099e3/output.webp b/records/sd1/sweep/5bfcf957ae141a1099e3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..823777314b2e4e27ab88f76dd5225387156cf999
--- /dev/null
+++ b/records/sd1/sweep/5bfcf957ae141a1099e3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0529d8cbdafa22552a7c546146af2b70453cf3d0a83e86b4a59e9b2e41066a34
+size 27682
diff --git a/records/sd1/sweep/5c004f148af4aab20a45/output.webp b/records/sd1/sweep/5c004f148af4aab20a45/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1738c5e5374673e1a50b535e71b864fa35bfec2a
--- /dev/null
+++ b/records/sd1/sweep/5c004f148af4aab20a45/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb0536d7b64f50f0d2664449da52a648c8a19cb1c3740d42ecba2dda1bc1eded
+size 29352
diff --git a/records/sd1/sweep/5c0b7d809fcd0148e69c/output.webp b/records/sd1/sweep/5c0b7d809fcd0148e69c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..302bc053c50c6555b6c430cc6b1df27362e6355e
--- /dev/null
+++ b/records/sd1/sweep/5c0b7d809fcd0148e69c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:737323b4ac7d6af5cc05f2e0342e5ff32ed773f04e353f6a113401a5522f5cab
+size 50352
diff --git a/records/sd1/sweep/5c0cd82edaae46698767/output.webp b/records/sd1/sweep/5c0cd82edaae46698767/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..55593ee40bcb20c74b6b52df5e544480ea266c54
--- /dev/null
+++ b/records/sd1/sweep/5c0cd82edaae46698767/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e4a0366c7cbcfa870f560d44577d0f5af8f0a1004896f52d5e7d870e9de27130
+size 56650
diff --git a/records/sd1/sweep/5c1225272cdbdb67696e/output.webp b/records/sd1/sweep/5c1225272cdbdb67696e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..66a1613ddb5bd832b52041f49b3a00141970d726
--- /dev/null
+++ b/records/sd1/sweep/5c1225272cdbdb67696e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d8de8f109d9fc39faaf5f36ae35f03114862148d5e7f0e467d0b3c934722616
+size 36814
diff --git a/records/sd1/sweep/5c26b4de57547ac7123d/output.webp b/records/sd1/sweep/5c26b4de57547ac7123d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d3d1f87ce1d23ce70de9c1eebe749bd8f4d8454f
--- /dev/null
+++ b/records/sd1/sweep/5c26b4de57547ac7123d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45859d46c653d4b167de7b7f16a1ca0de7769dc654ebfd7e1481f20f3a15c558
+size 29952
diff --git a/records/sd1/sweep/5c2d3e3b23f4b857a8be/output.webp b/records/sd1/sweep/5c2d3e3b23f4b857a8be/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..931064819bc6d94d52a4e3573d29ddef68c8c71a
--- /dev/null
+++ b/records/sd1/sweep/5c2d3e3b23f4b857a8be/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d452831082485eced7eedc8d2122e392b3817ab3798fa21697734fa1dfefeac5
+size 40656
diff --git a/records/sd1/sweep/5c399554804436463ef8/output.webp b/records/sd1/sweep/5c399554804436463ef8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a27d5d79cb7f31193f099e3fd56c40940b1fd79c
--- /dev/null
+++ b/records/sd1/sweep/5c399554804436463ef8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adab92a204aea13dcb5f3dfdd32b9d599cc8b09216af1599f09d8c4c1746e6dc
+size 39068
diff --git a/records/sd1/sweep/5c4233c3500b7429344d/output.webp b/records/sd1/sweep/5c4233c3500b7429344d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..78d5177d13fe05b20ec38fde4ca0686a1c8aaa5e
--- /dev/null
+++ b/records/sd1/sweep/5c4233c3500b7429344d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2efa02613a089a5a165fed98a8b0d98933ae1b8309823015ad118a300ac55b5
+size 31974
diff --git a/records/sd1/sweep/5c4686a2919e3bea308b/output.webp b/records/sd1/sweep/5c4686a2919e3bea308b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..84ffea6b2a7653fcbbf63562c14a78bf2745819a
--- /dev/null
+++ b/records/sd1/sweep/5c4686a2919e3bea308b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:048639397516539eeacd6081c7a4008ff72a5963d9b5c085847da3ee977ed560
+size 33680
diff --git a/records/sd1/sweep/5c4c8e4e40d4725711a8/output.webp b/records/sd1/sweep/5c4c8e4e40d4725711a8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e8543f8d4ecf02ce2a40d025e59f3cc654b59154
--- /dev/null
+++ b/records/sd1/sweep/5c4c8e4e40d4725711a8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ed9d68e63b3d144df72eb6eafb52278739cd1ee21036fbe9eb7c0395ae39b46
+size 33188
diff --git a/records/sd1/sweep/5c53cae57671aa55d7c2/output.webp b/records/sd1/sweep/5c53cae57671aa55d7c2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..129aa641d31cb265cdaa999da75bfe75faf4faf1
--- /dev/null
+++ b/records/sd1/sweep/5c53cae57671aa55d7c2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:deec2d482dc424921878e54abd5bd132219baf5d64c243bf16a7e04da5ea6e3a
+size 15102
diff --git a/records/sd1/sweep/5c53cc95a95edf74dcc2/output.webp b/records/sd1/sweep/5c53cc95a95edf74dcc2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b5c09789db74f96c903fc7328e15c1ea0c27dd61
--- /dev/null
+++ b/records/sd1/sweep/5c53cc95a95edf74dcc2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25128cf861b7b060ecb8d392d88707c72884121623c861bd74e021bc35280076
+size 46772
diff --git a/records/sd1/sweep/5c55e15ab4ecea4b1361/output.webp b/records/sd1/sweep/5c55e15ab4ecea4b1361/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8b2c1537ca6b9df1c1e21fd7796e22a710ef8436
--- /dev/null
+++ b/records/sd1/sweep/5c55e15ab4ecea4b1361/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f22f2008fed5eb350eb206159659439e8121c4e465422e87fcd13391905512d
+size 19774
diff --git a/records/sd1/sweep/5c674e2d12035bbf6c2c/output.webp b/records/sd1/sweep/5c674e2d12035bbf6c2c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8c023e37789b92f188002966350b59b69e8e477b
--- /dev/null
+++ b/records/sd1/sweep/5c674e2d12035bbf6c2c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:affddf5232d293712bd952d1c4c9a6aed999b25fb88d541dff2edafe10e4ecfc
+size 49722
diff --git a/records/sd1/sweep/5c6e6c2d2d885bb01dc1/output.webp b/records/sd1/sweep/5c6e6c2d2d885bb01dc1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dfb24016955ccb5433098c90501e19f0a1972d83
--- /dev/null
+++ b/records/sd1/sweep/5c6e6c2d2d885bb01dc1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff5d1e23d4fa5ed88fe88e21897495d31411f3941525c73852281ece91165410
+size 28832
diff --git a/records/sd1/sweep/5c75d98753e8c9293053/output.webp b/records/sd1/sweep/5c75d98753e8c9293053/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e55986498d7d5a0d97fd8414243201e8ae17f8ff
--- /dev/null
+++ b/records/sd1/sweep/5c75d98753e8c9293053/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b450508c42ffb5593f8fa6c031576d374eae1939f5767f7c9f0c28743e63e16b
+size 17634
diff --git a/records/sd1/sweep/5c891f4a525eb6edf6ca/output.webp b/records/sd1/sweep/5c891f4a525eb6edf6ca/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7564f76bf5f4f1dcde6f6973a5dbf8bdb2763220
--- /dev/null
+++ b/records/sd1/sweep/5c891f4a525eb6edf6ca/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ed3c78edb96277a40af99c4eff4190dc58eb0749141ef0ab4b431cf6025f69c
+size 17808
diff --git a/records/sd1/sweep/5c96a9980d4f5f5315b1/output.webp b/records/sd1/sweep/5c96a9980d4f5f5315b1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1018d9fe209a49e42350ce7d61a6e15fc580200b
--- /dev/null
+++ b/records/sd1/sweep/5c96a9980d4f5f5315b1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd5efab094d06e7529fda481ef138299d806ee7d3f30aad349a67eecdb349b76
+size 26664
diff --git a/records/sd1/sweep/5ca327b7c72c033f925e/output.webp b/records/sd1/sweep/5ca327b7c72c033f925e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c4dc021ed16d9f7ee5455d4c256a035ad0d048a9
--- /dev/null
+++ b/records/sd1/sweep/5ca327b7c72c033f925e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a86532b9afb63479cd8c34e1e4fc3153baa5008f7a947a009f9edcd1690a9725
+size 51586
diff --git a/records/sd1/sweep/5ca7a1510172d9d7a423/output.webp b/records/sd1/sweep/5ca7a1510172d9d7a423/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5854d051b03957a133e666ec171f00763ceb5017
--- /dev/null
+++ b/records/sd1/sweep/5ca7a1510172d9d7a423/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18c64e2dfdda6dc1aaa787a4dc7cea047ea910a590611fbd30bc4b89abe425ac
+size 45112
diff --git a/records/sd1/sweep/5ca8fc00870bf9a6dbf4/output.webp b/records/sd1/sweep/5ca8fc00870bf9a6dbf4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..14768820f0bb4b696f9f01ed5a4c066daa081890
--- /dev/null
+++ b/records/sd1/sweep/5ca8fc00870bf9a6dbf4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2729ae07bdcd226fb41d1c2eb1df4f4caad0c4de8d985c265b13356bafc3199b
+size 36498
diff --git a/records/sd1/sweep/5cd683e8b39f020243df/output.webp b/records/sd1/sweep/5cd683e8b39f020243df/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..76a1b4a47ad2b7b14515df798f2f30e0344e834d
--- /dev/null
+++ b/records/sd1/sweep/5cd683e8b39f020243df/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5b9ac3ca7f24d7110a0080b491a437ba45ea8c90bbaf94ce7e0ae40b25b0f830
+size 15274
diff --git a/records/sd1/sweep/5cedcb541faed40e55ad/output.webp b/records/sd1/sweep/5cedcb541faed40e55ad/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3d537fe6c906ebe5939bea6396039c55c8c17b78
--- /dev/null
+++ b/records/sd1/sweep/5cedcb541faed40e55ad/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e1ed9e920703509f98321288da1e37a0ba4b94e0e9ffbdfbadb42d3f903f8bc
+size 54650
diff --git a/records/sd1/sweep/5cedf8f4ee18a146d10e/output.webp b/records/sd1/sweep/5cedf8f4ee18a146d10e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fda617595e2335db3b6408d8e337da5f95554989
--- /dev/null
+++ b/records/sd1/sweep/5cedf8f4ee18a146d10e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:84f69ec10f84a8d63552d1adad934d3b27df28a319e84acc20f1c72b59ca4332
+size 17848
diff --git a/records/sd1/sweep/5cf11e6056f8a854f1ed/output.webp b/records/sd1/sweep/5cf11e6056f8a854f1ed/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6123d27b7445c2727018ec882fbec8953510fb18
--- /dev/null
+++ b/records/sd1/sweep/5cf11e6056f8a854f1ed/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac20f90ce4f1ddf1f4330a77756a8039b8c3bc8a094be86bf497d7148f693740
+size 53152
diff --git a/records/sd1/sweep/5cf98e118f78e3805e4d/output.webp b/records/sd1/sweep/5cf98e118f78e3805e4d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e820af533f7b95f6227fc45be293a90fc558683f
--- /dev/null
+++ b/records/sd1/sweep/5cf98e118f78e3805e4d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0bff57277d5303eccccb028189b94c49b0c94d1fb1fa2b7b23884f66c584e63
+size 46574
diff --git a/records/sd1/sweep/5cfe34a26a7fa2a3b1e6/output.webp b/records/sd1/sweep/5cfe34a26a7fa2a3b1e6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f1e03175575da3b83d167608ed48f3119b42c6bb
--- /dev/null
+++ b/records/sd1/sweep/5cfe34a26a7fa2a3b1e6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a05afcb35b8e9ab127eb809a5ab57c2b452fb41f37cac02d9874b8f73a9b7131
+size 19934
diff --git a/records/sd1/sweep/5d1910c4485073103886/output.webp b/records/sd1/sweep/5d1910c4485073103886/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5a13e6a29be076bb9d201879df7292311776d25b
--- /dev/null
+++ b/records/sd1/sweep/5d1910c4485073103886/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbb6a1c31a1e21e39a15b9bcc2ebf31affe51123c6583e78738050cde20c5d35
+size 19678
diff --git a/records/sd1/sweep/5d1daf0b55a44880c153/output.webp b/records/sd1/sweep/5d1daf0b55a44880c153/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1a609770f12db7ea336d7626bf7da27141e20be4
--- /dev/null
+++ b/records/sd1/sweep/5d1daf0b55a44880c153/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07c2e69f0d0d40142689c8086df2ca9ea0376ef8f001f54be41a5e6c028c9be4
+size 42764
diff --git a/records/sd1/sweep/5d330c41852d60eda000/output.webp b/records/sd1/sweep/5d330c41852d60eda000/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0b2b68bcbde864e6372d948db4bda089e93e58a1
--- /dev/null
+++ b/records/sd1/sweep/5d330c41852d60eda000/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d969b621cb888b8bdb92a39d309583a2c7cfc21ef5fdc11de69c8055fbf255f1
+size 41700
diff --git a/records/sd1/sweep/5d3b2d499be15674db8f/output.webp b/records/sd1/sweep/5d3b2d499be15674db8f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e417c05af2cf14e3bcf87cfa4de6d8338187bb65
--- /dev/null
+++ b/records/sd1/sweep/5d3b2d499be15674db8f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:929772600d810026e9006044ff15ce7c68812287d33dd48531ac7b31279b6566
+size 45508
diff --git a/records/sd1/sweep/5d420902221af94645b2/output.webp b/records/sd1/sweep/5d420902221af94645b2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d11b4f41ed46e8157457d6273df48e557994b77d
--- /dev/null
+++ b/records/sd1/sweep/5d420902221af94645b2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a88200fdc53cf52cea2cd0e8daa9e1da126e624e1b460231ab9ff9d40149f0b
+size 23638
diff --git a/records/sd1/sweep/5d48aeb78917bcf0a139/output.webp b/records/sd1/sweep/5d48aeb78917bcf0a139/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c044715cc346c49e6239d923a110e8a1a3b993f8
--- /dev/null
+++ b/records/sd1/sweep/5d48aeb78917bcf0a139/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfb103ade9f69de5fb113602834afa8b96802510342c05b28fda6d9072544057
+size 33052
diff --git a/records/sd1/sweep/5d4a67abbc8dd712a8ac/output.webp b/records/sd1/sweep/5d4a67abbc8dd712a8ac/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e726864d9d49ddb578ce772db944610d8892cc0c
--- /dev/null
+++ b/records/sd1/sweep/5d4a67abbc8dd712a8ac/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd5c2236c09c31115bdaf2c9806234d9130131160dc9f9173a19402bbd61f615
+size 25752
diff --git a/records/sd1/sweep/5d4cf2e5c06e7f1dc0dc/output.webp b/records/sd1/sweep/5d4cf2e5c06e7f1dc0dc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..51d61a0b1bec7201eced6c80f788c0df063127db
--- /dev/null
+++ b/records/sd1/sweep/5d4cf2e5c06e7f1dc0dc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9ee048552d89515d06d3b6d869f4ee748fe56263afc72e081e5636176486098
+size 39896
diff --git a/records/sd1/sweep/5d50c77f0e43a1c6e504/output.webp b/records/sd1/sweep/5d50c77f0e43a1c6e504/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fa85df26e6bb3026e9334debe10a71f82b828887
--- /dev/null
+++ b/records/sd1/sweep/5d50c77f0e43a1c6e504/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52d7cc92ff1ea6fd16ffe9b816b4825ceaf9200a1e16e91d8755d930b8985754
+size 18808
diff --git a/records/sd1/sweep/5d52527a4e3ef933f3f8/output.webp b/records/sd1/sweep/5d52527a4e3ef933f3f8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b87c0d8ceae688f579221486626c922876aab445
--- /dev/null
+++ b/records/sd1/sweep/5d52527a4e3ef933f3f8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfc2c4f6a63efa76a1668a92a5fac10732f6d8bfffb047f652520bea84163686
+size 26420
diff --git a/records/sd1/sweep/5d55fd417849fbb3235d/output.webp b/records/sd1/sweep/5d55fd417849fbb3235d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..800e17b013c8ea33afe985131f036c15cc6aba78
--- /dev/null
+++ b/records/sd1/sweep/5d55fd417849fbb3235d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2634455c82624e74e8730556ecaad544815f9353b9e9653994b1455c89dbd680
+size 19526
diff --git a/records/sd1/sweep/5d5a8932a03928a89d3b/output.webp b/records/sd1/sweep/5d5a8932a03928a89d3b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..34cf25c7624c2557b8e777e6ac8f13e6dd7afff9
--- /dev/null
+++ b/records/sd1/sweep/5d5a8932a03928a89d3b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0eefc943768fc30ee073b38f928b79825250d8e28b87ced4863794ed23493599
+size 47292
diff --git a/records/sd1/sweep/5d6ab4799a5d127986a4/output.webp b/records/sd1/sweep/5d6ab4799a5d127986a4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..336926aae802ff21e75c0b19d9fc0e72f3351504
--- /dev/null
+++ b/records/sd1/sweep/5d6ab4799a5d127986a4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e9bcc0908240359e0551c0fc28bb4434f2d2b081bd42acd40fb5e70f01ab736
+size 32208
diff --git a/records/sd1/sweep/5d7c6492345a85e0ba14/output.webp b/records/sd1/sweep/5d7c6492345a85e0ba14/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9ff468dd0dbefd92333ae50f6fed87aff8f07682
--- /dev/null
+++ b/records/sd1/sweep/5d7c6492345a85e0ba14/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86461b9fe3abe5d681f3cda332439451943e76e6479a809eac10506cfc3cdfff
+size 20018
diff --git a/records/sd1/sweep/5d8aa03e68d7a3685f10/output.webp b/records/sd1/sweep/5d8aa03e68d7a3685f10/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4fb9cbfb6d70658d5cc9ee02ea4d93998e9f378a
--- /dev/null
+++ b/records/sd1/sweep/5d8aa03e68d7a3685f10/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42f63137430269382ee2fa866eaddb63173933eb7b908a492fda7f56b2762208
+size 39192
diff --git a/records/sd1/sweep/5d913653850ede1ab9c1/output.webp b/records/sd1/sweep/5d913653850ede1ab9c1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..73167fe227c958076583cde5aa62a9fea97396b2
--- /dev/null
+++ b/records/sd1/sweep/5d913653850ede1ab9c1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aeb3318c6b74fe27934f03c83f057275114abedaa8710d47b101984fe837986b
+size 44298
diff --git a/records/sd1/sweep/5d98eef6bed32214a524/output.webp b/records/sd1/sweep/5d98eef6bed32214a524/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1e0eb813b4f9cb08d1f87c2fa91f8d966dbb729a
--- /dev/null
+++ b/records/sd1/sweep/5d98eef6bed32214a524/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:096c17d18870fa007e42ebdbc89ec062dc86f770fc2c6756f19165df7f6c1c7d
+size 45486
diff --git a/records/sd1/sweep/5d9b568932675f1759eb/output.webp b/records/sd1/sweep/5d9b568932675f1759eb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..16f2db71467530fd744b39ae41e497fe89c22c9c
--- /dev/null
+++ b/records/sd1/sweep/5d9b568932675f1759eb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64740e904b79ee883685a38fe2045d72b818ecc764fd6dddfb9735f39507319c
+size 20772
diff --git a/records/sd1/sweep/5da966deaca638185e50/output.webp b/records/sd1/sweep/5da966deaca638185e50/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6a7e61ddc250fbdc02ac84d987d93cd0dde1aa75
--- /dev/null
+++ b/records/sd1/sweep/5da966deaca638185e50/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:511356d4014cac4ab4f36845573dd1058ac12eca498427945d5168047c7099da
+size 28170
diff --git a/records/sd1/sweep/5dad2ad9cb9284b84885/output.webp b/records/sd1/sweep/5dad2ad9cb9284b84885/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..080974f001184ed43a3fbb36786392b820c01ab6
--- /dev/null
+++ b/records/sd1/sweep/5dad2ad9cb9284b84885/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48120f7f289b9c32e7ff0046656dd5987eb73af3f1bd85a1057a62a9f6c603b7
+size 26406
diff --git a/records/sd1/sweep/5dada8be10447493bf89/output.webp b/records/sd1/sweep/5dada8be10447493bf89/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fd087a2d550a4b326c88d476ddf2e0f1698fde8f
--- /dev/null
+++ b/records/sd1/sweep/5dada8be10447493bf89/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a2a053a7de6fe94fd3be957788746c8be8edd169d52584e5fce397e95138ba6
+size 35768
diff --git a/records/sd1/sweep/5db2fa9d8ff82e84b32d/output.webp b/records/sd1/sweep/5db2fa9d8ff82e84b32d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..81bd141a403f620f9406c7c92e27fb472d27ec7c
--- /dev/null
+++ b/records/sd1/sweep/5db2fa9d8ff82e84b32d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d8784e0538d990dba74423ea0b948545f2c50efe83ace790c7dc17b67a084ec
+size 40490
diff --git a/records/sd1/sweep/5dbb4e73af61f1b6d111/output.webp b/records/sd1/sweep/5dbb4e73af61f1b6d111/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..86c98034e79fc750ae9be11aed57e632bc65cf94
--- /dev/null
+++ b/records/sd1/sweep/5dbb4e73af61f1b6d111/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16bca97cc625224db65c49ebee9530a6963359dad39ad2e4e8496eed29a24996
+size 38338
diff --git a/records/sd1/sweep/5dbee34f2677d06716d4/output.webp b/records/sd1/sweep/5dbee34f2677d06716d4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e28f7784089c87ba2828e484763f93e41018dc06
--- /dev/null
+++ b/records/sd1/sweep/5dbee34f2677d06716d4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:47dcb362032bc5f7187db4fb131ecb345ace53c53c44c01a671bdf9aa0c4654c
+size 57616
diff --git a/records/sd1/sweep/5dc69f8fece9f4f8f32c/output.webp b/records/sd1/sweep/5dc69f8fece9f4f8f32c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..84610fb8dcf66ab8c9e364f990950daf1261ac2b
--- /dev/null
+++ b/records/sd1/sweep/5dc69f8fece9f4f8f32c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e16c111c0f4e67b5684f8e413e057a9234e2b0d815aaa0e165a8a7f7de15761
+size 38368
diff --git a/records/sd1/sweep/5dca82cf481540a5d10e/output.webp b/records/sd1/sweep/5dca82cf481540a5d10e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cc6cc577cc94054b3e5d6bee34295bc522533a28
--- /dev/null
+++ b/records/sd1/sweep/5dca82cf481540a5d10e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b689d8ae77f8ede6197f86dd1b1656b769bd3140eb07cdf0c291c4521a8dba4d
+size 17764
diff --git a/records/sd1/sweep/5dcb5eba529bb4b08e00/output.webp b/records/sd1/sweep/5dcb5eba529bb4b08e00/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..40e9906f17769bdd73b10c89d043ac983f87cb9f
--- /dev/null
+++ b/records/sd1/sweep/5dcb5eba529bb4b08e00/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f6aa3638593d632d819f34b2edb213da39ec10cf8109b15c438f32210d300f0
+size 27618
diff --git a/records/sd1/sweep/5dcc5f3e32792f22168e/output.webp b/records/sd1/sweep/5dcc5f3e32792f22168e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7d2bbb49b1901513dfdce4eff903a902d7396518
--- /dev/null
+++ b/records/sd1/sweep/5dcc5f3e32792f22168e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ceaaeabcbfe24b7a4797e1ef2d985329397a1494c7532a6d4959749055ddb183
+size 36504
diff --git a/records/sd1/sweep/5de55890df6de428ffa9/output.webp b/records/sd1/sweep/5de55890df6de428ffa9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4a98236da62236c6dc379e166e67de38e05d3dc3
--- /dev/null
+++ b/records/sd1/sweep/5de55890df6de428ffa9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c29d815ad28a8d1c52e364c394b4c5c78cd2d1a767052c3148cd7bd05a071576
+size 18738
diff --git a/records/sd1/sweep/5de56edaea52cc9eb2db/output.webp b/records/sd1/sweep/5de56edaea52cc9eb2db/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..03fd834a63040b1e6f9ba961d4d2a49e0bc5312c
--- /dev/null
+++ b/records/sd1/sweep/5de56edaea52cc9eb2db/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:590531942b38b7bf43e625318575b7e27cc1b4632148b076ff37aaf66a480546
+size 23714
diff --git a/records/sd1/sweep/5de5914887d25949809b/output.webp b/records/sd1/sweep/5de5914887d25949809b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..40e9c0a2eefeba64c981c3eb154175484d1f2fac
--- /dev/null
+++ b/records/sd1/sweep/5de5914887d25949809b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e46b4feab5882341260502f3ad02fedfc72a46b26cc65d9f4e056708cc03c1dc
+size 21324
diff --git a/records/sd1/sweep/5df35a96209f87b23133/output.webp b/records/sd1/sweep/5df35a96209f87b23133/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..788764d5e8890da9af5acd743f133f408e533d09
--- /dev/null
+++ b/records/sd1/sweep/5df35a96209f87b23133/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45287abdda02fc495386e88b27e2e32d73649fbc4e44f7c3a247db6386932224
+size 27466
diff --git a/records/sd1/sweep/5dfffea679984872bb06/output.webp b/records/sd1/sweep/5dfffea679984872bb06/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b275a6034a656c2f50ed1ce95f3d8dccd38f9f49
--- /dev/null
+++ b/records/sd1/sweep/5dfffea679984872bb06/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebaf62ea8a61be167e710f0418d0ba61e2c37716efc729b5efb9da6f457e7809
+size 36194
diff --git a/records/sd1/sweep/5e02a6415b0ef2e9ee8a/output.webp b/records/sd1/sweep/5e02a6415b0ef2e9ee8a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..89bb282fc63530853899360b67e184ba624aae9a
--- /dev/null
+++ b/records/sd1/sweep/5e02a6415b0ef2e9ee8a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e19a22d5b7d12952e162a0a2650a5a501c27deb2f1a2318b1eac03c8c6587e40
+size 36504
diff --git a/records/sd1/sweep/5e0bd8b8825fb780112d/output.webp b/records/sd1/sweep/5e0bd8b8825fb780112d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e6100d77506e2c740b9b138ee9670ccd5a4384f5
--- /dev/null
+++ b/records/sd1/sweep/5e0bd8b8825fb780112d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e621f2cd35a5b18f030e14eeff48a16ab65a316bca90d535c7aa555f52636fd8
+size 32930
diff --git a/records/sd1/sweep/5e0dd89b18ed3aa51422/output.webp b/records/sd1/sweep/5e0dd89b18ed3aa51422/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..631754c4ec3aba31d9bc374b36ced41e35594d49
--- /dev/null
+++ b/records/sd1/sweep/5e0dd89b18ed3aa51422/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e63361fd8ea1d3ede5db5a7bae857fb7b898937e057bb8e26b5e5e54b505a49
+size 19936
diff --git a/records/sd1/sweep/5e10f236b44488454a9f/output.webp b/records/sd1/sweep/5e10f236b44488454a9f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1ba1b62e008935579af4f6bcf0ab90c5229be8f9
--- /dev/null
+++ b/records/sd1/sweep/5e10f236b44488454a9f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cf71fd36363bd32b12edb6ecf4d0ae63e0c579022eaf08cbf63a55e797ba0c4
+size 18890
diff --git a/records/sd1/sweep/5e1250f1c0c0a3307186/output.webp b/records/sd1/sweep/5e1250f1c0c0a3307186/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..24d8ceeb3952d087647b6bf58b7fe7e127aa61fb
--- /dev/null
+++ b/records/sd1/sweep/5e1250f1c0c0a3307186/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c862cd5e67c2f97f43aa79d377108ca68700e5190eea7e49ddcbcbd697d6f3b7
+size 28392
diff --git a/records/sd1/sweep/5e20751d44de8a76ec6a/output.webp b/records/sd1/sweep/5e20751d44de8a76ec6a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0aaafc7df9f6896207aac9b5efef4097095f4872
--- /dev/null
+++ b/records/sd1/sweep/5e20751d44de8a76ec6a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8de351f2ca2a040b391e5f6f45d8256c51081957da8d9dc3c096d08ff6624fdf
+size 40888
diff --git a/records/sd1/sweep/5e236125e1c8a61345cf/output.webp b/records/sd1/sweep/5e236125e1c8a61345cf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1586c53fa02b05be3be06a2af25e95db99d681e0
--- /dev/null
+++ b/records/sd1/sweep/5e236125e1c8a61345cf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:151ea14859300d0239b0cb36f6635fb40c3b6b810b709308e5449c5e5ec2beef
+size 40236
diff --git a/records/sd1/sweep/5e23c9a007ad8ae7a469/output.webp b/records/sd1/sweep/5e23c9a007ad8ae7a469/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bdca67f6f2cd799008e4d584c551c3d7616b8546
--- /dev/null
+++ b/records/sd1/sweep/5e23c9a007ad8ae7a469/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c5f6ad384771dae3d4188d4fd965ddc4c60a39c43dfb5e225f284a7441a45d79
+size 41836
diff --git a/records/sd1/sweep/5e2c9053af71efdb289e/output.webp b/records/sd1/sweep/5e2c9053af71efdb289e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a5987588ebff67884f39c850a5d4fa21ba82ca5f
--- /dev/null
+++ b/records/sd1/sweep/5e2c9053af71efdb289e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c5f9fb6454ce6fb49818c148c41e1b674d8d87d543773bcdf45972acb137a76
+size 28674
diff --git a/records/sd1/sweep/5e3076579076d2a11e3b/output.webp b/records/sd1/sweep/5e3076579076d2a11e3b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..95d90ee032624293a76483ad23145074aee98ebb
--- /dev/null
+++ b/records/sd1/sweep/5e3076579076d2a11e3b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:48a2f2eaeab0f88af5a1953aa0701b95eaa37f84dbf276cd9df7a1629a823264
+size 31878
diff --git a/records/sd1/sweep/5e393bf924e25ae301e7/output.webp b/records/sd1/sweep/5e393bf924e25ae301e7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6f4a20a2dee6b66dc0b48870c44df0dddd824309
--- /dev/null
+++ b/records/sd1/sweep/5e393bf924e25ae301e7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c16746cad4658cb00ed7aa6e2c98ccb5ee53cbd10c2b372b4ae2c654a8a1be5
+size 58210
diff --git a/records/sd1/sweep/5e3a18faa0a0f72e0c31/output.webp b/records/sd1/sweep/5e3a18faa0a0f72e0c31/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5eace294d3f0e1a40d33fd03e61f5339f40590bf
--- /dev/null
+++ b/records/sd1/sweep/5e3a18faa0a0f72e0c31/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4132e18b7d9c0f3197a175a1c35c5602ce97b30dc7ecc1c4fb500e1e34f0b387
+size 35552
diff --git a/records/sd1/sweep/5e3ef4347ad8befe7af6/output.webp b/records/sd1/sweep/5e3ef4347ad8befe7af6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5cf30053ac6b86074901d7bb8e2ee2f8c3178318
--- /dev/null
+++ b/records/sd1/sweep/5e3ef4347ad8befe7af6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c17a4250a1c8c325f5517bb5a41e367fa808acd50366ca7f757a7f9bd47524ee
+size 36606
diff --git a/records/sd1/sweep/5e4b202f379457fef1bf/output.webp b/records/sd1/sweep/5e4b202f379457fef1bf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3afb2d26a9531d5bbb3ad0603dbe0c5afe5337e6
--- /dev/null
+++ b/records/sd1/sweep/5e4b202f379457fef1bf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dba4bf355ba9b455666ea25ed0a61565b3e978e1151e272c809639f27827c21f
+size 20370
diff --git a/records/sd1/sweep/5e54649d73049851f2cf/output.webp b/records/sd1/sweep/5e54649d73049851f2cf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b482aea3c4761d1de5ee319bbc8456e2c332fbbf
--- /dev/null
+++ b/records/sd1/sweep/5e54649d73049851f2cf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e8b37230e5e3580aa8e0ad09d2e04875eeb264b3c175c563d087b01a1d8e779
+size 19076
diff --git a/records/sd1/sweep/5e5edff93035d4dced32/output.webp b/records/sd1/sweep/5e5edff93035d4dced32/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2c3d38be3045fe317526067f65cba7ab13d661cc
--- /dev/null
+++ b/records/sd1/sweep/5e5edff93035d4dced32/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:63e42721de5d8b9b3ab56dc4b062c4fd477468cae67c63c02e375b8e2b771f46
+size 38502
diff --git a/records/sd1/sweep/5e821db6827cc41ee07d/output.webp b/records/sd1/sweep/5e821db6827cc41ee07d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cbdbd9002f0f668a5b4913d23a9d2462032818e5
--- /dev/null
+++ b/records/sd1/sweep/5e821db6827cc41ee07d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73a3fbc27e80d78b1799745fc2a4bcccc60d4efcad4bc19917b3a16e27f6bf5e
+size 40284
diff --git a/records/sd1/sweep/5e94d88621865ba9b97e/output.webp b/records/sd1/sweep/5e94d88621865ba9b97e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..43852a5dc805f150b7497bbe4d829d3766c7a2ed
--- /dev/null
+++ b/records/sd1/sweep/5e94d88621865ba9b97e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0140484aa4517577f61e9b1f26e8659e2513dcf36b007c03b3209813ad57b722
+size 22008
diff --git a/records/sd1/sweep/5e996d1ce220ebc83309/output.webp b/records/sd1/sweep/5e996d1ce220ebc83309/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c775a2db89b5e56e17ca78df000dbf34c3569576
--- /dev/null
+++ b/records/sd1/sweep/5e996d1ce220ebc83309/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4c0acf0a3662d9903c70a81deb57398e1840202e73022c29bfcea9aa72ed5ea
+size 25704
diff --git a/records/sd1/sweep/5e9a1c07a2c3579deb79/output.webp b/records/sd1/sweep/5e9a1c07a2c3579deb79/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..83c93f9e88dad0a97eddaa6adcf07f979e2341ca
--- /dev/null
+++ b/records/sd1/sweep/5e9a1c07a2c3579deb79/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:099f78ab526c768de34f4ca670bf28794b0eecf74902538e1076b2532ffd8613
+size 39796
diff --git a/records/sd1/sweep/5ea3cbf74e29eff72969/output.webp b/records/sd1/sweep/5ea3cbf74e29eff72969/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..80e15137949aa0430f5963eb32727b6457a61459
--- /dev/null
+++ b/records/sd1/sweep/5ea3cbf74e29eff72969/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b434e44c01802987df067b6e691a2b1a51031323a95e87acc29e539072a3abcf
+size 58348
diff --git a/records/sd1/sweep/5eaad04ee57a17c0b152/output.webp b/records/sd1/sweep/5eaad04ee57a17c0b152/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e92e220a460e3ce396ff855db7f29dcf6f54f7f1
--- /dev/null
+++ b/records/sd1/sweep/5eaad04ee57a17c0b152/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3285042de949a1f91a91137569fd88d84dc0a82695942c77378194d70f30129f
+size 47290
diff --git a/records/sd1/sweep/5eaad2c289ef12ec1a94/output.webp b/records/sd1/sweep/5eaad2c289ef12ec1a94/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8409fd1e3401b4a5442b113b9b89ad1a5df6d179
--- /dev/null
+++ b/records/sd1/sweep/5eaad2c289ef12ec1a94/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a78c07bedf5eb210ba6d1b9f06b2c16b342dfb97e0a2c4fdf51506d71081bc8
+size 31690
diff --git a/records/sd1/sweep/5eb220d49d771f273469/output.webp b/records/sd1/sweep/5eb220d49d771f273469/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..81003c2de0d2ca775665605b5cbffa2480389f4c
--- /dev/null
+++ b/records/sd1/sweep/5eb220d49d771f273469/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3e6896f52d0fa2c8d71b38aeeee71530df7094f6c5199b4d3e219e287781bff
+size 39472
diff --git a/records/sd1/sweep/5eb82a197abbdc46267c/output.webp b/records/sd1/sweep/5eb82a197abbdc46267c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..676f8eb3037f87cd329a3d7398d30c7d307b7522
--- /dev/null
+++ b/records/sd1/sweep/5eb82a197abbdc46267c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b9172d7ade3b0df660c2d205323cd995ece37cf2f90b25105f44658efe6eae6e
+size 30204
diff --git a/records/sd1/sweep/5ec4de85c6028c53ca37/output.webp b/records/sd1/sweep/5ec4de85c6028c53ca37/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..574354f31d424d843df576d0db48d0b97f4e6ecf
--- /dev/null
+++ b/records/sd1/sweep/5ec4de85c6028c53ca37/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:caf39f3b9807a29ef12ab5a52b94743a71988044aa5168613601b45a8b9ccbf5
+size 45074
diff --git a/records/sd1/sweep/5ec8afc3f1f62e5c8090/output.webp b/records/sd1/sweep/5ec8afc3f1f62e5c8090/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..90187a3f3a3777ce06fddddb0007590e0cb843d5
--- /dev/null
+++ b/records/sd1/sweep/5ec8afc3f1f62e5c8090/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac530347921c4bbfbc4c939ef24b8d2f569a18d7be8b132cf17adc5d47228586
+size 49626
diff --git a/records/sd1/sweep/5ecc5ac0d6db26c8c99f/output.webp b/records/sd1/sweep/5ecc5ac0d6db26c8c99f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..935f9baf51fb967c776b8f20c18e91e5fc74c25e
--- /dev/null
+++ b/records/sd1/sweep/5ecc5ac0d6db26c8c99f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a36b500b9a9a584fbc8aff1f3b18493e4bf74af6a781632c7d002b77725ce838
+size 38506
diff --git a/records/sd1/sweep/5ece1ab2e3ae2d1c14a8/output.webp b/records/sd1/sweep/5ece1ab2e3ae2d1c14a8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8254e7be023054662ad4cc26ddebe6a9218e0ca3
--- /dev/null
+++ b/records/sd1/sweep/5ece1ab2e3ae2d1c14a8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d25a0e20c77188cc12a8c83c867aa575ee063eddb1e3be93afd2257ed4a1d00
+size 30186
diff --git a/records/sd1/sweep/5ece74e2df222862d591/output.webp b/records/sd1/sweep/5ece74e2df222862d591/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b078f92277d5e9d5bdcb91c7330d33c16293b4e5
--- /dev/null
+++ b/records/sd1/sweep/5ece74e2df222862d591/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:927544ff102baa285ef957a73d12e05ec0e5368d1cf2050a6ae53f5ea2958dc2
+size 57636
diff --git a/records/sd1/sweep/5ecf7d7724e827db318e/output.webp b/records/sd1/sweep/5ecf7d7724e827db318e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ee88f57bb6629273b2638087f5d58a9e6656d475
--- /dev/null
+++ b/records/sd1/sweep/5ecf7d7724e827db318e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85cafd3fdf9ccb8d15cf214e417359b7076eded989c967b01fc7717d11e03583
+size 36124
diff --git a/records/sd1/sweep/5ed009f508e6509033cf/output.webp b/records/sd1/sweep/5ed009f508e6509033cf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2a61d7d6c3d54b89d777498e6a27cc97206c3be1
--- /dev/null
+++ b/records/sd1/sweep/5ed009f508e6509033cf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a8d1b74d21e2a73453451b62916e576bf1469cc9c763cdb257b932479c04114b
+size 22412
diff --git a/records/sd1/sweep/5ed255c974e03307b05a/output.webp b/records/sd1/sweep/5ed255c974e03307b05a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..18e90004d87c654c55e640cf214b23242757437e
--- /dev/null
+++ b/records/sd1/sweep/5ed255c974e03307b05a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b84b2b569251ad844bbf8cbe081ae6cdf7439daab62b6ee5365d20bc681d4775
+size 20852
diff --git a/records/sd1/sweep/5ee4543e11bd0bb47fbd/output.webp b/records/sd1/sweep/5ee4543e11bd0bb47fbd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a49bd941d7140cc63866533482145efdaadeac64
--- /dev/null
+++ b/records/sd1/sweep/5ee4543e11bd0bb47fbd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d3defc15235355624d30ee542b00a848127a25772e90fc34b2c273c39a0c8718
+size 32242
diff --git a/records/sd1/sweep/5ef0c1c7e966fa896093/output.webp b/records/sd1/sweep/5ef0c1c7e966fa896093/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a8f44b05e925d39507037554ad2086504771c444
--- /dev/null
+++ b/records/sd1/sweep/5ef0c1c7e966fa896093/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00aa864f7a97ebd6b8a6be41dcbd0b7739fa6e4be0b192a7b32157f000b9e8ae
+size 21536
diff --git a/records/sd1/sweep/5efc348ae779b0eb280e/output.webp b/records/sd1/sweep/5efc348ae779b0eb280e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ad76abcfd9303229bff90bf67efd00102155a6dc
--- /dev/null
+++ b/records/sd1/sweep/5efc348ae779b0eb280e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eac0c13377f416ee7997daee4f07678cfc94f38a74ebbe89df84491c9af4d5e3
+size 39672
diff --git a/records/sd1/sweep/5efeb0aecbd89d86846a/output.webp b/records/sd1/sweep/5efeb0aecbd89d86846a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..21d4bfb354bcf0ef5bc580d4e302037200cfbcea
--- /dev/null
+++ b/records/sd1/sweep/5efeb0aecbd89d86846a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef42954aa9059df8d4370ff56bd9429046f452a021dfbac423e6865a27284375
+size 26496
diff --git a/records/sd1/sweep/5f009e46a37ce44241e2/output.webp b/records/sd1/sweep/5f009e46a37ce44241e2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0f85ca84bb38766a6539b7a9c0de65527f5d6edf
--- /dev/null
+++ b/records/sd1/sweep/5f009e46a37ce44241e2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:217cdd69583388943ffdadd32706f9e5d9025f9c0f4d52a571d42bc4eaed3a2a
+size 19142
diff --git a/records/sd1/sweep/5f146ea0f814e815d636/output.webp b/records/sd1/sweep/5f146ea0f814e815d636/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..10b657d27df9dc92286a6a4ea3e2c471d4b56273
--- /dev/null
+++ b/records/sd1/sweep/5f146ea0f814e815d636/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:11ab24cc01aa8356f442303aace3b0dbc28014700eef84d4952fe63165e0f596
+size 30480
diff --git a/records/sd1/sweep/5f47614bf043b688d077/output.webp b/records/sd1/sweep/5f47614bf043b688d077/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..140cb78389c80a46b2db588757d1502bdb4d3e2b
--- /dev/null
+++ b/records/sd1/sweep/5f47614bf043b688d077/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7259f4a5cc35c2b5791b93057e2c0701096ade16252caf0f9a7f9d1671fdfee5
+size 46572
diff --git a/records/sd1/sweep/5f47c122bbd5db924008/output.webp b/records/sd1/sweep/5f47c122bbd5db924008/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9aebc8e50ae007d01cc0276841b732a0c4a0343c
--- /dev/null
+++ b/records/sd1/sweep/5f47c122bbd5db924008/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:613f63689adadcbb9080742dee23a08036e96459f3eefb244d92f269302db746
+size 39626
diff --git a/records/sd1/sweep/5f4ccbcff11606ab5ee9/output.webp b/records/sd1/sweep/5f4ccbcff11606ab5ee9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e862560ad19329a60a4dc3b9ba9a4e33604beef4
--- /dev/null
+++ b/records/sd1/sweep/5f4ccbcff11606ab5ee9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7860583c14be582686e55fcb1810fbf7a3f7aac287cf05f53dcbe00952a1f150
+size 32018
diff --git a/records/sd1/sweep/5f5e31ede2f9664a3be9/output.webp b/records/sd1/sweep/5f5e31ede2f9664a3be9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60f37c448310eb173455da40a0e8d0b1ba8b2296
--- /dev/null
+++ b/records/sd1/sweep/5f5e31ede2f9664a3be9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76020d4992b52195f05847c7853689ae60b19f81e3546bf834091ddc90bfea36
+size 19420
diff --git a/records/sd1/sweep/5f7cc6209276db09fb89/output.webp b/records/sd1/sweep/5f7cc6209276db09fb89/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4e26db54aa4f4a8af9dd92bc58d9a2b2d94884ee
--- /dev/null
+++ b/records/sd1/sweep/5f7cc6209276db09fb89/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a00c4a2de9a595de4ec7d66bac90028901751d01bddc7f97084050644664235d
+size 48728
diff --git a/records/sd1/sweep/5f86c5586cc6c41f35e8/output.webp b/records/sd1/sweep/5f86c5586cc6c41f35e8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7a1d66742eeee755fdea571ac646556ca6fc4681
--- /dev/null
+++ b/records/sd1/sweep/5f86c5586cc6c41f35e8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd402ec071cce4884a99e4e1d2a22fd4af51e7bdfe0ae273bf3d8f1301da9406
+size 28768
diff --git a/records/sd1/sweep/5fa47c6eace0e8dc89b3/output.webp b/records/sd1/sweep/5fa47c6eace0e8dc89b3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7f393a0b2cadc5b33c5b5a2f82c6fd5e8454a237
--- /dev/null
+++ b/records/sd1/sweep/5fa47c6eace0e8dc89b3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ffacd15d2bbc95333f4a99412fa8aa5b645b3fd2a978da9918019cd1285a71b
+size 37242
diff --git a/records/sd1/sweep/5fa4f7ab307153c74bcb/output.webp b/records/sd1/sweep/5fa4f7ab307153c74bcb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a2a9aff3b42bbc32f22825ac671563046b735749
--- /dev/null
+++ b/records/sd1/sweep/5fa4f7ab307153c74bcb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d088cbae1e49748603b32db978697a6f0a3aca085636d90ed3809b7e2727021c
+size 46738
diff --git a/records/sd1/sweep/5fa78fad48840116799c/output.webp b/records/sd1/sweep/5fa78fad48840116799c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0766efa65e75d22df0b72bbc89545c0769d23f38
--- /dev/null
+++ b/records/sd1/sweep/5fa78fad48840116799c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5cf5f8d0f70ae5141fae2746ca00453094fb8b8b8dbde2c40ac9bc4822192d82
+size 32876
diff --git a/records/sd1/sweep/5faa3e6be3e5a9a36a2b/output.webp b/records/sd1/sweep/5faa3e6be3e5a9a36a2b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..644a1c46b0f93b99d011fa1c5d05ef23db466c4f
--- /dev/null
+++ b/records/sd1/sweep/5faa3e6be3e5a9a36a2b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1363b9a7b2e980658ca199bc92809f06b76d20a74a7586f90ba74cd72afd394
+size 38872
diff --git a/records/sd1/sweep/5fab970fa7247bb68617/output.webp b/records/sd1/sweep/5fab970fa7247bb68617/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dd1c720a337fb6574105eabe54bbaa0cfbda0832
--- /dev/null
+++ b/records/sd1/sweep/5fab970fa7247bb68617/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c52250079069283893a286d5274da711edb125219984624d6eba623e52bd1d25
+size 37938
diff --git a/records/sd1/sweep/5fafa3e2af5f7668204d/output.webp b/records/sd1/sweep/5fafa3e2af5f7668204d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..02ffa625533659c3e1c5fdeb62a262fbd22693e5
--- /dev/null
+++ b/records/sd1/sweep/5fafa3e2af5f7668204d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e76fbbd16cf0121331e908c4ddf077429dee4b50042e20f3a6e1da3ba1033b4
+size 15774
diff --git a/records/sd1/sweep/5fb1a35e796a7a8d1981/output.webp b/records/sd1/sweep/5fb1a35e796a7a8d1981/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..064689a2d7c71e7dcbb43be7f14ca7bb3d45fca8
--- /dev/null
+++ b/records/sd1/sweep/5fb1a35e796a7a8d1981/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b08865318064fda3a9e2bf7bf2172b41de826152ad1ec7ff17df2e597bb4c04
+size 15732
diff --git a/records/sd1/sweep/5fd852b320be950eeaf2/output.webp b/records/sd1/sweep/5fd852b320be950eeaf2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6d14033cc589fabe8212b9d42d706a05eedefd57
--- /dev/null
+++ b/records/sd1/sweep/5fd852b320be950eeaf2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1325723b218bbee7dc6f85d129b7b6f7c7d7a73c538a3534d50444ff3f91fddc
+size 21722
diff --git a/records/sd1/sweep/5fe363aa9b75cd3499a2/output.webp b/records/sd1/sweep/5fe363aa9b75cd3499a2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4e54fdbfae8076bae35a8812c336cc59846d8df7
--- /dev/null
+++ b/records/sd1/sweep/5fe363aa9b75cd3499a2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4fe47e9b55c71cd077e57c30dca78bc19c97b91c8a9906942890e6b40dff593
+size 20852
diff --git a/records/sd1/sweep/5ff19eff847e8a359fbe/output.webp b/records/sd1/sweep/5ff19eff847e8a359fbe/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..56580b8dce9b00b279b31271a4b835cecbff27d9
--- /dev/null
+++ b/records/sd1/sweep/5ff19eff847e8a359fbe/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:09583df684c7374632ffa1a895d21618b2730d021f69c7c3013b9c844febad27
+size 20642
diff --git a/records/sd1/sweep/5ff8a939937ccce9265f/output.webp b/records/sd1/sweep/5ff8a939937ccce9265f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8131e6b3504480cec9f0bf8bd93af3cdaea2b721
--- /dev/null
+++ b/records/sd1/sweep/5ff8a939937ccce9265f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:032725a26eb5ac6cb0b226c097624e69705ee31b6702a1bbc79a3a0809ac0c38
+size 31136
diff --git a/records/sd1/sweep/600f13ae2e92ecb54deb/output.webp b/records/sd1/sweep/600f13ae2e92ecb54deb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ab04684fb40f29fbfd347448ba00cc8458641a2b
--- /dev/null
+++ b/records/sd1/sweep/600f13ae2e92ecb54deb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72d479f23feffc9db341b4cfac848e7eaaf638acb0335d8c3e03bf952785c286
+size 30338
diff --git a/records/sd1/sweep/60113b5af0b89072f1df/output.webp b/records/sd1/sweep/60113b5af0b89072f1df/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d181756089a17249cf7ce9cc2b064e8b8c438c92
--- /dev/null
+++ b/records/sd1/sweep/60113b5af0b89072f1df/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:91d2d1f9eaaa68742b26afb4237f6846e8bd9b11bcac936b8361493c6398192e
+size 24566
diff --git a/records/sd1/sweep/601f5a6aeebef6fead8e/output.webp b/records/sd1/sweep/601f5a6aeebef6fead8e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2344ed6cebffb266c9134a320d2d49e4892efbae
--- /dev/null
+++ b/records/sd1/sweep/601f5a6aeebef6fead8e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebee646386fab35d1d22145426ba10f5c945279f4b9b02c3603aaa44600a4070
+size 36262
diff --git a/records/sd1/sweep/602ae14ec6eab4d3b901/output.webp b/records/sd1/sweep/602ae14ec6eab4d3b901/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..03f2dc6b7b6a6ef56ca80fc417f292c13741b583
--- /dev/null
+++ b/records/sd1/sweep/602ae14ec6eab4d3b901/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29d4b5836ecd70fea30d2de0e87bdecf06764d72662ac9b1de991e4e35f82ca7
+size 36498
diff --git a/records/sd1/sweep/602c08207af6b8c07c24/output.webp b/records/sd1/sweep/602c08207af6b8c07c24/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..50fcb659a080266de4998ce404fc74f65555a6df
--- /dev/null
+++ b/records/sd1/sweep/602c08207af6b8c07c24/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb64d26e6b79ed4c26f2a739a074426e9673c6a1580dc1b94d4ff9f19d909d95
+size 34692
diff --git a/records/sd1/sweep/6034c2c137e21f0319fc/output.webp b/records/sd1/sweep/6034c2c137e21f0319fc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bedf6c10d46f89156d3776e689da5b16abc3c760
--- /dev/null
+++ b/records/sd1/sweep/6034c2c137e21f0319fc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36e1c8b36719eafedca2407f02d41845c0ae3776fc1ca6d5e9a17420da5aa5f4
+size 24626
diff --git a/records/sd1/sweep/6049c5c33b38726ea1bf/output.webp b/records/sd1/sweep/6049c5c33b38726ea1bf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4e7d2bea7512a6f19b91209a00a9509565d54585
--- /dev/null
+++ b/records/sd1/sweep/6049c5c33b38726ea1bf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:849b5ba1f45705ff67365e5e2e6852499b3acd55392d2c21b77c384ec1aa48b2
+size 42648
diff --git a/records/sd1/sweep/605600bde8f7941f713a/output.webp b/records/sd1/sweep/605600bde8f7941f713a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..21faa2c125d1d39a5099cb6a2edf957e737c2974
--- /dev/null
+++ b/records/sd1/sweep/605600bde8f7941f713a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d1b85d4a8cf1ed2710e9b35ce8b36cb5236c5162f362b8f2e1b745f9f70f19e
+size 36522
diff --git a/records/sd1/sweep/6061b801e9aeb07ecf76/output.webp b/records/sd1/sweep/6061b801e9aeb07ecf76/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3fa02cd5effe3b250e295416c3dd8ceb3f8ffc24
--- /dev/null
+++ b/records/sd1/sweep/6061b801e9aeb07ecf76/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa079f9f7e3fec3c24b29eb95025a2c52cd7d55f7122baa0fdc894972e128f8a
+size 27880
diff --git a/records/sd1/sweep/606b5a6b83655b7ab96a/output.webp b/records/sd1/sweep/606b5a6b83655b7ab96a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d94eac172536823d18bbb34cea33b1d9f527055e
--- /dev/null
+++ b/records/sd1/sweep/606b5a6b83655b7ab96a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1339544e99f88cdcf16c8c3c8512724181401fa1967c2418c174871f2eda265e
+size 18874
diff --git a/records/sd1/sweep/606dc63f27bec4e7056e/output.webp b/records/sd1/sweep/606dc63f27bec4e7056e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9a63f072c1275ee915222cc58c0ea0667dac9c2d
--- /dev/null
+++ b/records/sd1/sweep/606dc63f27bec4e7056e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0fea35ff4b6f632d6de697e9f4e0e0f25d6055a7b1200e952aaa8165ef31f7d5
+size 72476
diff --git a/records/sd1/sweep/607d538c40ffba399992/output.webp b/records/sd1/sweep/607d538c40ffba399992/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..90194fd99ee5b1bd913a2a3b401aa0d25351de98
--- /dev/null
+++ b/records/sd1/sweep/607d538c40ffba399992/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c6c16746b721897ecc0b0a6ee425e70adfd4e8253a14faf30dcf7b83116e928
+size 18706
diff --git a/records/sd1/sweep/608067885ab65c4c5937/output.webp b/records/sd1/sweep/608067885ab65c4c5937/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4f25a94c38137bc1bacdf593c935540c3702f519
--- /dev/null
+++ b/records/sd1/sweep/608067885ab65c4c5937/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:20be8a3bb87db5f1074de97c00147f7aa43f849854cb1ce1466e9a7f44d20012
+size 34314
diff --git a/records/sd1/sweep/6081cab8791e23c65521/output.webp b/records/sd1/sweep/6081cab8791e23c65521/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9e8fa75a14a6e177d4066007938525389988f7ca
--- /dev/null
+++ b/records/sd1/sweep/6081cab8791e23c65521/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:604703214f086d0c4ee7e28f70b6b30d5349785c14053e40ca46a02bd371149e
+size 27294
diff --git a/records/sd1/sweep/608e50eff55b49961244/output.webp b/records/sd1/sweep/608e50eff55b49961244/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1483a844871c05af267fc3f54c8ab4d5c47a5d6a
--- /dev/null
+++ b/records/sd1/sweep/608e50eff55b49961244/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bac5f8ae274ac97e113332d4722ab7f937095e2dbb53b2d06e2d879a2b94adb0
+size 13536
diff --git a/records/sd1/sweep/6093cf2ac9431739a09a/output.webp b/records/sd1/sweep/6093cf2ac9431739a09a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2fe7fd6dd5e15a6e2211fdec76ae992f41a256b0
--- /dev/null
+++ b/records/sd1/sweep/6093cf2ac9431739a09a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90934e661db5f36e1a1e5207f9dd518bbc5eee3a3e880ff6d32995e64426baf2
+size 40592
diff --git a/records/sd1/sweep/60af47cf5754588a3307/output.webp b/records/sd1/sweep/60af47cf5754588a3307/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d63538936505eabe31a2a0434403a1aeea11c8ac
--- /dev/null
+++ b/records/sd1/sweep/60af47cf5754588a3307/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bf3512a9899c4e3c3d28e48dcb5818ec542483c416ac6846ae1b2f1b91e9f6f
+size 20870
diff --git a/records/sd1/sweep/60b1296d6c3171d3e4d6/output.webp b/records/sd1/sweep/60b1296d6c3171d3e4d6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0d661dec6e4a82e965e10dd1d4c8d9a7ea21aee9
--- /dev/null
+++ b/records/sd1/sweep/60b1296d6c3171d3e4d6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0055c2a867e444173f8e926daf9661811ee5cdae64d60a33d25ec87363f164c0
+size 75808
diff --git a/records/sd1/sweep/60b3665d09e5caa87ad1/output.webp b/records/sd1/sweep/60b3665d09e5caa87ad1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6cb0d2143329f8f05386008f06c39c8fa6a0aae6
--- /dev/null
+++ b/records/sd1/sweep/60b3665d09e5caa87ad1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9304ad1d4de8f58495204ed963728cd5cc12fc0556d20e4fbccdf14392dadc38
+size 56426
diff --git a/records/sd1/sweep/60c3af26aeb012fdef4d/output.webp b/records/sd1/sweep/60c3af26aeb012fdef4d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6a9ca69b92d5a492196609c31c778f5e06d182c0
--- /dev/null
+++ b/records/sd1/sweep/60c3af26aeb012fdef4d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec4001231ce5a5b397613da823719ae9f3b17f1b95cedc5d91522e7b1885ca5c
+size 44510
diff --git a/records/sd1/sweep/60cda0346e54d8562b9e/output.webp b/records/sd1/sweep/60cda0346e54d8562b9e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..523e77ac8fc63fe6a5a9e70a2ef27e8573783100
--- /dev/null
+++ b/records/sd1/sweep/60cda0346e54d8562b9e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d0f2c665b013f51d8e8d578bf909c56d2a9f35ea78a4a214cb362786a2688a4
+size 20590
diff --git a/records/sd1/sweep/60ce612dc1856f642e4f/output.webp b/records/sd1/sweep/60ce612dc1856f642e4f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..45b53dc4bfe9de94649d106cf95ed980297b4135
--- /dev/null
+++ b/records/sd1/sweep/60ce612dc1856f642e4f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3dea214fd720567f9dbfd7abf3866dc3a66a22928f07910311d8898d802a243
+size 28600
diff --git a/records/sd1/sweep/60d14700364509e6a525/output.webp b/records/sd1/sweep/60d14700364509e6a525/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c5c104c563733ec0701473b902092d11668d9484
--- /dev/null
+++ b/records/sd1/sweep/60d14700364509e6a525/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:887c52f451a9e298266d8583ceb35ede18bd1a5f933818a72a8df78544fa1bef
+size 22366
diff --git a/records/sd1/sweep/60d2adba07909dba1166/output.webp b/records/sd1/sweep/60d2adba07909dba1166/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9d6b58f74ce1d7edb6e7a2f8e0464513a98f9e2d
--- /dev/null
+++ b/records/sd1/sweep/60d2adba07909dba1166/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f3bb629b9110475e6adf943d937c1704124e37e2f2338392e09e9142a9534200
+size 30492
diff --git a/records/sd1/sweep/60dd25efa4a3ac0f12df/output.webp b/records/sd1/sweep/60dd25efa4a3ac0f12df/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d549d1f021b4be40f32eb37a4765ef7d0c686c5e
--- /dev/null
+++ b/records/sd1/sweep/60dd25efa4a3ac0f12df/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a3eb996f94bcc2ee17da720c612a2cb8c1dfc1ed206b8a0a09b6d201c015f35
+size 36496
diff --git a/records/sd1/sweep/60ea1f258091f2a03313/output.webp b/records/sd1/sweep/60ea1f258091f2a03313/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1528507c55043612757409ad7e6d12c7be2587e8
--- /dev/null
+++ b/records/sd1/sweep/60ea1f258091f2a03313/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10d54c08a2ea1010c9548a148a803348b8f5991fdbf8b7a71233810410a10815
+size 16406
diff --git a/records/sd1/sweep/60ed4468a69ae08865af/output.webp b/records/sd1/sweep/60ed4468a69ae08865af/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c757e3c9f1896c564584b756ac2fdcb0abe6af39
--- /dev/null
+++ b/records/sd1/sweep/60ed4468a69ae08865af/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f6a6f121f6c2c71d5b2c49ffd4aad3ab08f40313713a38833b50a8793fa19e9
+size 12444
diff --git a/records/sd1/sweep/60eefb255d934d9920aa/output.webp b/records/sd1/sweep/60eefb255d934d9920aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2d9a2e07e7b04d0794c69ef7d197ebd3d4413ec8
--- /dev/null
+++ b/records/sd1/sweep/60eefb255d934d9920aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:420214c6bcd040124f9ce732eb581b5f30eac53e3296af58d80783517618b4c8
+size 7786
diff --git a/records/sd1/sweep/60f33568ef9376f0322a/output.webp b/records/sd1/sweep/60f33568ef9376f0322a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4327af1137eacec243a01313fc978c4e74fffe45
--- /dev/null
+++ b/records/sd1/sweep/60f33568ef9376f0322a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddef606fb2d896599ee7b8377c7d1aba8ca218094b9daedb381fda9d335d03cc
+size 32444
diff --git a/records/sd1/sweep/60fbc128218cbd48f96c/output.webp b/records/sd1/sweep/60fbc128218cbd48f96c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..53bdffc5b85c570d78c94d4390449fda1fa8ae0a
--- /dev/null
+++ b/records/sd1/sweep/60fbc128218cbd48f96c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ed85174c7ecae3fe094c60f64185db078ee24fd3285bc2d6adc5c12dba6b6118
+size 53790
diff --git a/records/sd1/sweep/60fc45bfa46795301431/output.webp b/records/sd1/sweep/60fc45bfa46795301431/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1e50b1ddb47f72854fc446c3608ce5039ae6f894
--- /dev/null
+++ b/records/sd1/sweep/60fc45bfa46795301431/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:26ff00b829465486e28bed79a2632cacb830a79253d21e8384fcb56fd9e51610
+size 40618
diff --git a/records/sd1/sweep/60ff67ff3debd4dd450c/output.webp b/records/sd1/sweep/60ff67ff3debd4dd450c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fd049f51c8d1dac6f503c83eabd83d3cee77f2ba
--- /dev/null
+++ b/records/sd1/sweep/60ff67ff3debd4dd450c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e6b371053030414925267a194706edec8cd469a276b0d03e9b346a688ea9a3f0
+size 36498
diff --git a/records/sd1/sweep/61085603b3b74eae3b33/output.webp b/records/sd1/sweep/61085603b3b74eae3b33/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ff8beba4ca17c6d1e6a17ec47de87c9ebdcf6c5c
--- /dev/null
+++ b/records/sd1/sweep/61085603b3b74eae3b33/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80d737c235cf700ce0bbf966628904efbca3970f7dc03e9dd0cb204f8657d08f
+size 32222
diff --git a/records/sd1/sweep/61138032bcd0af65ef09/output.webp b/records/sd1/sweep/61138032bcd0af65ef09/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0da6047a992536efe311574cbf349e5b25de2d95
--- /dev/null
+++ b/records/sd1/sweep/61138032bcd0af65ef09/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b14f95f4ea85a344c841dadb4a7d9f799b0b6f7370f5c41e8e05787a4fa36a31
+size 19994
diff --git a/records/sd1/sweep/611ce13589a5924d61aa/output.webp b/records/sd1/sweep/611ce13589a5924d61aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0af4c5762451df17aca6c7d43239615ae2cb9555
--- /dev/null
+++ b/records/sd1/sweep/611ce13589a5924d61aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:99a2ddc783babd6bbd9f0f63a9d8800922032d68ab54140ac78a8780023ef310
+size 17206
diff --git a/records/sd1/sweep/612192a6201531ebac0a/output.webp b/records/sd1/sweep/612192a6201531ebac0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9f9ea4adeac066b5aea664dac82e2817222c1b70
--- /dev/null
+++ b/records/sd1/sweep/612192a6201531ebac0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9842289f7a616d770cca529f4829ba01be413d845225b25829707f5eba7ba57e
+size 40876
diff --git a/records/sd1/sweep/61222dcce98b32083deb/output.webp b/records/sd1/sweep/61222dcce98b32083deb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dc846a3c1e6199956eb929ab8cbf985842f4b67c
--- /dev/null
+++ b/records/sd1/sweep/61222dcce98b32083deb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a4a0ed7050b318e31bbb5fbfc90ad93967c7e85e4690230f6d51556d5f2f467
+size 34580
diff --git a/records/sd1/sweep/612999db4e45da987e0b/output.webp b/records/sd1/sweep/612999db4e45da987e0b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..76865bcbc029ba6550d9ea8325db78eb35558cc7
--- /dev/null
+++ b/records/sd1/sweep/612999db4e45da987e0b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b617f4ac460a7832bb76134ca016482f8c9875ff4cc7455557b326306958adb
+size 51392
diff --git a/records/sd1/sweep/612fe5cd6706ec15daa7/output.webp b/records/sd1/sweep/612fe5cd6706ec15daa7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..861e115d31223439b879efae44067999a4821140
--- /dev/null
+++ b/records/sd1/sweep/612fe5cd6706ec15daa7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:930f5547d3bd1d09a5bd95753f75513d228c77c06489ddf5eaca9b04234bad19
+size 98342
diff --git a/records/sd1/sweep/61376b7a37371a9d8088/output.webp b/records/sd1/sweep/61376b7a37371a9d8088/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ee44b5af20d781742fc264698bcbfb73ad8c0ea6
--- /dev/null
+++ b/records/sd1/sweep/61376b7a37371a9d8088/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:724b096076afd0e0a1885fe4a565fbb98195f7fc45fa4ad301959b3264fd1fe9
+size 31836
diff --git a/records/sd1/sweep/613f53c493ddc593f014/output.webp b/records/sd1/sweep/613f53c493ddc593f014/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..64d8f0b40aac458051e4c974a211d4df6ca2d083
--- /dev/null
+++ b/records/sd1/sweep/613f53c493ddc593f014/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71fb69629489127d271605e6d31e82da2c6baee85412c2d324b7fe43df27a3de
+size 18596
diff --git a/records/sd1/sweep/6141ec8633121cc0bcb8/output.webp b/records/sd1/sweep/6141ec8633121cc0bcb8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e983714900efdf31f023135c5dff27183ddbfd3b
--- /dev/null
+++ b/records/sd1/sweep/6141ec8633121cc0bcb8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d39f6f9f116629d46d038cd94acc22c27f09a98101b2498c1b305cfe1a809577
+size 35248
diff --git a/records/sd1/sweep/614446789daf6c7b17bc/output.webp b/records/sd1/sweep/614446789daf6c7b17bc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2248f4db20eacb5d1e11f768bd7d7923c661893d
--- /dev/null
+++ b/records/sd1/sweep/614446789daf6c7b17bc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b7648a60d2505f1a44d6ec2b2d08f009b1a9406a8f7a5db1ef0c22b6b109545
+size 31072
diff --git a/records/sd1/sweep/61470ad0bbc840117ea7/output.webp b/records/sd1/sweep/61470ad0bbc840117ea7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..09eba027ca2f02e77bed03d32b3f272305e2abd3
--- /dev/null
+++ b/records/sd1/sweep/61470ad0bbc840117ea7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6c7c1955ea929d00e73261f39818eed22592652d9cb39500ac0f9d224c045cb3
+size 24160
diff --git a/records/sd1/sweep/614cc4051684d12ab9a1/output.webp b/records/sd1/sweep/614cc4051684d12ab9a1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c117552991f1d9ed5a82fe9aa2fcc5e6f9ad3e49
--- /dev/null
+++ b/records/sd1/sweep/614cc4051684d12ab9a1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1535d3f00c6f34ff8cbea613fab9a5c2d7b1ad3ae3b6e71c5f678096379bc8d8
+size 37212
diff --git a/records/sd1/sweep/6154288c816d3d710e32/output.webp b/records/sd1/sweep/6154288c816d3d710e32/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a84339949f315b021ce84e00ea4b2de6ea63c2f3
--- /dev/null
+++ b/records/sd1/sweep/6154288c816d3d710e32/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5e6e08318d670e3b94db3751ff063b47a915f229d2750bffa30d3da4def4ad8
+size 26568
diff --git a/records/sd1/sweep/6154f07db2cbc2f187e1/output.webp b/records/sd1/sweep/6154f07db2cbc2f187e1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f41fa425ddec22d3134ccb49cebd7b3e89f72417
--- /dev/null
+++ b/records/sd1/sweep/6154f07db2cbc2f187e1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:885f455b009b6f79fda4710c2674b8bab292c2ca20f8343ccde3c8022657bf69
+size 27484
diff --git a/records/sd1/sweep/6155094013278d2fbc62/output.webp b/records/sd1/sweep/6155094013278d2fbc62/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3bbdc6c45096f80b108952a5644223099396ff1a
--- /dev/null
+++ b/records/sd1/sweep/6155094013278d2fbc62/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:faec0f22cdc5950d879bfd25374cead350015a4db134f409967b26696643fd3c
+size 36170
diff --git a/records/sd1/sweep/616a47b6b5d0841f6156/output.webp b/records/sd1/sweep/616a47b6b5d0841f6156/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..44bf462745c300d589eb3acc508400c747508d25
--- /dev/null
+++ b/records/sd1/sweep/616a47b6b5d0841f6156/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4a5ff2eea737cd736df9758d202a233cb97a034c700bca91299430149555dc0
+size 20486
diff --git a/records/sd1/sweep/616df0a0cc15d227e1da/output.webp b/records/sd1/sweep/616df0a0cc15d227e1da/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d5af42a90fe3b5728516163cb85a2e6f2619bd1c
--- /dev/null
+++ b/records/sd1/sweep/616df0a0cc15d227e1da/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa1cfdccdd20ee2f61a7473dd86e4eee9f86ffc8ac70b6617a05c71ffc5f1687
+size 15444
diff --git a/records/sd1/sweep/6176be5ed5a267c4c388/output.webp b/records/sd1/sweep/6176be5ed5a267c4c388/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7c79f5a00aec842fdb1db80cf968655740636617
--- /dev/null
+++ b/records/sd1/sweep/6176be5ed5a267c4c388/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:355d3ed883fe0db974cb95b5461a95ad2a0e942b53be8bc24bad851cba3889a7
+size 47090
diff --git a/records/sd1/sweep/61786cc3534a96a1d0ea/output.webp b/records/sd1/sweep/61786cc3534a96a1d0ea/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3eb77a7f7fa1550df062f08aa0a4ed4fac8e9450
--- /dev/null
+++ b/records/sd1/sweep/61786cc3534a96a1d0ea/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d3f1bfd050e0763e7dbdbb41f23055c5d9c47eac64c07aa33414ba5daf371af
+size 19728
diff --git a/records/sd1/sweep/617b2023b641a3904cb8/output.webp b/records/sd1/sweep/617b2023b641a3904cb8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..962c2cd49028350c81db63e3c5d4802849d170d8
--- /dev/null
+++ b/records/sd1/sweep/617b2023b641a3904cb8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a97667410197a4341980147a0e8a922b1d84430bb6cca236a941285e4d92d4c
+size 28536
diff --git a/records/sd1/sweep/6182c4c3b84d9c86261c/output.webp b/records/sd1/sweep/6182c4c3b84d9c86261c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..88c915607bec2ab498cdd33ba71a55e5a5264c10
--- /dev/null
+++ b/records/sd1/sweep/6182c4c3b84d9c86261c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6ccd115fa1e52d462922b960cc0ab5833bee06fa752d42309270b3768a6a77e
+size 18324
diff --git a/records/sd1/sweep/6189ba33257131cae4c4/output.webp b/records/sd1/sweep/6189ba33257131cae4c4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9cfeb979c201192aad86b5d174482a327164e6a2
--- /dev/null
+++ b/records/sd1/sweep/6189ba33257131cae4c4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3be07d5983d856a5b900f738e61246daf5fca9a6cf9d626a431c89be58ec79a7
+size 37976
diff --git a/records/sd1/sweep/61921c6c00bc9aba8f95/output.webp b/records/sd1/sweep/61921c6c00bc9aba8f95/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e18290146c2b3649fd3a237c40e5a7a3ca2de42e
--- /dev/null
+++ b/records/sd1/sweep/61921c6c00bc9aba8f95/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a9c3375410769386464f28f20b5b01b0f4fc974a4802f95e80869d960c9aa962
+size 36504
diff --git a/records/sd1/sweep/61948521b3f0531494a3/output.webp b/records/sd1/sweep/61948521b3f0531494a3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2da5e3f0657197d2944e8de36f5c770967f0f784
--- /dev/null
+++ b/records/sd1/sweep/61948521b3f0531494a3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b49f9d5bbae386200ce3ce4b00a9a865701b59c2dcdbcb601c62a08db116f946
+size 58194
diff --git a/records/sd1/sweep/619f8bba3e20da2325ce/output.webp b/records/sd1/sweep/619f8bba3e20da2325ce/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7f33afacefc88b1c6dc463a65035e3250a49240f
--- /dev/null
+++ b/records/sd1/sweep/619f8bba3e20da2325ce/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b6f323eb9d6da1330a8469c41b4da0f51e22c2c6276e1b049ac765c08593c82
+size 56592
diff --git a/records/sd1/sweep/61a38ae6af3202d4d570/output.webp b/records/sd1/sweep/61a38ae6af3202d4d570/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c758b15632bab843a00369dd27c3d824a032bc21
--- /dev/null
+++ b/records/sd1/sweep/61a38ae6af3202d4d570/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a830d725351f134047a22f7e01907f05c4eb467ef820b77de7910ea1a25ab88
+size 98770
diff --git a/records/sd1/sweep/61a68ffa9adc69025a9f/output.webp b/records/sd1/sweep/61a68ffa9adc69025a9f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dbee04b88ce0105bd91e6cf0cdfa624c8fd456e3
--- /dev/null
+++ b/records/sd1/sweep/61a68ffa9adc69025a9f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b56678e755c33c91becef349c1b65b8d360f91dc9341fd9f276a4f4ad2f963d
+size 53122
diff --git a/records/sd1/sweep/61a8aedf0164eda2b074/output.webp b/records/sd1/sweep/61a8aedf0164eda2b074/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..580f45f22fc04d2806e71f79903db1a59c87868d
--- /dev/null
+++ b/records/sd1/sweep/61a8aedf0164eda2b074/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9546394df1cfde4047b94e9cfe9aa64d783ad7855426d103ad01fb038a103963
+size 32518
diff --git a/records/sd1/sweep/61c99de71740ae5cb1a8/output.webp b/records/sd1/sweep/61c99de71740ae5cb1a8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..54f058c14076a1306551a6b9e064f3948f623b7c
--- /dev/null
+++ b/records/sd1/sweep/61c99de71740ae5cb1a8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81a917c8b88c3bf66b5dd26ac6d882b7bbd95735d96e389b2e581ad2a719b5b6
+size 22432
diff --git a/records/sd1/sweep/61d6eece6fd7022fbf20/output.webp b/records/sd1/sweep/61d6eece6fd7022fbf20/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1de2eb4138f370feeb1ba9cc850107733774cf81
--- /dev/null
+++ b/records/sd1/sweep/61d6eece6fd7022fbf20/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9af9f3da788d88fdc4414a0852159b5b9ed31f0b9222f8c3613f2a293161b76d
+size 17192
diff --git a/records/sd1/sweep/61e088de6e9095d55c5d/output.webp b/records/sd1/sweep/61e088de6e9095d55c5d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..418c36d424c1b0af05ca5595a3dd4694566ba002
--- /dev/null
+++ b/records/sd1/sweep/61e088de6e9095d55c5d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98935cf52d962c79229741a727c3196d5f05ff2603ce96f32665c99f24c1e228
+size 61286
diff --git a/records/sd1/sweep/61f0826ccff7af942f31/output.webp b/records/sd1/sweep/61f0826ccff7af942f31/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d207b386530d1ae4b9947a168b28e1a2cc95f0f4
--- /dev/null
+++ b/records/sd1/sweep/61f0826ccff7af942f31/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83215bc6389bd1ce12bb877b0a0dcfc01b9dddbd2dd0d320ca40a898959136a1
+size 34106
diff --git a/records/sd1/sweep/61f38e086ee3f065aa0b/output.webp b/records/sd1/sweep/61f38e086ee3f065aa0b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b5503f421ca7ea8b0fa3d6f80c475ca67ef343fd
--- /dev/null
+++ b/records/sd1/sweep/61f38e086ee3f065aa0b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f90d1ab882a2557f8be131138ff183e0ce68a0a01544d7f308890b051107cf67
+size 31116
diff --git a/records/sd1/sweep/61f88e287fe9fad3d0fd/output.webp b/records/sd1/sweep/61f88e287fe9fad3d0fd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..393fa670fc38703c07c6c6165f45e4f38b04b414
--- /dev/null
+++ b/records/sd1/sweep/61f88e287fe9fad3d0fd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3b369a797f73b96562f778b3f21a35013b5346fa163eed6db57499c1d9f6b28
+size 26496
diff --git a/records/sd1/sweep/61ff3e3396c911fa32d4/output.webp b/records/sd1/sweep/61ff3e3396c911fa32d4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..884767b30173f5d5d2b03f8c9b07ba9d065d4b69
--- /dev/null
+++ b/records/sd1/sweep/61ff3e3396c911fa32d4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:752d3d9d4a55e17f5568fd1b752b1205e2d96cd17128846e62bc73fcd04e2edf
+size 57120
diff --git a/records/sd1/sweep/620d5e5123ec4f4ca134/output.webp b/records/sd1/sweep/620d5e5123ec4f4ca134/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9384035515d23724624e051c18677a6e90c84103
--- /dev/null
+++ b/records/sd1/sweep/620d5e5123ec4f4ca134/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5aeec087c7c3642156096094a5a699a2dfad789d5fc181336018467761347001
+size 51610
diff --git a/records/sd1/sweep/6215a71b68f13e85826b/output.webp b/records/sd1/sweep/6215a71b68f13e85826b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..98be257830315fb9a98de3de53d3c19c03cfa4bb
--- /dev/null
+++ b/records/sd1/sweep/6215a71b68f13e85826b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:529372589d54a563eb2dab9b127545b1fc7a34123bc5b24db7454e4647d40d2d
+size 42050
diff --git a/records/sd1/sweep/6216eac76a811a996c44/output.webp b/records/sd1/sweep/6216eac76a811a996c44/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9e699146108a0ad0dfdce9f6639d936b50da6171
--- /dev/null
+++ b/records/sd1/sweep/6216eac76a811a996c44/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25f7b2597be6b242b17963033609767a213f24ed941445d880eced0a928eca4d
+size 19930
diff --git a/records/sd1/sweep/6219a15832d6eef0d39a/output.webp b/records/sd1/sweep/6219a15832d6eef0d39a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d4dd1e59bd254c6ca2b72f835541d7c79a525412
--- /dev/null
+++ b/records/sd1/sweep/6219a15832d6eef0d39a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:889e1c6e45557c94128045135c17c5da4adebb096f74f5c0f70b02159d396c11
+size 32708
diff --git a/records/sd1/sweep/621c217a1c40c75e3395/output.webp b/records/sd1/sweep/621c217a1c40c75e3395/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8405b38387dfb83af13dab8f5e9227d46c5d39f7
--- /dev/null
+++ b/records/sd1/sweep/621c217a1c40c75e3395/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4b4c803d8815d4864300931ceab994ce83e1c7cfc7d55162a0127422d544ed4
+size 44042
diff --git a/records/sd1/sweep/621d7f547779a45dc6cc/output.webp b/records/sd1/sweep/621d7f547779a45dc6cc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0b7ae134821373b9412764fa3ebc35c690dfd711
--- /dev/null
+++ b/records/sd1/sweep/621d7f547779a45dc6cc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ce4eaf8952daa8791cd37e0bd45aaa111032b77b2c0dd789af0994c0f50b49c
+size 21262
diff --git a/records/sd1/sweep/623174b4f8d9376c5e3a/output.webp b/records/sd1/sweep/623174b4f8d9376c5e3a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a949fa3b8646e2dd772f264d8f1912991eee6f43
--- /dev/null
+++ b/records/sd1/sweep/623174b4f8d9376c5e3a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:781b4323113b90624e1c535a71c15a63d576075c3fd962045fb27ce6380721ed
+size 15002
diff --git a/records/sd1/sweep/6231a4f70c31eed8c732/output.webp b/records/sd1/sweep/6231a4f70c31eed8c732/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..62e019b933932b1b464192de2dc060a8e89f02f6
--- /dev/null
+++ b/records/sd1/sweep/6231a4f70c31eed8c732/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56bf1e184313775e125ad4f865df23cb0c6c9d4f685ca966b0d4ca01326f0135
+size 62132
diff --git a/records/sd1/sweep/6245be818e8fb9902fda/output.webp b/records/sd1/sweep/6245be818e8fb9902fda/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d0eac9e408d59f776c2aa6c10ddf0a4dffbaa4bb
--- /dev/null
+++ b/records/sd1/sweep/6245be818e8fb9902fda/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e92bd14cabac7c801f12e66e61bf2cbe91269ba6ebf169e2a4acee551052144
+size 25454
diff --git a/records/sd1/sweep/6249edc6a1ca80b03820/output.webp b/records/sd1/sweep/6249edc6a1ca80b03820/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5e3e675afb84b9424bf1fe21c435ccd1de4d6d3f
--- /dev/null
+++ b/records/sd1/sweep/6249edc6a1ca80b03820/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37df79fad3aa1f5e701305e7b28628fb078dbe51aba64d56216f5fd604db164e
+size 42518
diff --git a/records/sd1/sweep/624e51d917a437cac429/output.webp b/records/sd1/sweep/624e51d917a437cac429/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f2ad69a52c6d0826989d15420268e6313ec101c4
--- /dev/null
+++ b/records/sd1/sweep/624e51d917a437cac429/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d87beda9185169e985729ea3d867eff1437d9b68a5fa993983db6c7a1e45f13
+size 19936
diff --git a/records/sd1/sweep/62587aaa72324b5235ea/output.webp b/records/sd1/sweep/62587aaa72324b5235ea/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1ec99ff1b44b525df3879dcf5e0b3c698149e17d
--- /dev/null
+++ b/records/sd1/sweep/62587aaa72324b5235ea/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cab6826143e0a275457c564df20060ce8ffe97effa9f6610a6b45abcc5a9b249
+size 15874
diff --git a/records/sd1/sweep/6258fd6185d008fd60ec/output.webp b/records/sd1/sweep/6258fd6185d008fd60ec/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fbaf9a5786e1c7e41151dd30fd503ef17c857c85
--- /dev/null
+++ b/records/sd1/sweep/6258fd6185d008fd60ec/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:212a5b31e4956b11f07d4285a4c51aff6008c975648388f9bed2d1ab9701a3a4
+size 26680
diff --git a/records/sd1/sweep/625a43c793ec3e9b8d9c/output.webp b/records/sd1/sweep/625a43c793ec3e9b8d9c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..46dc469f82877d524b928aa094b420b1ca95eb37
--- /dev/null
+++ b/records/sd1/sweep/625a43c793ec3e9b8d9c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76598822573b8a3ea1f7c6eb934ffa6fd31d71cfcb4e1b563daa8ef8bdecc260
+size 33364
diff --git a/records/sd1/sweep/62616a5b74ff4e880df9/output.webp b/records/sd1/sweep/62616a5b74ff4e880df9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3bdb858edf3114f238daa13405a2020f94e982fa
--- /dev/null
+++ b/records/sd1/sweep/62616a5b74ff4e880df9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0531f33574391c9319ae3545a4fd2e53e849f4a642d8f4d2a76a9fceb1cb8234
+size 22548
diff --git a/records/sd1/sweep/626cc4450d5a11b939d7/output.webp b/records/sd1/sweep/626cc4450d5a11b939d7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60f095fe092268ccddfa856676dc45c82afafbda
--- /dev/null
+++ b/records/sd1/sweep/626cc4450d5a11b939d7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d92ae7f02805e578a5c0ab00ef00ff8f598476289664c9a996c7069b3626ba3
+size 20390
diff --git a/records/sd1/sweep/6271c186e7dce0dc8efd/output.webp b/records/sd1/sweep/6271c186e7dce0dc8efd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..68efe3be4188e860046fad65c8103def72dd0923
--- /dev/null
+++ b/records/sd1/sweep/6271c186e7dce0dc8efd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac2f58f97b6b2ae2b42db54d0b9da609037bd8e16687bd6996ecb963f9334cee
+size 19934
diff --git a/records/sd1/sweep/6272b9136d96cc963cdd/output.webp b/records/sd1/sweep/6272b9136d96cc963cdd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0a20b8ed65a49ad9271c58dd582fa00df7ba8196
--- /dev/null
+++ b/records/sd1/sweep/6272b9136d96cc963cdd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4f5da443c56d3054d6f743e993918d35d7f2b9ecc98d40618da22dc6c37fc86
+size 19176
diff --git a/records/sd1/sweep/6275f7b5016c918f6c52/output.webp b/records/sd1/sweep/6275f7b5016c918f6c52/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8285def3109921d39a5bdc17cbad6c5b765166f5
--- /dev/null
+++ b/records/sd1/sweep/6275f7b5016c918f6c52/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a75f5f45377504fa3495d94c4f3f3e3f98419856380bb7d7e92df31601f80db2
+size 47784
diff --git a/records/sd1/sweep/627a66437b21f2e2e6ae/output.webp b/records/sd1/sweep/627a66437b21f2e2e6ae/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e5f2000e53825b4abd9a0aafac70b71ec6ff700d
--- /dev/null
+++ b/records/sd1/sweep/627a66437b21f2e2e6ae/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:79ce6490ad1fb2465b187a53d733d5851082e445a02de1bda2134e523b5e84d8
+size 32862
diff --git a/records/sd1/sweep/62859dc65a081f1b109e/output.webp b/records/sd1/sweep/62859dc65a081f1b109e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7bd1985d2056c5d7a76d5c6aa3a47643ac5ce8b8
--- /dev/null
+++ b/records/sd1/sweep/62859dc65a081f1b109e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b2b1ebb9d7f9495956907d0e9abc5da7f92d4b47af3b41be409143e08cc43b7
+size 24912
diff --git a/records/sd1/sweep/6294fa62c1951d9df5d4/output.webp b/records/sd1/sweep/6294fa62c1951d9df5d4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..63495dca3c301d2eac5150eb403969fa22d48b78
--- /dev/null
+++ b/records/sd1/sweep/6294fa62c1951d9df5d4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f6f294b63da403ad60bd76a7cb3462744a5958c4e1e0985c6d9745a90a772b74
+size 18786
diff --git a/records/sd1/sweep/629d24ec160fb7ac12e9/output.webp b/records/sd1/sweep/629d24ec160fb7ac12e9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2f9535aab161018594dc9af11b88ce11a68fbd09
--- /dev/null
+++ b/records/sd1/sweep/629d24ec160fb7ac12e9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3d60c3b3c5b0d0d2117298dea9f1c0f46837d46471c138f76830282716edd2c0
+size 47566
diff --git a/records/sd1/sweep/62a339c7b1d88a423859/output.webp b/records/sd1/sweep/62a339c7b1d88a423859/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..036ba1e5d202083bd9ed6de70e41ece90173f1eb
--- /dev/null
+++ b/records/sd1/sweep/62a339c7b1d88a423859/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8dbf840a0d764ecc64f3f0d729922f8b41f87a4279a0c036ff50d3919d5c631f
+size 22508
diff --git a/records/sd1/sweep/62a8ca9576aee7b52c78/output.webp b/records/sd1/sweep/62a8ca9576aee7b52c78/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..96b49a227b7358bb4161ef83169987a1be8934d9
--- /dev/null
+++ b/records/sd1/sweep/62a8ca9576aee7b52c78/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77a2b8e0647567155601711c3b8495d77123d5875453aa6dfe2f2acca1bea4b1
+size 19930
diff --git a/records/sd1/sweep/62acd1563105331760ca/output.webp b/records/sd1/sweep/62acd1563105331760ca/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1d7342e47f3e53d17709116a1bd83a76bae0558a
--- /dev/null
+++ b/records/sd1/sweep/62acd1563105331760ca/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eeb995611e8b63e7e6158008a953b283c2f2a6a12a3c739d9caf378d38696042
+size 36928
diff --git a/records/sd1/sweep/62ad7d6cf162ab909f4a/output.webp b/records/sd1/sweep/62ad7d6cf162ab909f4a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..954c667774275399b88b641f0eae63a1ebb8ef8f
--- /dev/null
+++ b/records/sd1/sweep/62ad7d6cf162ab909f4a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcbd48278d066ebd24178a01f7a48c47adf7b2b4ad0c0ba2e8ff7de36cb25e85
+size 18538
diff --git a/records/sd1/sweep/62af59016e131ca7334b/output.webp b/records/sd1/sweep/62af59016e131ca7334b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1bb4b4cfd8bc2a1870064a46de840e7eecd5bc89
--- /dev/null
+++ b/records/sd1/sweep/62af59016e131ca7334b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:559f9dc234948bdab695ae636efbdab34f0517cb3b67e3bb06796de57b59941c
+size 37136
diff --git a/records/sd1/sweep/62b844fce457ead32a13/output.webp b/records/sd1/sweep/62b844fce457ead32a13/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..30367cf47952dd1a9b6c7775e48d394d6e6ee193
--- /dev/null
+++ b/records/sd1/sweep/62b844fce457ead32a13/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b70387d42cf2542ce2abdfaae391e92c683175783c4d9406786d7c275df5e3e
+size 36664
diff --git a/records/sd1/sweep/62c3181f7e84f151528c/output.webp b/records/sd1/sweep/62c3181f7e84f151528c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..199f7149909e414398ae0b75bcbca4ae1ff00ef3
--- /dev/null
+++ b/records/sd1/sweep/62c3181f7e84f151528c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:400b5b5ea1ca2cfa22c3b37b71a7c584044c7b7daa4e9b50800b43ee759f3c04
+size 51492
diff --git a/records/sd1/sweep/62db9511a2ed94f7b3c6/output.webp b/records/sd1/sweep/62db9511a2ed94f7b3c6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6df83f3d4cb94bca1345d0261ddec9d59f4026d9
--- /dev/null
+++ b/records/sd1/sweep/62db9511a2ed94f7b3c6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:07b523ff734b407e55743dd32d2c88949016ebbaefb167e68d8c794b30ef6b35
+size 36516
diff --git a/records/sd1/sweep/62ec513df3fd75a5ab0a/output.webp b/records/sd1/sweep/62ec513df3fd75a5ab0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..00b1fde98a053d86110389eb5761bc58d1f97bed
--- /dev/null
+++ b/records/sd1/sweep/62ec513df3fd75a5ab0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4e1c285e0f98ea2fceca296850de5711f45ed5ff6ba53b13b1cc29b464a0290
+size 52346
diff --git a/records/sd1/sweep/62eed29732d1b82417a5/output.webp b/records/sd1/sweep/62eed29732d1b82417a5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0937fd2e2902f109f5a079d8dea1fdc0d6dcc606
--- /dev/null
+++ b/records/sd1/sweep/62eed29732d1b82417a5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:228f708ced6936100dfc506d3ce2c29a35f61b90161456950010e8bfa99380b6
+size 32438
diff --git a/records/sd1/sweep/630e47d2a6480293c305/output.webp b/records/sd1/sweep/630e47d2a6480293c305/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d4ac10fdf221142ab30fa4d3b7908e04372c4fbd
--- /dev/null
+++ b/records/sd1/sweep/630e47d2a6480293c305/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3c27235a02efb624ef2efc74e7b385e72ec13a6f3dae1130156ccaf0cd02f00
+size 67602
diff --git a/records/sd1/sweep/6313424020d570aea0c8/output.webp b/records/sd1/sweep/6313424020d570aea0c8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..07c8a6c478275aaf7b512d7e0527317eacf3ed1a
--- /dev/null
+++ b/records/sd1/sweep/6313424020d570aea0c8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d50a684a8ae2a78bb32866fdb0802c7180ce8edb21f52fc1d0c167e13b1350dd
+size 33028
diff --git a/records/sd1/sweep/63189329bf69d69fb7f6/output.webp b/records/sd1/sweep/63189329bf69d69fb7f6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..735e5a295b11194206141ffcceb2b1583bed2af4
--- /dev/null
+++ b/records/sd1/sweep/63189329bf69d69fb7f6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4968a6bf7cc0c02ab563de44aabaac0eb70a79a7a4b245f4c928f92b24314ae5
+size 17064
diff --git a/records/sd1/sweep/632b8fcd07c1b8ce5cef/output.webp b/records/sd1/sweep/632b8fcd07c1b8ce5cef/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e0089d7a9969f795bfa89aa1b4682cde956406e5
--- /dev/null
+++ b/records/sd1/sweep/632b8fcd07c1b8ce5cef/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17765d26a3db78a1280f7e473c6c6464522cb4e90de5343bd4e23657b2416944
+size 29602
diff --git a/records/sd1/sweep/632ff1f509ccba8f9a14/output.webp b/records/sd1/sweep/632ff1f509ccba8f9a14/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1efce2af85e941f72c63e5ccd46fa8591d1ad171
--- /dev/null
+++ b/records/sd1/sweep/632ff1f509ccba8f9a14/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b05a77137aef4cf9d9541fa69fddae6a52583eca0f7d4255a3d4c20c4a15c618
+size 18480
diff --git a/records/sd1/sweep/6cac9a5c7126b241e383/record.json b/records/sd1/sweep/6cac9a5c7126b241e383/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..28b907ce7cac1be6b45204e1f36e892221f5b8a6
--- /dev/null
+++ b/records/sd1/sweep/6cac9a5c7126b241e383/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6cac9a5c7126b241e383",
+ "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.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/0e6c5e1c5497b8fc3349f94f08409d73bf8b4b193b9d36fc8fda52ea4464f341.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0e6c5e1c5497b8fc3349f94f08409d73bf8b4b193b9d36fc8fda52ea4464f341",
+  "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": "55b5d0965dd796a06e0b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cac9a5c7126b241e383/workflow.json b/records/sd1/sweep/6cac9a5c7126b241e383/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..10ac1c8dc151e8e7a2b39eabd6e78403031907f4
--- /dev/null
+++ b/records/sd1/sweep/6cac9a5c7126b241e383/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 6cac9a5c7126b241e383)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.h_upd\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cca646b3465cb5d49f9/interpretations.jsonl b/records/sd1/sweep/6cca646b3465cb5d49f9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2de4ffd4e0c265e9a6d65f39ed00a21824e64d34
--- /dev/null
+++ b/records/sd1/sweep/6cca646b3465cb5d49f9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6cca646b3465cb5d49f9"}, "kind": "caption", "value": "Painterly woman with dark wavy bob, blue-grey eyes, grey brushed background with rusty edges", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2ec1e102c5f73341"}
+{"target": {"record": "6cca646b3465cb5d49f9"}, "kind": "change", "value": "The portrait becomes a looser oil-painting style with darker, fuller hair, a slightly tilted gaze and a lighter grey scumbled background with rust-stained edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2ace650a7bddd0aa"}
+{"target": {"record": "6cca646b3465cb5d49f9"}, "kind": "keywords", "value": ["oil portrait", "dark wavy hair", "scumbled background", "rusty edges", "painterly"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2b4a12b84432c775"}
+{"target": {"record": "6cca646b3465cb5d49f9"}, "kind": "effect_tags", "value": ["painterly", "texture", "subject_kept", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:13:59Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d0038e8d99821c25"}
diff --git a/records/sd1/sweep/6cca646b3465cb5d49f9/measurements.json b/records/sd1/sweep/6cca646b3465cb5d49f9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..47853a467cfa0bf5d260804c4efc1e49709aec03
--- /dev/null
+++ b/records/sd1/sweep/6cca646b3465cb5d49f9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6cca646b3465cb5d49f9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.248728,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.304053,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.096307,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.046274,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.552,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01251,
+   "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": 33.127,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.939,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0942,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cca646b3465cb5d49f9/record.json b/records/sd1/sweep/6cca646b3465cb5d49f9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..80900f83f90da80f75af0760c68e4855df0b420c
--- /dev/null
+++ b/records/sd1/sweep/6cca646b3465cb5d49f9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6cca646b3465cb5d49f9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:50Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "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/a5cf26daec90e6c6b9f048a3258f0f043ec8b332afcc334e7cb7a6e668814c3b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a5cf26daec90e6c6b9f048a3258f0f043ec8b332afcc334e7cb7a6e668814c3b",
+  "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": "610fcd44f550b9cc09ea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cca646b3465cb5d49f9/workflow.json b/records/sd1/sweep/6cca646b3465cb5d49f9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..354ff7bf6b3d9b17184b7c91dc2c39d5beb51586
--- /dev/null
+++ b/records/sd1/sweep/6cca646b3465cb5d49f9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6cca646b3465cb5d49f9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.transformer_blocks.0.ff.net.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/6ccac331b6d4d5ab18b9/measurements.json b/records/sd1/sweep/6ccac331b6d4d5ab18b9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d244d871393f6926987aff2832af6f317f381a42
--- /dev/null
+++ b/records/sd1/sweep/6ccac331b6d4d5ab18b9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ccac331b6d4d5ab18b9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.331755,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.381572,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.439404,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.158528,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.722,
+   "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": 38.506,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9999,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0887,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ccac331b6d4d5ab18b9/record.json b/records/sd1/sweep/6ccac331b6d4d5ab18b9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d57177be30e461fc6806ea20bcbeee92662f5ef
--- /dev/null
+++ b/records/sd1/sweep/6ccac331b6d4d5ab18b9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6ccac331b6d4d5ab18b9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.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/a1997c5dd0ff8e843b31abe0eaa078d2493450b38813f5f05dbb5d8c7269d144.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a1997c5dd0ff8e843b31abe0eaa078d2493450b38813f5f05dbb5d8c7269d144",
+  "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": "d5571bb6dc2bf83c049d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ccac331b6d4d5ab18b9/workflow.json b/records/sd1/sweep/6ccac331b6d4d5ab18b9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ef2ba890e27cc890ae3fff998c5c2fc61c2ca03
--- /dev/null
+++ b/records/sd1/sweep/6ccac331b6d4d5ab18b9/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 6ccac331b6d4d5ab18b9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.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/6ccc5b5ee4f23011f28b/measurements.json b/records/sd1/sweep/6ccc5b5ee4f23011f28b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..348c4b02332c89da0b1bb5ce386961d0666983b8
--- /dev/null
+++ b/records/sd1/sweep/6ccc5b5ee4f23011f28b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ccc5b5ee4f23011f28b",
+ "values": {
+  "std": {
+   "value": 29.969,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02929,
+   "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": 6.199,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.016505,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.052864,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.058885,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008671,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.3521,
+   "method": "CLIP zero-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/6ccc5b5ee4f23011f28b/record.json b/records/sd1/sweep/6ccc5b5ee4f23011f28b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c191fc7318882765d6b823cf94413d1fea6bb713
--- /dev/null
+++ b/records/sd1/sweep/6ccc5b5ee4f23011f28b/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6ccc5b5ee4f23011f28b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:01:58Z",
+ "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 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": "1bbda6ca1f7efc631dfb",
+  "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": "27638e9211d1fc245727"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ccc5b5ee4f23011f28b/workflow.json b/records/sd1/sweep/6ccc5b5ee4f23011f28b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..191b8f55f5c4affd1ac890e270b158d7617d629b
--- /dev/null
+++ b/records/sd1/sweep/6ccc5b5ee4f23011f28b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ccc5b5ee4f23011f28b)"
+  },
+  "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": 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/6ccca1b8a1ff258e439a/measurements.json b/records/sd1/sweep/6ccca1b8a1ff258e439a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f16a4a8b7d5fae725452adc907624bf81585793
--- /dev/null
+++ b/records/sd1/sweep/6ccca1b8a1ff258e439a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6ccca1b8a1ff258e439a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.028525,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.047624,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.011468,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.009613,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.878,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00528,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.258,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0125,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2063,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ccca1b8a1ff258e439a/record.json b/records/sd1/sweep/6ccca1b8a1ff258e439a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2612563fd6c6fc6ac0523c7c1713cd4e3b2c42a3
--- /dev/null
+++ b/records/sd1/sweep/6ccca1b8a1ff258e439a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6ccca1b8a1ff258e439a",
+ "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": 1.5
+   },
+   "group": "out.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/b6042c339e9e9a6ea8a9c07be6769b3dfe88b8fefe5eab087ad10b38efada7eb.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b6042c339e9e9a6ea8a9c07be6769b3dfe88b8fefe5eab087ad10b38efada7eb",
+  "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": "f8fdcab2eb94fc942f1c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ccca1b8a1ff258e439a/workflow.json b/records/sd1/sweep/6ccca1b8a1ff258e439a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9053e93397ce0586783f914b28da57de8774796
--- /dev/null
+++ b/records/sd1/sweep/6ccca1b8a1ff258e439a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ccca1b8a1ff258e439a)"
+  },
+  "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\": 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/6cde3664debcb0a1f9c6/measurements.json b/records/sd1/sweep/6cde3664debcb0a1f9c6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..79b27df2c321d355b5d616748ac961c894c128f7
--- /dev/null
+++ b/records/sd1/sweep/6cde3664debcb0a1f9c6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6cde3664debcb0a1f9c6",
+ "values": {
+  "std": {
+   "value": 28.693,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02058,
+   "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": 31.519,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.33798,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.578008,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.559958,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.252572,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0262,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0153,
+   "method": "CLIP zero-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/6cde3664debcb0a1f9c6/record.json b/records/sd1/sweep/6cde3664debcb0a1f9c6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3225962a17abe77972e65e1f89bcb4df31ca87a
--- /dev/null
+++ b/records/sd1/sweep/6cde3664debcb0a1f9c6/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6cde3664debcb0a1f9c6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:57:16Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "slight",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   30.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "16a21c75606e807f571a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cde3664debcb0a1f9c6/workflow.json b/records/sd1/sweep/6cde3664debcb0a1f9c6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..62452a9804bfb2eaca5b8ececbb31dba7415f6a6
--- /dev/null
+++ b/records/sd1/sweep/6cde3664debcb0a1f9c6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6cde3664debcb0a1f9c6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"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/6ce79139a2fe4f021701/measurements.json b/records/sd1/sweep/6ce79139a2fe4f021701/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cce939a96a5d71c72156cbc13437477bc14cc0ec
--- /dev/null
+++ b/records/sd1/sweep/6ce79139a2fe4f021701/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ce79139a2fe4f021701",
+ "values": {
+  "std": {
+   "value": 37.733,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01469,
+   "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": 2.191,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.003577,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.011385,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.007702,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001423,
+   "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": 0.0134,
+   "method": "CLIP zero-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/6ce79139a2fe4f021701/record.json b/records/sd1/sweep/6ce79139a2fe4f021701/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c66c8c0d27a217337533bb81ed5f7548982f0757
--- /dev/null
+++ b/records/sd1/sweep/6ce79139a2fe4f021701/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6ce79139a2fe4f021701",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:06:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2640d0bad4926cceb0de"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ce79139a2fe4f021701/workflow.json b/records/sd1/sweep/6ce79139a2fe4f021701/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d8b403e41f097afc3e4b0a0a2385d2f64ec45e2
--- /dev/null
+++ b/records/sd1/sweep/6ce79139a2fe4f021701/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ce79139a2fe4f021701)"
+  },
+  "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\": 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/6ced0f95f4e3f2c0fbba/measurements.json b/records/sd1/sweep/6ced0f95f4e3f2c0fbba/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f77cfdb2d1cd7168c017cbe7df5fd5f4583effa9
--- /dev/null
+++ b/records/sd1/sweep/6ced0f95f4e3f2c0fbba/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ced0f95f4e3f2c0fbba",
+ "values": {
+  "cosine_distance": {
+   "value": 0.200012,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.356021,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.240205,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.1087,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.539,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00609,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.187,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9732,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0245,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ced0f95f4e3f2c0fbba/record.json b/records/sd1/sweep/6ced0f95f4e3f2c0fbba/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f60c947adbbcdb7db7212a7e3c35c161880cea4f
--- /dev/null
+++ b/records/sd1/sweep/6ced0f95f4e3f2c0fbba/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6ced0f95f4e3f2c0fbba",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:27Z",
+ "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.8.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/953d9c6b8f94a61a8dfd7522f46c9ae1f1d1cb8043a7019f17798325351feefc.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "953d9c6b8f94a61a8dfd7522f46c9ae1f1d1cb8043a7019f17798325351feefc",
+  "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": "1bd250cb7f975cae9590"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ced0f95f4e3f2c0fbba/workflow.json b/records/sd1/sweep/6ced0f95f4e3f2c0fbba/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e075f6bcb60509d2e95ede63ed4c74543d2ecf3
--- /dev/null
+++ b/records/sd1/sweep/6ced0f95f4e3f2c0fbba/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ced0f95f4e3f2c0fbba)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.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/6cedff8639793560a4bf/measurements.json b/records/sd1/sweep/6cedff8639793560a4bf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6015bc6d0731c876a1a251fd0df2aa1727a6126b
--- /dev/null
+++ b/records/sd1/sweep/6cedff8639793560a4bf/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6cedff8639793560a4bf",
+ "values": {
+  "std": {
+   "value": 41.351,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01507,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 58.688,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.598703,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.432756,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.140529,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.087596,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.02,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1123,
+   "method": "CLIP zero-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/6cedff8639793560a4bf/record.json b/records/sd1/sweep/6cedff8639793560a4bf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2eed595d373667fc053ef03e4cd756ca5873d77d
--- /dev/null
+++ b/records/sd1/sweep/6cedff8639793560a4bf/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6cedff8639793560a4bf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:28:51Z",
+ "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": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.mid",
+   "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": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   180.0,
+   "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": "c8b7a5e66fb87f8f6e33"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cedff8639793560a4bf/workflow.json b/records/sd1/sweep/6cedff8639793560a4bf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..07bbaa6138bd3447823d4a142f76354bd2fabb6a
--- /dev/null
+++ b/records/sd1/sweep/6cedff8639793560a4bf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6cedff8639793560a4bf)"
+  },
+  "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\": 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/6cf06d1a05122ad97c1e/interpretations.jsonl b/records/sd1/sweep/6cf06d1a05122ad97c1e/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7bee027d80e25e14e28889adf09aefe16bdbd42c
--- /dev/null
+++ b/records/sd1/sweep/6cf06d1a05122ad97c1e/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6cf06d1a05122ad97c1e"}, "kind": "caption", "value": "Warm-skinned woman with a dark tousled bob and blue eyes on a light grey-blue background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "36203b77152b3e9b"}
+{"target": {"record": "6cf06d1a05122ad97c1e"}, "kind": "change", "value": "The face turns warmer and rounder with a dark, tousled bob, the eyes become bluer, and the background lightens to a cool grey-blue.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "489d90e62ddba865"}
+{"target": {"record": "6cf06d1a05122ad97c1e"}, "kind": "keywords", "value": ["warm skin", "dark bob", "blue eyes", "light background", "rounder face"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6a30a9631f4ac400"}
+{"target": {"record": "6cf06d1a05122ad97c1e"}, "kind": "effect_tags", "value": ["subject_kept", "colour_shift", "painterly"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5c1abcfe8b0a8379"}
diff --git a/records/sd1/sweep/6cf06d1a05122ad97c1e/measurements.json b/records/sd1/sweep/6cf06d1a05122ad97c1e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..37441af964202f79d147f03cd6de7f64812ff971
--- /dev/null
+++ b/records/sd1/sweep/6cf06d1a05122ad97c1e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6cf06d1a05122ad97c1e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.311598,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.360718,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.240138,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.079811,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.528,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00475,
+   "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": 35.409,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9714,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0427,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cf06d1a05122ad97c1e/record.json b/records/sd1/sweep/6cf06d1a05122ad97c1e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1643317d3fb74fe4d7c7c6c4fdae5b44681c707e
--- /dev/null
+++ b/records/sd1/sweep/6cf06d1a05122ad97c1e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6cf06d1a05122ad97c1e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:12Z",
+ "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.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/c183e0996537e4018591c54c96083aa3879c36002028f4a0d92dad91eaa1e96e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c183e0996537e4018591c54c96083aa3879c36002028f4a0d92dad91eaa1e96e",
+  "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": "f0af41bdd715ac5ddd04"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cf06d1a05122ad97c1e/workflow.json b/records/sd1/sweep/6cf06d1a05122ad97c1e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..be7b0e017a3fea09f1099e6a58e034bc636e97ab
--- /dev/null
+++ b/records/sd1/sweep/6cf06d1a05122ad97c1e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6cf06d1a05122ad97c1e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.proj_out\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cffbc055b1ec2ffb383/measurements.json b/records/sd1/sweep/6cffbc055b1ec2ffb383/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c883ba914ee800e34d7275442cf9e399b88150df
--- /dev/null
+++ b/records/sd1/sweep/6cffbc055b1ec2ffb383/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6cffbc055b1ec2ffb383",
+ "values": {
+  "cosine_distance": {
+   "value": 0.145575,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.199417,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.1452,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.047837,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.638,
+   "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": 26.163,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.044,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0481,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cffbc055b1ec2ffb383/record.json b/records/sd1/sweep/6cffbc055b1ec2ffb383/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f343aff475fa103eba6efd880dedd40ba7fccd2a
--- /dev/null
+++ b/records/sd1/sweep/6cffbc055b1ec2ffb383/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6cffbc055b1ec2ffb383",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:56Z",
+ "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": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/275e2faef5841b314461809955f8c3a904dec19b89b6ead3858bad743e4967d9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "275e2faef5841b314461809955f8c3a904dec19b89b6ead3858bad743e4967d9",
+  "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": "0823e605b0b2862e7cfb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6cffbc055b1ec2ffb383/workflow.json b/records/sd1/sweep/6cffbc055b1ec2ffb383/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b02e50ea3586c7eb9fd8a2c5a9b39e47918ea155
--- /dev/null
+++ b/records/sd1/sweep/6cffbc055b1ec2ffb383/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6cffbc055b1ec2ffb383)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.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": [
+    "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/6d10d2134b22b3232477/measurements.json b/records/sd1/sweep/6d10d2134b22b3232477/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe0bf2c351b78f27d3b75e088ee4311b68e22b98
--- /dev/null
+++ b/records/sd1/sweep/6d10d2134b22b3232477/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d10d2134b22b3232477",
+ "values": {
+  "std": {
+   "value": 58.125,
+   "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"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.784,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.082538,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.201733,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.293512,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.109762,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9589,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.387,
+   "method": "CLIP zero-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/6d10d2134b22b3232477/record.json b/records/sd1/sweep/6d10d2134b22b3232477/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0da3684061d9d17fd107f821ace136a02c5c6d88
--- /dev/null
+++ b/records/sd1/sweep/6d10d2134b22b3232477/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "6d10d2134b22b3232477",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:41:28Z",
+ "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": 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": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "zero",
+   "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": [
+   "out.lo",
+   "multiply",
+   0.0,
+   "late"
+  ],
+  "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": "2b75d85c99b3393ab023"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d10d2134b22b3232477/workflow.json b/records/sd1/sweep/6d10d2134b22b3232477/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..023ed11cafe06611c395c6cdb6123b2c9a2c9a52
--- /dev/null
+++ b/records/sd1/sweep/6d10d2134b22b3232477/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d10d2134b22b3232477)"
+  },
+  "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\": 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/6d17ecb1a9cb97ecbe9b/interpretations.jsonl b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ffe4d45b4e334afea5730ed0281170d600b032b5
--- /dev/null
+++ b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6d17ecb1a9cb97ecbe9b"}, "kind": "caption", "value": "Smaller painted portrait of a woman inside a dark panel framed by a wide tan border", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e74a77f7509b2d75"}
+{"target": {"record": "6d17ecb1a9cb97ecbe9b"}, "kind": "change", "value": "The figure is smaller and set inside a dark rectangular panel with a wider tan paper border, so it reads like a framed picture. The face is softer and younger.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "528815216210029a"}
+{"target": {"record": "6d17ecb1a9cb97ecbe9b"}, "kind": "keywords", "value": ["framed portrait", "wide border", "smaller figure", "dark panel"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "909f16d8fc57733b"}
+{"target": {"record": "6d17ecb1a9cb97ecbe9b"}, "kind": "effect_tags", "value": ["framing", "subject_kept", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:16:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e9605da8e11db76e"}
diff --git a/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/measurements.json b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ad6079979555f9f0fc4fc414b064d7f84a696bf
--- /dev/null
+++ b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d17ecb1a9cb97ecbe9b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.11253,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.203992,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.05313,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.038083,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.07,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00589,
+   "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": 18.753,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.968,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1544,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/record.json b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1527d5da7037becb70958c86f9f09489c14a821c
--- /dev/null
+++ b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d17ecb1a9cb97ecbe9b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:41Z",
+ "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": 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.9.0.op",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.lo",
+   "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/6718721c5852267645e5658cdf7c54112ade693ff25849466abc0a417ae0ae49.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6718721c5852267645e5658cdf7c54112ade693ff25849466abc0a417ae0ae49",
+  "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": "a66b51b37a70f3f79aa9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/workflow.json b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6317c59ed61587f1f76ef2152a10dd20ae0edb00
--- /dev/null
+++ b/records/sd1/sweep/6d17ecb1a9cb97ecbe9b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d17ecb1a9cb97ecbe9b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.9.0.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/6d1911be314ecad1a5fe/measurements.json b/records/sd1/sweep/6d1911be314ecad1a5fe/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..74985aca10ec0b0803d1ed3b8307ffd235ae8669
--- /dev/null
+++ b/records/sd1/sweep/6d1911be314ecad1a5fe/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d1911be314ecad1a5fe",
+ "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/6d1911be314ecad1a5fe/record.json b/records/sd1/sweep/6d1911be314ecad1a5fe/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ac4c870f0d2db10c91a0d33549255f7c5d2f74b
--- /dev/null
+++ b/records/sd1/sweep/6d1911be314ecad1a5fe/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d1911be314ecad1a5fe",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:35Z",
+ "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.11.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/99d6c80120ca7114f97be52dd12706be4ffeb6ecd812bfb34c43b241168ed07a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "99d6c80120ca7114f97be52dd12706be4ffeb6ecd812bfb34c43b241168ed07a",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "a1ea1d7bf1e7b94a0590"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d1911be314ecad1a5fe/workflow.json b/records/sd1/sweep/6d1911be314ecad1a5fe/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c41dfddf62911a048d8fe1eb36d497e1b07d84a
--- /dev/null
+++ b/records/sd1/sweep/6d1911be314ecad1a5fe/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 6d1911be314ecad1a5fe)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/6d1b240b9362a6001ebd/interpretations.jsonl b/records/sd1/sweep/6d1b240b9362a6001ebd/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a5db3b2171499c7d167c09e5c55724ac294c6ffe
--- /dev/null
+++ b/records/sd1/sweep/6d1b240b9362a6001ebd/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6d1b240b9362a6001ebd"}, "kind": "caption", "value": "Dark smoky frame around a burst of colourful neon fragments in the center", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5291465848f9ae15"}
+{"target": {"record": "6d1b240b9362a6001ebd"}, "kind": "change", "value": "The street is swallowed by a black, smoky vignette that surrounds a chaotic, glitchy cluster of blue, orange and white neon fragments, so the image becomes dramatic and abstract.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f1a7f7a8eca11609"}
+{"target": {"record": "6d1b240b9362a6001ebd"}, "kind": "keywords", "value": ["smoky vignette", "neon burst", "dark abstraction", "explosive center", "city lost"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4fb6e97d94d9994d"}
+{"target": {"record": "6d1b240b9362a6001ebd"}, "kind": "effect_tags", "value": ["abstract", "dark", "framing", "subject_lost", "high_contrast", "surreal"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "910fa777f6ffcc3a"}
diff --git a/records/sd1/sweep/6d1b240b9362a6001ebd/measurements.json b/records/sd1/sweep/6d1b240b9362a6001ebd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c27cad8e74a129b8f13d78a1f6a13b2f53dd7f28
--- /dev/null
+++ b/records/sd1/sweep/6d1b240b9362a6001ebd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d1b240b9362a6001ebd",
+ "values": {
+  "std": {
+   "value": 43.655,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00594,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.18,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.401024,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.628925,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.981879,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.329726,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6921,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4811,
+   "method": "CLIP zero-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/6d1b240b9362a6001ebd/record.json b/records/sd1/sweep/6d1b240b9362a6001ebd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fdac9da073dfc1259a869e3a716664825870df91
--- /dev/null
+++ b/records/sd1/sweep/6d1b240b9362a6001ebd/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "6d1b240b9362a6001ebd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:14:00Z",
+ "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": 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": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "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",
+   "scale",
+   0.7,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.7,
+    "applied": 0.7,
+    "reason": "outside safe range [0.95, 1.05] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c2ba0e8ed9783617ae6e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d1b240b9362a6001ebd/workflow.json b/records/sd1/sweep/6d1b240b9362a6001ebd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5272a517379e54abd008a2a83de65b15796f91a
--- /dev/null
+++ b/records/sd1/sweep/6d1b240b9362a6001ebd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d1b240b9362a6001ebd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\n   }\n  }\n ]\n}",
+   "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/6d22cb26885710fefd77/measurements.json b/records/sd1/sweep/6d22cb26885710fefd77/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f67308aeb891d8180cce4b60baf0690f445307e9
--- /dev/null
+++ b/records/sd1/sweep/6d22cb26885710fefd77/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d22cb26885710fefd77",
+ "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/6d22cb26885710fefd77/record.json b/records/sd1/sweep/6d22cb26885710fefd77/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4a691789c439ec363e36e73f6aba4fbfb9d0545
--- /dev/null
+++ b/records/sd1/sweep/6d22cb26885710fefd77/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d22cb26885710fefd77",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:27Z",
+ "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": "input_blocks.5.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/dba4bb36fadc59abf6b57d78a16d13b09b2d079c60a9c7a7afddd70e4eec1a7d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "dba4bb36fadc59abf6b57d78a16d13b09b2d079c60a9c7a7afddd70e4eec1a7d",
+  "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": "39364bb71f5e875b7626"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d22cb26885710fefd77/workflow.json b/records/sd1/sweep/6d22cb26885710fefd77/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a7a86b3ac559b0c053a26bd42f65d8aa3b2c32d
--- /dev/null
+++ b/records/sd1/sweep/6d22cb26885710fefd77/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d22cb26885710fefd77)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d260d8cc61dbe392f63/measurements.json b/records/sd1/sweep/6d260d8cc61dbe392f63/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..66c1b0521cd57683ce608815bad6afcb561a053f
--- /dev/null
+++ b/records/sd1/sweep/6d260d8cc61dbe392f63/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d260d8cc61dbe392f63",
+ "values": {
+  "std": {
+   "value": 42.379,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0067,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.782,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.393748,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.654355,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.760554,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.293984,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0208,
+   "method": "CLIP zero-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/6d260d8cc61dbe392f63/record.json b/records/sd1/sweep/6d260d8cc61dbe392f63/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..04b3dd46f481c4fa6ff33347aeb9b2723b30615f
--- /dev/null
+++ b/records/sd1/sweep/6d260d8cc61dbe392f63/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6d260d8cc61dbe392f63",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:33:42Z",
+ "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": "input_blocks.5.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "rotate",
+   30.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4535e11c7be7952eac93"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d260d8cc61dbe392f63/workflow.json b/records/sd1/sweep/6d260d8cc61dbe392f63/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5302eb844dd80c41668221f1b460337e5f45006e
--- /dev/null
+++ b/records/sd1/sweep/6d260d8cc61dbe392f63/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d260d8cc61dbe392f63)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ]\n}",
+   "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/6d26534af981b6360ce8/measurements.json b/records/sd1/sweep/6d26534af981b6360ce8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc084c19fbe02e2578badc57830248527efc6024
--- /dev/null
+++ b/records/sd1/sweep/6d26534af981b6360ce8/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d26534af981b6360ce8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.163487,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.184372,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.214465,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.200766,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.902,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00508,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.686,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9038,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7172,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d26534af981b6360ce8/record.json b/records/sd1/sweep/6d26534af981b6360ce8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7231321488e6506c8e70506f54995ac593bf6daf
--- /dev/null
+++ b/records/sd1/sweep/6d26534af981b6360ce8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d26534af981b6360ce8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:27Z",
+ "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": "output_blocks.5.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "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/3be12da57fa7f98df890d2ea683894d4a7dfba4a11f5a87c6ecbf48d7ae16b3e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3be12da57fa7f98df890d2ea683894d4a7dfba4a11f5a87c6ecbf48d7ae16b3e",
+  "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": "bfe9986d4ad9584debb7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d26534af981b6360ce8/workflow.json b/records/sd1/sweep/6d26534af981b6360ce8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9494905bb14f3c42116143fee76567a1e0c75630
--- /dev/null
+++ b/records/sd1/sweep/6d26534af981b6360ce8/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 6d26534af981b6360ce8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.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": [
+    "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/6d4450edef416e993662/interpretations.jsonl b/records/sd1/sweep/6d4450edef416e993662/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5a1088831c222d6ac19da7d7f7854faf8f6ecd96
--- /dev/null
+++ b/records/sd1/sweep/6d4450edef416e993662/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6d4450edef416e993662"}, "kind": "caption", "value": "Sharp sepia portrait of a slicked-hair young person on a beige, rounded vintage card", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8875675bb2529c3f"}
+{"target": {"record": "6d4450edef416e993662"}, "kind": "change", "value": "The subject becomes a different young person with slicked hair and a stronger jaw, in warmer beige-sepia tones on a rounded vintage card.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7d798a14ed1e30a8"}
+{"target": {"record": "6d4450edef416e993662"}, "kind": "keywords", "value": ["sepia portrait", "slicked hair", "vintage card", "beige tones", "new face"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "45af246af4e36993"}
+{"target": {"record": "6d4450edef416e993662"}, "kind": "effect_tags", "value": ["colour_shift", "subject_kept", "framing", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "86a069f004fa6cf0"}
diff --git a/records/sd1/sweep/6d4450edef416e993662/measurements.json b/records/sd1/sweep/6d4450edef416e993662/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..63b3a8c4532766472d0d69956df53c88660b0236
--- /dev/null
+++ b/records/sd1/sweep/6d4450edef416e993662/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d4450edef416e993662",
+ "values": {
+  "cosine_distance": {
+   "value": 0.395525,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.427931,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.228057,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.188296,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.424,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00369,
+   "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": 44.853,
+   "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": 0.1751,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d4450edef416e993662/record.json b/records/sd1/sweep/6d4450edef416e993662/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c41d10f04a73d6b1526da2b9ec56f6734e404a1
--- /dev/null
+++ b/records/sd1/sweep/6d4450edef416e993662/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d4450edef416e993662",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:13Z",
+ "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.2.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "skip",
+   "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/458aabff03840543905a4226a5d6d24dd9d1e5f7684bd72a63507e7d1385cf62.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "458aabff03840543905a4226a5d6d24dd9d1e5f7684bd72a63507e7d1385cf62",
+  "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": "d9cc6f8b8a8b0177dd21"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d4450edef416e993662/workflow.json b/records/sd1/sweep/6d4450edef416e993662/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ad689be5c054624fc166fdf9990ab5954e454f0
--- /dev/null
+++ b/records/sd1/sweep/6d4450edef416e993662/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 6d4450edef416e993662)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.skip_connection\",\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/6d52d965b20bd46f7c9e/measurements.json b/records/sd1/sweep/6d52d965b20bd46f7c9e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fde06facf87fa891713a0e414b7a859784c99b9f
--- /dev/null
+++ b/records/sd1/sweep/6d52d965b20bd46f7c9e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d52d965b20bd46f7c9e",
+ "values": {
+  "std": {
+   "value": 44.198,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01671,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 1.908,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001453,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.007575,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004551,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001165,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9884,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0645,
+   "method": "CLIP zero-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/6d52d965b20bd46f7c9e/record.json b/records/sd1/sweep/6d52d965b20bd46f7c9e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..14224b0a0ae29aada62787dbf4bc1688d8ac126a
--- /dev/null
+++ b/records/sd1/sweep/6d52d965b20bd46f7c9e/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6d52d965b20bd46f7c9e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:53:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "9cff7ccefd3537cd38b9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d52d965b20bd46f7c9e/workflow.json b/records/sd1/sweep/6d52d965b20bd46f7c9e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..89eaeffac5a70b802d3a82928a94f46ecac3d7b7
--- /dev/null
+++ b/records/sd1/sweep/6d52d965b20bd46f7c9e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d52d965b20bd46f7c9e)"
+  },
+  "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\": 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/6d5363f88a6c3e6d311b/measurements.json b/records/sd1/sweep/6d5363f88a6c3e6d311b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f02dadc8b07e43af4cebb1a666221843c75f7de5
--- /dev/null
+++ b/records/sd1/sweep/6d5363f88a6c3e6d311b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d5363f88a6c3e6d311b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.327723,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.365493,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.173349,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.076931,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.097,
+   "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": 35.446,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.0897,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d5363f88a6c3e6d311b/record.json b/records/sd1/sweep/6d5363f88a6c3e6d311b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..268e93451354aa15f8e2309f1cc48c1a4a1328ce
--- /dev/null
+++ b/records/sd1/sweep/6d5363f88a6c3e6d311b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d5363f88a6c3e6d311b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:59Z",
+ "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": "input_blocks.4.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.mid",
+   "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/d12a557dd987a7df635d916f03464eceb356961b15e148a56b67f524192a7583.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d12a557dd987a7df635d916f03464eceb356961b15e148a56b67f524192a7583",
+  "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": "0e048bcfd58909e7018c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d5363f88a6c3e6d311b/workflow.json b/records/sd1/sweep/6d5363f88a6c3e6d311b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..12ae92a2eed6e2eb696f4ad9e287b0f9d25f9f27
--- /dev/null
+++ b/records/sd1/sweep/6d5363f88a6c3e6d311b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d5363f88a6c3e6d311b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d5a137d13d3060f0bc9/measurements.json b/records/sd1/sweep/6d5a137d13d3060f0bc9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..04fd2c1335c5922c5058e46f724828759b1e8bb4
--- /dev/null
+++ b/records/sd1/sweep/6d5a137d13d3060f0bc9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d5a137d13d3060f0bc9",
+ "values": {
+  "std": {
+   "value": 33.982,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01295,
+   "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": 36.714,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.358262,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.534781,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.239472,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0478,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9482,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4253,
+   "method": "CLIP zero-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/6d5a137d13d3060f0bc9/record.json b/records/sd1/sweep/6d5a137d13d3060f0bc9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb279f0b065ffabaa61b3c0a27deff9b20492f5f
--- /dev/null
+++ b/records/sd1/sweep/6d5a137d13d3060f0bc9/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6d5a137d13d3060f0bc9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:19:25Z",
+ "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 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": 1.4
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "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",
+   "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": "7c6fda9e42e57f5cfae8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d5a137d13d3060f0bc9/workflow.json b/records/sd1/sweep/6d5a137d13d3060f0bc9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbbc0d004ca5a7a6646fb84d3e9810685581638e
--- /dev/null
+++ b/records/sd1/sweep/6d5a137d13d3060f0bc9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d5a137d13d3060f0bc9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/6d5fcd9315b487bb5736/measurements.json b/records/sd1/sweep/6d5fcd9315b487bb5736/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..93894d2f669bb6bdb622c56b6375e3c0bc335e93
--- /dev/null
+++ b/records/sd1/sweep/6d5fcd9315b487bb5736/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d5fcd9315b487bb5736",
+ "values": {
+  "cosine_distance": {
+   "value": 0.674649,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.73211,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.931039,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.495495,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 94.158,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01498,
+   "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": 88.382,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8257,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8816,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d5fcd9315b487bb5736/record.json b/records/sd1/sweep/6d5fcd9315b487bb5736/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5428c83275c453655a3ab659cab12ffb7e03fc56
--- /dev/null
+++ b/records/sd1/sweep/6d5fcd9315b487bb5736/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d5fcd9315b487bb5736",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:23Z",
+ "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.10.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/4065d5fe4068450ea6dc4aefbadd2dc226795be1eee9b4f72e340a3d314fb675.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "4065d5fe4068450ea6dc4aefbadd2dc226795be1eee9b4f72e340a3d314fb675",
+  "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": "4847e438a58a7801bad8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d5fcd9315b487bb5736/workflow.json b/records/sd1/sweep/6d5fcd9315b487bb5736/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c94271561d40f756f4c1ac4ac1b1ff820b1f298b
--- /dev/null
+++ b/records/sd1/sweep/6d5fcd9315b487bb5736/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 6d5fcd9315b487bb5736)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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": [
+    "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/6d63b2122d42c5a622d3/measurements.json b/records/sd1/sweep/6d63b2122d42c5a622d3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..42309ab51c20b271d60f8566f2922746be0b7878
--- /dev/null
+++ b/records/sd1/sweep/6d63b2122d42c5a622d3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d63b2122d42c5a622d3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.257342,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.428078,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.504915,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.373004,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.779,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00607,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.92,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9451,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.439,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d63b2122d42c5a622d3/record.json b/records/sd1/sweep/6d63b2122d42c5a622d3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d08f9f478d356020936f6b388c7563f73d5ae34a
--- /dev/null
+++ b/records/sd1/sweep/6d63b2122d42c5a622d3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d63b2122d42c5a622d3",
+ "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.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.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/f00d5a809ce61194099105f8153a5e00cfa41c98c107ef15c0d4444f1c781398.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f00d5a809ce61194099105f8153a5e00cfa41c98c107ef15c0d4444f1c781398",
+  "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": "654782e62307a66550ee"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d63b2122d42c5a622d3/workflow.json b/records/sd1/sweep/6d63b2122d42c5a622d3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..869469419f1d3e1ceecb8850a2dd293839de3bda
--- /dev/null
+++ b/records/sd1/sweep/6d63b2122d42c5a622d3/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 6d63b2122d42c5a622d3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/6d6525e7e92d7d87eb80/measurements.json b/records/sd1/sweep/6d6525e7e92d7d87eb80/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba9f77e54ae20b2172c2d4e21c3e152bae231669
--- /dev/null
+++ b/records/sd1/sweep/6d6525e7e92d7d87eb80/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "6d6525e7e92d7d87eb80",
+ "values": {
+  "cosine_distance": {
+   "value": 0.762054,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.83352,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.923201,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.548493,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 90.133,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02342,
+   "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": 99.917,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.785,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9853,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d6525e7e92d7d87eb80/record.json b/records/sd1/sweep/6d6525e7e92d7d87eb80/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1d7b014acb6e6e3f6adfd22106d7568dc8986e5
--- /dev/null
+++ b/records/sd1/sweep/6d6525e7e92d7d87eb80/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d6525e7e92d7d87eb80",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.skip_connection",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.hi",
+   "kind": "skip",
+   "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/84ddface927dd710e60ccbf55b9fddffcebf0b054ffb37295cdbce7a0d80a9d0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "84ddface927dd710e60ccbf55b9fddffcebf0b054ffb37295cdbce7a0d80a9d0",
+  "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": "4799eab1a23989a08957"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d6525e7e92d7d87eb80/workflow.json b/records/sd1/sweep/6d6525e7e92d7d87eb80/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecebeaae861b02f1541497b02f0eeb8e6d2b6bcf
--- /dev/null
+++ b/records/sd1/sweep/6d6525e7e92d7d87eb80/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 6d6525e7e92d7d87eb80)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.skip_connection\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 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/6d67bbacb5edf7a318ec/measurements.json b/records/sd1/sweep/6d67bbacb5edf7a318ec/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..93119a668677f6779fc45118150b8191e8e01463
--- /dev/null
+++ b/records/sd1/sweep/6d67bbacb5edf7a318ec/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d67bbacb5edf7a318ec",
+ "values": {
+  "std": {
+   "value": 44.141,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01964,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.352,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.215821,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.351882,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.099333,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01557,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9924,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.026,
+   "method": "CLIP zero-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/6d67bbacb5edf7a318ec/record.json b/records/sd1/sweep/6d67bbacb5edf7a318ec/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c27042920119b6a21713d0b1327d760b83d31b5d
--- /dev/null
+++ b/records/sd1/sweep/6d67bbacb5edf7a318ec/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6d67bbacb5edf7a318ec",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:52:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "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": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.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": "a6dcd9a4042c1c6feab2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d67bbacb5edf7a318ec/workflow.json b/records/sd1/sweep/6d67bbacb5edf7a318ec/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..38e30d9a386e6261cffd6a9cd4c6b672cae6622a
--- /dev/null
+++ b/records/sd1/sweep/6d67bbacb5edf7a318ec/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d67bbacb5edf7a318ec)"
+  },
+  "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}",
+   "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/6d6b5e18effd4e79ec00/interpretations.jsonl b/records/sd1/sweep/6d6b5e18effd4e79ec00/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c607421ec43ce4e87e9c8a9e162316a251abc671
--- /dev/null
+++ b/records/sd1/sweep/6d6b5e18effd4e79ec00/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6d6b5e18effd4e79ec00"}, "kind": "caption", "value": "Dark glitchy portrait with ghostly face dissolving into blue-grey static texture", "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": "f7619b3a057ba497"}
+{"target": {"record": "6d6b5e18effd4e79ec00"}, "kind": "change", "value": "The image 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": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a0d5157f11360721"}
+{"target": {"record": "6d6b5e18effd4e79ec00"}, "kind": "keywords", "value": ["glitch", "static", "ghostly face", "dark noise", "degraded"], "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": "e81020cf207489b8"}
+{"target": {"record": "6d6b5e18effd4e79ec00"}, "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": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "55bce293acfe7b2c"}
diff --git a/records/sd1/sweep/6d6b5e18effd4e79ec00/measurements.json b/records/sd1/sweep/6d6b5e18effd4e79ec00/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..61252742c9ef57e3f237c5913ee37c0245407441
--- /dev/null
+++ b/records/sd1/sweep/6d6b5e18effd4e79ec00/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d6b5e18effd4e79ec00",
+ "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/6d6b5e18effd4e79ec00/record.json b/records/sd1/sweep/6d6b5e18effd4e79ec00/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c0e96f945a97f595ac68de0a4b0f90930236f81
--- /dev/null
+++ b/records/sd1/sweep/6d6b5e18effd4e79ec00/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d6b5e18effd4e79ec00",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:38Z",
+ "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": "input_blocks.1.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/d22f5133a4f182ffc73a3638a25da556a0f28ad9ace90db2e1f8ca1d1eda7b15.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d22f5133a4f182ffc73a3638a25da556a0f28ad9ace90db2e1f8ca1d1eda7b15",
+  "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": "a07ddca1d86f9e7e7dc1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d6b5e18effd4e79ec00/workflow.json b/records/sd1/sweep/6d6b5e18effd4e79ec00/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..50861eae0b74a2e1798e50f7f627a1abb978234f
--- /dev/null
+++ b/records/sd1/sweep/6d6b5e18effd4e79ec00/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d6b5e18effd4e79ec00)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/6d7101f1f4876c806ae3/measurements.json b/records/sd1/sweep/6d7101f1f4876c806ae3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..25e49bb368b2651702f227031d673485ddace1aa
--- /dev/null
+++ b/records/sd1/sweep/6d7101f1f4876c806ae3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d7101f1f4876c806ae3",
+ "values": {
+  "std": {
+   "value": 27.856,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03799,
+   "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": 32.342,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.310737,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.522826,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.271322,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.043949,
+   "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": 0.3602,
+   "method": "CLIP zero-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/6d7101f1f4876c806ae3/record.json b/records/sd1/sweep/6d7101f1f4876c806ae3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe5059c9267f63234c42f6ba60d995256b04ad26
--- /dev/null
+++ b/records/sd1/sweep/6d7101f1f4876c806ae3/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "6d7101f1f4876c806ae3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:26:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d7101f1f4876c806ae3/workflow.json b/records/sd1/sweep/6d7101f1f4876c806ae3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea28c5e26706b1e812e2b4e4e42a8b8af5dd677c
--- /dev/null
+++ b/records/sd1/sweep/6d7101f1f4876c806ae3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d7101f1f4876c806ae3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/6d7618af6a801463b8e3/measurements.json b/records/sd1/sweep/6d7618af6a801463b8e3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..838d52b1445cb6031e5758f0004f65116a338f3a
--- /dev/null
+++ b/records/sd1/sweep/6d7618af6a801463b8e3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d7618af6a801463b8e3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.092224,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.13542,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.079418,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.037116,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.147,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00477,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "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": 13.005,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9803,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1125,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d7618af6a801463b8e3/record.json b/records/sd1/sweep/6d7618af6a801463b8e3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f1c692258577080b64b29d022c8374bbb1c08bd
--- /dev/null
+++ b/records/sd1/sweep/6d7618af6a801463b8e3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d7618af6a801463b8e3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:34Z",
+ "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.8.1.proj_out",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.mid",
+   "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/c86003b4e12c237193044e1811ede0c6f580120026f0e5ac9a4a5d9e84da4018.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c86003b4e12c237193044e1811ede0c6f580120026f0e5ac9a4a5d9e84da4018",
+  "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": "86d9ccc8b9979dd14c26"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d7618af6a801463b8e3/workflow.json b/records/sd1/sweep/6d7618af6a801463b8e3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d206c7149e70b63c5604f5d993540dd1a03fdf8
--- /dev/null
+++ b/records/sd1/sweep/6d7618af6a801463b8e3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d7618af6a801463b8e3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.proj_out\",\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/6d794d0672dc70fab290/measurements.json b/records/sd1/sweep/6d794d0672dc70fab290/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4819901d0da668f8803620d9b5dba078b8477565
--- /dev/null
+++ b/records/sd1/sweep/6d794d0672dc70fab290/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d794d0672dc70fab290",
+ "values": {
+  "cosine_distance": {
+   "value": 0.150969,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.222772,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.052805,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.030887,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.851,
+   "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": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.956,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.93,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2675,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d794d0672dc70fab290/record.json b/records/sd1/sweep/6d794d0672dc70fab290/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5637ccaea6e2d6f9d02aefeefdce9d2a84f6ff42
--- /dev/null
+++ b/records/sd1/sweep/6d794d0672dc70fab290/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d794d0672dc70fab290",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:43Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.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/fc6a0a5405b4da7dfa63be4d6b8a1cb9931212a38bf038b0a17bb6cbc85c8362.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "fc6a0a5405b4da7dfa63be4d6b8a1cb9931212a38bf038b0a17bb6cbc85c8362",
+  "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": "b49e525d970d11c37f54"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d794d0672dc70fab290/workflow.json b/records/sd1/sweep/6d794d0672dc70fab290/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..16e905ddc7d1bdacc5eaf6d7b9df5022ac3152c2
--- /dev/null
+++ b/records/sd1/sweep/6d794d0672dc70fab290/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d794d0672dc70fab290)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.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/6d8d9ea4b0351c61bf3a/measurements.json b/records/sd1/sweep/6d8d9ea4b0351c61bf3a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8785bbce9731310935793aa3df0eb29adb00078
--- /dev/null
+++ b/records/sd1/sweep/6d8d9ea4b0351c61bf3a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d8d9ea4b0351c61bf3a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.450052,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.784624,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.822058,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.39838,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 30.938,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01046,
+   "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": 42.118,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8381,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4719,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d8d9ea4b0351c61bf3a/record.json b/records/sd1/sweep/6d8d9ea4b0351c61bf3a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..63294f25facaf7f098c482d7cd8e0dbc6c546a2b
--- /dev/null
+++ b/records/sd1/sweep/6d8d9ea4b0351c61bf3a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d8d9ea4b0351c61bf3a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.out_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/c3f7b7d35de3c6050ba9e2dbb9d8b82031548afbe8117348ed48ed3d12cd3bad.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c3f7b7d35de3c6050ba9e2dbb9d8b82031548afbe8117348ed48ed3d12cd3bad",
+  "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": "fd0d46e6e04bb6e884cb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d8d9ea4b0351c61bf3a/workflow.json b/records/sd1/sweep/6d8d9ea4b0351c61bf3a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..aef525ce6d98d555818d63d3dfc6b2ef4a8798be
--- /dev/null
+++ b/records/sd1/sweep/6d8d9ea4b0351c61bf3a/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 6d8d9ea4b0351c61bf3a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.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/6d8dca889239f9056c3a/measurements.json b/records/sd1/sweep/6d8dca889239f9056c3a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b514839ec476e943b199662f5361a45c7044812
--- /dev/null
+++ b/records/sd1/sweep/6d8dca889239f9056c3a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d8dca889239f9056c3a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.212922,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.301592,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.196414,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.060728,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.188,
+   "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": 28.846,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9727,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0412,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d8dca889239f9056c3a/record.json b/records/sd1/sweep/6d8dca889239f9056c3a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c77c9e805ac3df4498c6b1a3f2131c1097c195bb
--- /dev/null
+++ b/records/sd1/sweep/6d8dca889239f9056c3a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d8dca889239f9056c3a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/de4e94a622153a60c59b475be961f22f2c184d726fcfcb1c333b32c5a2b00506.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "de4e94a622153a60c59b475be961f22f2c184d726fcfcb1c333b32c5a2b00506",
+  "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": "5cba87ae562c28b4aef3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d8dca889239f9056c3a/workflow.json b/records/sd1/sweep/6d8dca889239f9056c3a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..29a27645cb596738048c23763314712b2b1be3be
--- /dev/null
+++ b/records/sd1/sweep/6d8dca889239f9056c3a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d8dca889239f9056c3a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/6d924ce0c052f64825d4/measurements.json b/records/sd1/sweep/6d924ce0c052f64825d4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..50d46eb8adb49017fe47ff6ccb1c882284568720
--- /dev/null
+++ b/records/sd1/sweep/6d924ce0c052f64825d4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d924ce0c052f64825d4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.198494,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.243577,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.087741,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.033324,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.598,
+   "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": 24.074,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9631,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1384,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d924ce0c052f64825d4/record.json b/records/sd1/sweep/6d924ce0c052f64825d4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..29004634809f9979e9af9672808748862f71fecb
--- /dev/null
+++ b/records/sd1/sweep/6d924ce0c052f64825d4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d924ce0c052f64825d4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "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/85a3e0acfc61086fe2c0c3d52d8b97de09e8a70bc025b36d01dc928603d8c5d6.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "85a3e0acfc61086fe2c0c3d52d8b97de09e8a70bc025b36d01dc928603d8c5d6",
+  "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": "ed46d0a1752c4f6cedd7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d924ce0c052f64825d4/workflow.json b/records/sd1/sweep/6d924ce0c052f64825d4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..552534c718e097cea4349b0169346c430bab6cb0
--- /dev/null
+++ b/records/sd1/sweep/6d924ce0c052f64825d4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d924ce0c052f64825d4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d945e0a39a34a84546c/measurements.json b/records/sd1/sweep/6d945e0a39a34a84546c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8b7c84d88416b67fbcdecd84b382f239872a6d2
--- /dev/null
+++ b/records/sd1/sweep/6d945e0a39a34a84546c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d945e0a39a34a84546c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.095085,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.115639,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.128589,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.034225,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.551,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00461,
+   "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": 13.94,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.1429,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d945e0a39a34a84546c/record.json b/records/sd1/sweep/6d945e0a39a34a84546c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..74aa6391b854e55df3ef15997a142774a13e17a1
--- /dev/null
+++ b/records/sd1/sweep/6d945e0a39a34a84546c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d945e0a39a34a84546c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:08Z",
+ "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": "middle_block.2.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9ea7da5385326aa315c3dad34c9337b75ccdd42a6eba8ecc2f95289f410e5193.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9ea7da5385326aa315c3dad34c9337b75ccdd42a6eba8ecc2f95289f410e5193",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "22f44f5c51e46f58179e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d945e0a39a34a84546c/workflow.json b/records/sd1/sweep/6d945e0a39a34a84546c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2bbb68e7c127582f711dbfdcb593946300e399c
--- /dev/null
+++ b/records/sd1/sweep/6d945e0a39a34a84546c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d945e0a39a34a84546c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9644c5d58c0ee2f9fc/measurements.json b/records/sd1/sweep/6d9644c5d58c0ee2f9fc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f22f91e788354683156cd7028fe5a407b910c7c
--- /dev/null
+++ b/records/sd1/sweep/6d9644c5d58c0ee2f9fc/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d9644c5d58c0ee2f9fc",
+ "values": {
+  "cosine_distance": {
+   "value": 0.133822,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.212751,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.072309,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.019316,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.348,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00727,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.806,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.958,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1132,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9644c5d58c0ee2f9fc/record.json b/records/sd1/sweep/6d9644c5d58c0ee2f9fc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e84b6fb668b3df8c4ec8e00d2296a4e0752f933
--- /dev/null
+++ b/records/sd1/sweep/6d9644c5d58c0ee2f9fc/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d9644c5d58c0ee2f9fc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:56Z",
+ "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": 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.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/771bfcfd993567367da72405e59008866ff3d782f0d51e1bce59fc0cb2316897.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "771bfcfd993567367da72405e59008866ff3d782f0d51e1bce59fc0cb2316897",
+  "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": "dc69d6263ec16443a708"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9644c5d58c0ee2f9fc/workflow.json b/records/sd1/sweep/6d9644c5d58c0ee2f9fc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..48e8f7073a95eb8136cfa29b346bef22b8231304
--- /dev/null
+++ b/records/sd1/sweep/6d9644c5d58c0ee2f9fc/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d9644c5d58c0ee2f9fc)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9bffb3b6f572c38660/measurements.json b/records/sd1/sweep/6d9bffb3b6f572c38660/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfdd49d4653f5fa8ad8577ca3c618a9a69c328ed
--- /dev/null
+++ b/records/sd1/sweep/6d9bffb3b6f572c38660/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d9bffb3b6f572c38660",
+ "values": {
+  "cosine_distance": {
+   "value": 0.282568,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.46816,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.325176,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.16869,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 34.88,
+   "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": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.941,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0308,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2183,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9bffb3b6f572c38660/record.json b/records/sd1/sweep/6d9bffb3b6f572c38660/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..626914728382e5c59abc925ea1180bae020e9862
--- /dev/null
+++ b/records/sd1/sweep/6d9bffb3b6f572c38660/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d9bffb3b6f572c38660",
+ "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.norm",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.mid",
+   "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/1aae6697b9e09836341b261ecde88c38dbf04af733ec7224b17e59a0537a7c40.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1aae6697b9e09836341b261ecde88c38dbf04af733ec7224b17e59a0537a7c40",
+  "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": "a6ef9438b48c72ecc19b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9bffb3b6f572c38660/workflow.json b/records/sd1/sweep/6d9bffb3b6f572c38660/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7e351869e2bf10c3f7e0d25fcc207b6d3cf4913
--- /dev/null
+++ b/records/sd1/sweep/6d9bffb3b6f572c38660/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d9bffb3b6f572c38660)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.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/6d9d2e507559437e4b27/measurements.json b/records/sd1/sweep/6d9d2e507559437e4b27/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..83632fd37defe58d13900358fbcf76a6eac970f4
--- /dev/null
+++ b/records/sd1/sweep/6d9d2e507559437e4b27/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6d9d2e507559437e4b27",
+ "values": {
+  "cosine_distance": {
+   "value": 0.141397,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.160912,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.152147,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.171931,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.842,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00526,
+   "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": 16.791,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9505,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/6d9d2e507559437e4b27/record.json b/records/sd1/sweep/6d9d2e507559437e4b27/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f04a67b058f080c1b0ea838582342ee53ca1d78c
--- /dev/null
+++ b/records/sd1/sweep/6d9d2e507559437e4b27/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d9d2e507559437e4b27",
+ "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": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5d470034982f2d3fdfed5d429bf6776d63a5f09e1bdd30f7d2932bf58651e6ac.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5d470034982f2d3fdfed5d429bf6776d63a5f09e1bdd30f7d2932bf58651e6ac",
+  "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": "bdc8759e49663a2b9d31"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9d2e507559437e4b27/workflow.json b/records/sd1/sweep/6d9d2e507559437e4b27/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f42ebadccd73e8d8af02879c0bfdcf9f830034a
--- /dev/null
+++ b/records/sd1/sweep/6d9d2e507559437e4b27/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 6d9d2e507559437e4b27)"
+  },
+  "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\": 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/6d9de95e0c0115332ce1/measurements.json b/records/sd1/sweep/6d9de95e0c0115332ce1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..016a73cbb227614043c16700cce343c462ddff8a
--- /dev/null
+++ b/records/sd1/sweep/6d9de95e0c0115332ce1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d9de95e0c0115332ce1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.440989,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.456271,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.33611,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.278907,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.863,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00528,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.275,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0157,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3884,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9de95e0c0115332ce1/record.json b/records/sd1/sweep/6d9de95e0c0115332ce1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a97ce22aa57fb15d279c978e172a751b9d54575c
--- /dev/null
+++ b/records/sd1/sweep/6d9de95e0c0115332ce1/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d9de95e0c0115332ce1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:59Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/6cb746d60d9c204621b8839883b09707ec89ca160e9f2bb58229410b5a66f3e4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6cb746d60d9c204621b8839883b09707ec89ca160e9f2bb58229410b5a66f3e4",
+  "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": "a51bca470c9e93da3b98"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9de95e0c0115332ce1/workflow.json b/records/sd1/sweep/6d9de95e0c0115332ce1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b3040a35b9fa7f2852f12d3d7d94e475a53ab4cb
--- /dev/null
+++ b/records/sd1/sweep/6d9de95e0c0115332ce1/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 6d9de95e0c0115332ce1)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/6d9f14013414526a1973/measurements.json b/records/sd1/sweep/6d9f14013414526a1973/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..05211b640abb288ad3257f2eb6cd6cd28ae30a24
--- /dev/null
+++ b/records/sd1/sweep/6d9f14013414526a1973/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d9f14013414526a1973",
+ "values": {
+  "cosine_distance": {
+   "value": 0.019815,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.041371,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.014318,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007192,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.658,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0061,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.771,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0313,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1012,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9f14013414526a1973/record.json b/records/sd1/sweep/6d9f14013414526a1973/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8151d83792a9c2a74fc3cb6fed0a188920d9b065
--- /dev/null
+++ b/records/sd1/sweep/6d9f14013414526a1973/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6d9f14013414526a1973",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:42Z",
+ "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": 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.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "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/00cd3c4160686ef6b8d1e3fd9ccfe758720863d57045d166311667e14cc8a44b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "00cd3c4160686ef6b8d1e3fd9ccfe758720863d57045d166311667e14cc8a44b",
+  "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": "325ffa54349cb54131eb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9f14013414526a1973/workflow.json b/records/sd1/sweep/6d9f14013414526a1973/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed768d9d2a4b763d1de91c08aa3bf4c01a72cc50
--- /dev/null
+++ b/records/sd1/sweep/6d9f14013414526a1973/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6d9f14013414526a1973)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.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/6d9f42243640fe56e058/measurements.json b/records/sd1/sweep/6d9f42243640fe56e058/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..caf48d4a2231370a8f3f060b271a6fb641f1fe66
--- /dev/null
+++ b/records/sd1/sweep/6d9f42243640fe56e058/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6d9f42243640fe56e058",
+ "values": {
+  "cosine_distance": {
+   "value": 0.514243,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.498482,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.405548,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.277055,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 78.748,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0122,
+   "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": 59.662,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9996,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3659,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9f42243640fe56e058/record.json b/records/sd1/sweep/6d9f42243640fe56e058/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1ee2989517528947e12d93fe5d27e7e0f3307c4
--- /dev/null
+++ b/records/sd1/sweep/6d9f42243640fe56e058/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6d9f42243640fe56e058",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:27Z",
+ "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.1.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.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/1de662527210e938bc142b8921793bd7471fea549ac3852142cf00c1e0e2423b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1de662527210e938bc142b8921793bd7471fea549ac3852142cf00c1e0e2423b",
+  "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": "a03784e9e9fcbd787e5d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6d9f42243640fe56e058/workflow.json b/records/sd1/sweep/6d9f42243640fe56e058/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae4b9d4f14361dd0a8aa3d4c3cccdc5fa0d02b40
--- /dev/null
+++ b/records/sd1/sweep/6d9f42243640fe56e058/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 6d9f42243640fe56e058)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.1.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/6da51ebc4ec72810d110/interpretations.jsonl b/records/sd1/sweep/6da51ebc4ec72810d110/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..95c004f3c750c5a8bb2008256810649f24d3a2e0
--- /dev/null
+++ b/records/sd1/sweep/6da51ebc4ec72810d110/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6da51ebc4ec72810d110"}, "kind": "caption", "value": "Chaotic abstract mix of teal, orange, cream and black painterly patches", "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": "ce7d2af9762a7ace"}
+{"target": {"record": "6da51ebc4ec72810d110"}, "kind": "change", "value": "The portrait dissolves into a chaotic, colourful abstract of loose teal, orange and white daubs with no recognisable figure.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:23:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2c479ae1fbf0434e"}
+{"target": {"record": "6da51ebc4ec72810d110"}, "kind": "keywords", "value": ["abstract", "colourful daubs", "teal orange", "chaotic", "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": "c27a720d1a9986b4"}
+{"target": {"record": "6da51ebc4ec72810d110"}, "kind": "effect_tags", "value": ["abstract", "painterly", "saturated", "colour_shift", "subject_lost", "fragmented"], "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": "ad42bffda20aadd7"}
diff --git a/records/sd1/sweep/6da51ebc4ec72810d110/measurements.json b/records/sd1/sweep/6da51ebc4ec72810d110/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5fc017d83ee92483e24d54cf6c3283f332103d2c
--- /dev/null
+++ b/records/sd1/sweep/6da51ebc4ec72810d110/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6da51ebc4ec72810d110",
+ "values": {
+  "cosine_distance": {
+   "value": 0.635987,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.72905,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.860406,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.420021,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.796,
+   "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": 65.999,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8115,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2894,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6da51ebc4ec72810d110/record.json b/records/sd1/sweep/6da51ebc4ec72810d110/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..65bc3b352c6a9319f7fb8784c2fe262aa9d40259
--- /dev/null
+++ b/records/sd1/sweep/6da51ebc4ec72810d110/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6da51ebc4ec72810d110",
+ "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": 60
+   },
+   "group": "out.mid",
+   "kind": "upsample",
+   "module_type": "Conv2d",
+   "container": "Upsample",
+   "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/d3fb3210c2bd293920a0b47203b7a20b804661e516f77e14b4cc989a8a99d565.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d3fb3210c2bd293920a0b47203b7a20b804661e516f77e14b4cc989a8a99d565",
+  "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": "20cffcf95cedb2cf67eb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6da51ebc4ec72810d110/workflow.json b/records/sd1/sweep/6da51ebc4ec72810d110/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9229bdaf7cb834488cdc1bc88622ae76547035d8
--- /dev/null
+++ b/records/sd1/sweep/6da51ebc4ec72810d110/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 6da51ebc4ec72810d110)"
+  },
+  "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\": 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/6da68d68a8598dfceee9/interpretations.jsonl b/records/sd1/sweep/6da68d68a8598dfceee9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7c651a85008ee96aa5f69f8d93250de13e7e17d7
--- /dev/null
+++ b/records/sd1/sweep/6da68d68a8598dfceee9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6da68d68a8598dfceee9"}, "kind": "caption", "value": "Chaotic abstract patches of electric blue, green, yellow, white and black", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:12:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "72de3f5f5d50141a"}
+{"target": {"record": "6da68d68a8598dfceee9"}, "kind": "change", "value": "As in C, the portrait is completely lost to a saturated, glitchy blotch texture in blues, greens and yellows with black gaps.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:12:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8a5ddfb2eb81729e"}
+{"target": {"record": "6da68d68a8598dfceee9"}, "kind": "keywords", "value": ["abstract noise", "glitch", "saturated blotches", "camouflage texture", "subject lost"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:12:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d92f3ca84c5ab1cf"}
+{"target": {"record": "6da68d68a8598dfceee9"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "glitch", "saturated", "colour_shift", "texture", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:12:51Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3aa8d4a149bde0e0"}
diff --git a/records/sd1/sweep/6da68d68a8598dfceee9/measurements.json b/records/sd1/sweep/6da68d68a8598dfceee9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..872ddc771ec9dff9e0642e311700e94940b2ed39
--- /dev/null
+++ b/records/sd1/sweep/6da68d68a8598dfceee9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6da68d68a8598dfceee9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.765431,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.8079,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.934913,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.424659,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 89.795,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0416,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 94.717,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9164,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8989,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6da68d68a8598dfceee9/record.json b/records/sd1/sweep/6da68d68a8598dfceee9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6138d920ae56773780d16585567f3e4681349c55
--- /dev/null
+++ b/records/sd1/sweep/6da68d68a8598dfceee9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6da68d68a8598dfceee9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "time_embed.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "embed",
+   "kind": "time_embed",
+   "module_type": "SiLU",
+   "container": "Sequential",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/57a5a1b7e8b069cde8f884b921610a48edf400a73a793e8b6be26c5a752914e0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "57a5a1b7e8b069cde8f884b921610a48edf400a73a793e8b6be26c5a752914e0",
+  "type_path": "UNetModel.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "31a666858dcb6ccac2e8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6da68d68a8598dfceee9/workflow.json b/records/sd1/sweep/6da68d68a8598dfceee9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..90c69731d96c2d452cdd77ae29b83c9c21157afc
--- /dev/null
+++ b/records/sd1/sweep/6da68d68a8598dfceee9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6da68d68a8598dfceee9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"time_embed.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6daf8ab7d506644a8033/measurements.json b/records/sd1/sweep/6daf8ab7d506644a8033/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..61bd9bc0d5f2c4e3885fcf5704f97b456adbce6e
--- /dev/null
+++ b/records/sd1/sweep/6daf8ab7d506644a8033/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6daf8ab7d506644a8033",
+ "values": {
+  "cosine_distance": {
+   "value": 0.156292,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.148307,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.067125,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.062153,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.966,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00405,
+   "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": 17.098,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9288,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5496,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6daf8ab7d506644a8033/record.json b/records/sd1/sweep/6daf8ab7d506644a8033/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..70a108761c52508ff4c0935124f7e9e69908f6d7
--- /dev/null
+++ b/records/sd1/sweep/6daf8ab7d506644a8033/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6daf8ab7d506644a8033",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e4303fee06675d59016fbef1312747f61d14703caa1ad6e3e12af15ded18677c.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e4303fee06675d59016fbef1312747f61d14703caa1ad6e3e12af15ded18677c",
+  "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": "61e5741e31c1927197da"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6daf8ab7d506644a8033/workflow.json b/records/sd1/sweep/6daf8ab7d506644a8033/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9fae63d687d5d77f4cdfdfff6c5cdf821c9deaa
--- /dev/null
+++ b/records/sd1/sweep/6daf8ab7d506644a8033/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 6daf8ab7d506644a8033)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.proj_out\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6db9194d017b94e3b0c4/measurements.json b/records/sd1/sweep/6db9194d017b94e3b0c4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ddec98c2718daf2b0651642b4b1db0a69a6b4784
--- /dev/null
+++ b/records/sd1/sweep/6db9194d017b94e3b0c4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6db9194d017b94e3b0c4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.150533,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.328416,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.307259,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.185308,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 72.425,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00345,
+   "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": 22.374,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0102,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1007,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6db9194d017b94e3b0c4/record.json b/records/sd1/sweep/6db9194d017b94e3b0c4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..778a5059d05814c2fa99884e4a00f286a28b94f7
--- /dev/null
+++ b/records/sd1/sweep/6db9194d017b94e3b0c4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6db9194d017b94e3b0c4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.hi",
+   "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/20fb3afcebe2c226c5ae289385b3eda5bae81d9df84ed92216b03638023d603f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "20fb3afcebe2c226c5ae289385b3eda5bae81d9df84ed92216b03638023d603f",
+  "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": "8d8cbc5e00dacf9afd2f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6db9194d017b94e3b0c4/workflow.json b/records/sd1/sweep/6db9194d017b94e3b0c4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d50b708584cbfa3662c4851c85b48af800b8580
--- /dev/null
+++ b/records/sd1/sweep/6db9194d017b94e3b0c4/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 6db9194d017b94e3b0c4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.in_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/6dbac813d0aaeeac1897/measurements.json b/records/sd1/sweep/6dbac813d0aaeeac1897/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0bade511701a02227e7dfd5992dbbd068121687f
--- /dev/null
+++ b/records/sd1/sweep/6dbac813d0aaeeac1897/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6dbac813d0aaeeac1897",
+ "values": {
+  "std": {
+   "value": 29.785,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03025,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.801,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.275396,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.428146,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131042,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023444,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0034,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1853,
+   "method": "CLIP zero-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/6dbac813d0aaeeac1897/record.json b/records/sd1/sweep/6dbac813d0aaeeac1897/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b91169415e4474c276d815540f56795e1867ae0f
--- /dev/null
+++ b/records/sd1/sweep/6dbac813d0aaeeac1897/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6dbac813d0aaeeac1897",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:57:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "slight",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.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": "19bfe4d4105668c264c6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dbac813d0aaeeac1897/workflow.json b/records/sd1/sweep/6dbac813d0aaeeac1897/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c31257535ce18e9094c6e4a735b4eefc397ff48d
--- /dev/null
+++ b/records/sd1/sweep/6dbac813d0aaeeac1897/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6dbac813d0aaeeac1897)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dbbd856ce17660582e0/measurements.json b/records/sd1/sweep/6dbbd856ce17660582e0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..94a281fde31720b38514ca25aadb7a7fc1d2aef1
--- /dev/null
+++ b/records/sd1/sweep/6dbbd856ce17660582e0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6dbbd856ce17660582e0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.308368,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.594854,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.567043,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.404259,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 72.525,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00192,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.131,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0201,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1375,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dbbd856ce17660582e0/record.json b/records/sd1/sweep/6dbbd856ce17660582e0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e621a8a52629f0c065504f818d8cde9ee12803f
--- /dev/null
+++ b/records/sd1/sweep/6dbbd856ce17660582e0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6dbbd856ce17660582e0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:38Z",
+ "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": "output_blocks.7.1.transformer_blocks.0.attn1.to_v",
+   "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": "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/55129885766d13b00bf1e77728a01bedce0aa5fa6d79d968c57e5ae276ece064.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "55129885766d13b00bf1e77728a01bedce0aa5fa6d79d968c57e5ae276ece064",
+  "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": "b46192e11a5cc1603fbc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dbbd856ce17660582e0/workflow.json b/records/sd1/sweep/6dbbd856ce17660582e0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2342d80f11bd58b7fb5c7eff62b910d23c8522e3
--- /dev/null
+++ b/records/sd1/sweep/6dbbd856ce17660582e0/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 6dbbd856ce17660582e0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.attn1.to_v\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/6dc02d4ea41b17b0b167/measurements.json b/records/sd1/sweep/6dc02d4ea41b17b0b167/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac57c2b1a14d9adb1f6e93f5d57356ee50de44bf
--- /dev/null
+++ b/records/sd1/sweep/6dc02d4ea41b17b0b167/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6dc02d4ea41b17b0b167",
+ "values": {
+  "cosine_distance": {
+   "value": 0.141504,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.205345,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.075196,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025274,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.741,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00694,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.983,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9599,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.097,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dc02d4ea41b17b0b167/record.json b/records/sd1/sweep/6dc02d4ea41b17b0b167/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..878e491a6d0fbcb6e854ee30e8dfba0bad41e873
--- /dev/null
+++ b/records/sd1/sweep/6dc02d4ea41b17b0b167/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6dc02d4ea41b17b0b167",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:50Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1d535afff9432d92698d6a50e4123417a82c306e5b73e35460b8086f39b3cdde.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1d535afff9432d92698d6a50e4123417a82c306e5b73e35460b8086f39b3cdde",
+  "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": "53bdaaf0d8c71f4329d5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dc02d4ea41b17b0b167/workflow.json b/records/sd1/sweep/6dc02d4ea41b17b0b167/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5acf7d710c66b20bde29e5e80aee6d2bcca9f655
--- /dev/null
+++ b/records/sd1/sweep/6dc02d4ea41b17b0b167/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6dc02d4ea41b17b0b167)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/6dc7712401ec851f5790/measurements.json b/records/sd1/sweep/6dc7712401ec851f5790/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..477231bbcad6a505734ac41db742aeafc4d9f042
--- /dev/null
+++ b/records/sd1/sweep/6dc7712401ec851f5790/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6dc7712401ec851f5790",
+ "values": {
+  "cosine_distance": {
+   "value": 0.059421,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.080027,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.052482,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01354,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.454,
+   "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": 9.595,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0076,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1146,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dc7712401ec851f5790/record.json b/records/sd1/sweep/6dc7712401ec851f5790/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..92dc3849ba461c6ba4b9f38c411e6abff442f18b
--- /dev/null
+++ b/records/sd1/sweep/6dc7712401ec851f5790/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6dc7712401ec851f5790",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:45Z",
+ "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.9.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.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/dd2d75d57bce17186162f49feab221f4afd1e9e7e904390a4e38b30f24415007.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "dd2d75d57bce17186162f49feab221f4afd1e9e7e904390a4e38b30f24415007",
+  "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": "39c4fb1360ca80d9a647"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dc7712401ec851f5790/workflow.json b/records/sd1/sweep/6dc7712401ec851f5790/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..77c3800352da8e8bdcae1fa19a261e5e9d3e9985
--- /dev/null
+++ b/records/sd1/sweep/6dc7712401ec851f5790/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6dc7712401ec851f5790)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.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/6dcda9f71d75a24f4ffa/measurements.json b/records/sd1/sweep/6dcda9f71d75a24f4ffa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c3fb97844fa1396ae3e77b56e3ee53b942ec74d
--- /dev/null
+++ b/records/sd1/sweep/6dcda9f71d75a24f4ffa/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6dcda9f71d75a24f4ffa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.364631,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.388329,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.235107,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.147191,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.825,
+   "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": 38.264,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0117,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2658,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dcda9f71d75a24f4ffa/record.json b/records/sd1/sweep/6dcda9f71d75a24f4ffa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..060cf1bd4b844e357f4087830c564536ddee8ea7
--- /dev/null
+++ b/records/sd1/sweep/6dcda9f71d75a24f4ffa/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6dcda9f71d75a24f4ffa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.proj_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.lo",
+   "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/24071f8721214cc3ee5bf82bec01fbc0214787f9d3fec6ced0ef8330861cf037.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "24071f8721214cc3ee5bf82bec01fbc0214787f9d3fec6ced0ef8330861cf037",
+  "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": "4584333723a3eb4b9bd4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dcda9f71d75a24f4ffa/workflow.json b/records/sd1/sweep/6dcda9f71d75a24f4ffa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f3d9743da42c38a792ebdfa81a44b8601cb838c6
--- /dev/null
+++ b/records/sd1/sweep/6dcda9f71d75a24f4ffa/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 6dcda9f71d75a24f4ffa)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.proj_out\",\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/6dcfb9cb3e8693e8b7d3/measurements.json b/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a80930a51888a81c63cda3d6ee40bd26ee196d4a
--- /dev/null
+++ b/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6dcfb9cb3e8693e8b7d3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.477408,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.432765,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.346721,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.156277,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.257,
+   "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": 49.089,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9459,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.528,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/record.json b/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..118222d5c02738af849438654f77c8f913020a8b
--- /dev/null
+++ b/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6dcfb9cb3e8693e8b7d3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:24Z",
+ "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": "input_blocks.7.0.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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8c3e2a16057f04af1ab05a03e5914a286b6511e72ca1829265d6c8befcc13cde.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8c3e2a16057f04af1ab05a03e5914a286b6511e72ca1829265d6c8befcc13cde",
+  "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": "00c6f477e75d31f28f0a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/workflow.json b/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ab25633b9b4fcc7118a5a98177de8fa5f7d2a74
--- /dev/null
+++ b/records/sd1/sweep/6dcfb9cb3e8693e8b7d3/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 6dcfb9cb3e8693e8b7d3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.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/6dd393537d8402283306/measurements.json b/records/sd1/sweep/6dd393537d8402283306/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c5903eb9659146ba994fd70470b6ef4b8b64f79
--- /dev/null
+++ b/records/sd1/sweep/6dd393537d8402283306/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6dd393537d8402283306",
+ "values": {
+  "cosine_distance": {
+   "value": 0.419342,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.410473,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.313389,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.247033,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 77.342,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0059,
+   "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": 48.771,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8865,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1844,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dd393537d8402283306/record.json b/records/sd1/sweep/6dd393537d8402283306/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e2e9c730d9931de78a347ff50d37938ef226800
--- /dev/null
+++ b/records/sd1/sweep/6dd393537d8402283306/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6dd393537d8402283306",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "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/3d0180d4f789e9f57a120ae480e89131b7e52d089bb6e61f7f90530126b4f872.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3d0180d4f789e9f57a120ae480e89131b7e52d089bb6e61f7f90530126b4f872",
+  "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": "f3945ae4a34aeae173e9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dd393537d8402283306/workflow.json b/records/sd1/sweep/6dd393537d8402283306/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..33c819b6da9b3738c240c6a0aa23fd7bb87615a0
--- /dev/null
+++ b/records/sd1/sweep/6dd393537d8402283306/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 6dd393537d8402283306)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.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/6dd749b23c1797e7c699/measurements.json b/records/sd1/sweep/6dd749b23c1797e7c699/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..263e53658bfc8083bf80b30d66e6b6417d1477b8
--- /dev/null
+++ b/records/sd1/sweep/6dd749b23c1797e7c699/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6dd749b23c1797e7c699",
+ "values": {
+  "std": {
+   "value": 32.432,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02161,
+   "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": 5.89,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.014353,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.03682,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.019965,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004963,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9979,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2191,
+   "method": "CLIP zero-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/6dd749b23c1797e7c699/record.json b/records/sd1/sweep/6dd749b23c1797e7c699/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..05c3e47b1885f31931584967f1893ab2e6df747e
--- /dev/null
+++ b/records/sd1/sweep/6dd749b23c1797e7c699/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "6dd749b23c1797e7c699",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:23:57Z",
+ "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": "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": "81ef55543b3b3c080780",
+  "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/6dd749b23c1797e7c699/workflow.json b/records/sd1/sweep/6dd749b23c1797e7c699/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8afebdcaebf2b908389fa0006c107b4d331b673
--- /dev/null
+++ b/records/sd1/sweep/6dd749b23c1797e7c699/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6dd749b23c1797e7c699)"
+  },
+  "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 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/6dd7fffe020efe2395a6/measurements.json b/records/sd1/sweep/6dd7fffe020efe2395a6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2261c9ec11259f7999788eab9e7762ed2b54e851
--- /dev/null
+++ b/records/sd1/sweep/6dd7fffe020efe2395a6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6dd7fffe020efe2395a6",
+ "values": {
+  "std": {
+   "value": 62.971,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00834,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.968,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.258831,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.268327,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.050947,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.06473,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9162,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2949,
+   "method": "CLIP zero-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/6dd7fffe020efe2395a6/record.json b/records/sd1/sweep/6dd7fffe020efe2395a6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cec6c3ba3ecbadb223512f9fa2c280223b140abc
--- /dev/null
+++ b/records/sd1/sweep/6dd7fffe020efe2395a6/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "6dd7fffe020efe2395a6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:34:48Z",
+ "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": "input_blocks.2.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "mid",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_scalar",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6dd7fffe020efe2395a6/workflow.json b/records/sd1/sweep/6dd7fffe020efe2395a6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b41456a054b7c4c9e67af174d0e42b17b83087d1
--- /dev/null
+++ b/records/sd1/sweep/6dd7fffe020efe2395a6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6dd7fffe020efe2395a6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/6ddd1aff73846a043eac/measurements.json b/records/sd1/sweep/6ddd1aff73846a043eac/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dfa1a06b0a3f4e6409235bcb1e06dd61b18dd901
--- /dev/null
+++ b/records/sd1/sweep/6ddd1aff73846a043eac/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ddd1aff73846a043eac",
+ "values": {
+  "std": {
+   "value": 30.433,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0059,
+   "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": 25.355,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.15113,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.508018,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.986213,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.337135,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6378,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8495,
+   "method": "CLIP zero-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/6ddd1aff73846a043eac/record.json b/records/sd1/sweep/6ddd1aff73846a043eac/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a02a8734fd8637e60571691f5cb6b8057ced83
--- /dev/null
+++ b/records/sd1/sweep/6ddd1aff73846a043eac/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "6ddd1aff73846a043eac",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:31:10Z",
+ "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": 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": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "slight",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   30.0,
+   "late"
+  ],
+  "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": "b659db08f6924dfd1299"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ddd1aff73846a043eac/workflow.json b/records/sd1/sweep/6ddd1aff73846a043eac/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..269eca5efd3783d920b0e7d4e4f42dab5ce13e50
--- /dev/null
+++ b/records/sd1/sweep/6ddd1aff73846a043eac/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ddd1aff73846a043eac)"
+  },
+  "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\": 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/6deceb913492b6f9802d/interpretations.jsonl b/records/sd1/sweep/6deceb913492b6f9802d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..64bebb83c3cec3c2bb037fe79998c13cf728333a
--- /dev/null
+++ b/records/sd1/sweep/6deceb913492b6f9802d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6deceb913492b6f9802d"}, "kind": "caption", "value": "Short-haired woman in black top, three-quarter view, grey framed studio backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b6ac18abed02682c"}
+{"target": {"record": "6deceb913492b6f9802d"}, "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": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4c8ebdc793b29841"}
+{"target": {"record": "6deceb913492b6f9802d"}, "kind": "keywords", "value": ["near identical", "grey studio portrait", "no change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f6add221a4771463"}
+{"target": {"record": "6deceb913492b6f9802d"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f56a51813c2e6744"}
diff --git a/records/sd1/sweep/6deceb913492b6f9802d/measurements.json b/records/sd1/sweep/6deceb913492b6f9802d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..32f4f0997462ee441b840acb2c0a80ce0ac932db
--- /dev/null
+++ b/records/sd1/sweep/6deceb913492b6f9802d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6deceb913492b6f9802d",
+ "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/6deceb913492b6f9802d/record.json b/records/sd1/sweep/6deceb913492b6f9802d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e42a07e688397cb72c2f260ca5d7ed6a3011590
--- /dev/null
+++ b/records/sd1/sweep/6deceb913492b6f9802d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6deceb913492b6f9802d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/0bf05a11e229c25145409a1bdaea8cba37590e5fa72d256c6b15e3b744600def.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0bf05a11e229c25145409a1bdaea8cba37590e5fa72d256c6b15e3b744600def",
+  "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": "9b7d7d48a52923185d43"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6deceb913492b6f9802d/workflow.json b/records/sd1/sweep/6deceb913492b6f9802d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..763b1281b3319fe195009585a1446121995ac094
--- /dev/null
+++ b/records/sd1/sweep/6deceb913492b6f9802d/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 6deceb913492b6f9802d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.h_upd\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6defc20b75d65847c264/measurements.json b/records/sd1/sweep/6defc20b75d65847c264/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0d32edc0181ac5b62b7b269378952f4fde5cb79
--- /dev/null
+++ b/records/sd1/sweep/6defc20b75d65847c264/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6defc20b75d65847c264",
+ "values": {
+  "std": {
+   "value": 41.327,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01952,
+   "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": 48.295,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.502027,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.56818,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.299595,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.084928,
+   "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": 0.0171,
+   "method": "CLIP zero-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/6defc20b75d65847c264/record.json b/records/sd1/sweep/6defc20b75d65847c264/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7180a1bdd3e6e8f65e9753c855453f741fd7f15
--- /dev/null
+++ b/records/sd1/sweep/6defc20b75d65847c264/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6defc20b75d65847c264",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:53:33Z",
+ "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": 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": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "mid",
+   "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": [
+   "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": "d7909c6ce307c621f6a0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6defc20b75d65847c264/workflow.json b/records/sd1/sweep/6defc20b75d65847c264/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3451340c89d65b34533d2e39d0f4b8c6f468485e
--- /dev/null
+++ b/records/sd1/sweep/6defc20b75d65847c264/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6defc20b75d65847c264)"
+  },
+  "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}",
+   "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/6df119b22c2e6668017d/measurements.json b/records/sd1/sweep/6df119b22c2e6668017d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebca94ca660ddc9ab8db027fc2bedef23561bc9d
--- /dev/null
+++ b/records/sd1/sweep/6df119b22c2e6668017d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6df119b22c2e6668017d",
+ "values": {
+  "std": {
+   "value": 32.087,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01374,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.664,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.372046,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.502171,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.157326,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.070655,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9653,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.065,
+   "method": "CLIP zero-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/6df119b22c2e6668017d/record.json b/records/sd1/sweep/6df119b22c2e6668017d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f30dc1e95324d24505c10ecfa91214cb13c5f70d
--- /dev/null
+++ b/records/sd1/sweep/6df119b22c2e6668017d/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "6df119b22c2e6668017d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:09:46Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "shrink",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.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/6df119b22c2e6668017d/workflow.json b/records/sd1/sweep/6df119b22c2e6668017d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2ce038a12e7ad2d60734448df42c326ac7b7204
--- /dev/null
+++ b/records/sd1/sweep/6df119b22c2e6668017d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6df119b22c2e6668017d)"
+  },
+  "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\": 0.7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6df6b48e05d550d32842/measurements.json b/records/sd1/sweep/6df6b48e05d550d32842/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e789b9aef7837323830d1f6dda861346ae751ca0
--- /dev/null
+++ b/records/sd1/sweep/6df6b48e05d550d32842/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6df6b48e05d550d32842",
+ "values": {
+  "std": {
+   "value": 42.663,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00244,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "clip_degenerate"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 52.779,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.604212,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.862816,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.910278,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.438243,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6001,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9624,
+   "method": "CLIP zero-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/6df6b48e05d550d32842/record.json b/records/sd1/sweep/6df6b48e05d550d32842/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a1494f663e353914bdcb886cac8d26cc942b4fa
--- /dev/null
+++ b/records/sd1/sweep/6df6b48e05d550d32842/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "6df6b48e05d550d32842",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:27:51Z",
+ "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": "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": 2.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "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",
+   2.0,
+   "all"
+  ],
+  "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": "54dc195e80e62a9b55d9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6df6b48e05d550d32842/workflow.json b/records/sd1/sweep/6df6b48e05d550d32842/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..70b8755b3e99655e98c5bf2b038944acacf651fc
--- /dev/null
+++ b/records/sd1/sweep/6df6b48e05d550d32842/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6df6b48e05d550d32842)"
+  },
+  "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}",
+   "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/6df7354908dc48285f82/measurements.json b/records/sd1/sweep/6df7354908dc48285f82/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..56e5e2832625002c0d2a98ad6ed969b8961e2753
--- /dev/null
+++ b/records/sd1/sweep/6df7354908dc48285f82/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6df7354908dc48285f82",
+ "values": {
+  "std": {
+   "value": 75.991,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01374,
+   "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": 39.982,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.251452,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.331377,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.0753,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.051113,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0176,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.181,
+   "method": "CLIP zero-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/6df7354908dc48285f82/record.json b/records/sd1/sweep/6df7354908dc48285f82/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..abf865ed57d7943e0373b9c61f133d7ef75a73e5
--- /dev/null
+++ b/records/sd1/sweep/6df7354908dc48285f82/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6df7354908dc48285f82",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:55:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "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": "d75f7d07ca225b6809fd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6df7354908dc48285f82/workflow.json b/records/sd1/sweep/6df7354908dc48285f82/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb95ad4927c398e7365b870c3172c7d9d6f1d329
--- /dev/null
+++ b/records/sd1/sweep/6df7354908dc48285f82/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6df7354908dc48285f82)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ]\n}",
+   "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/6e0a626706f0e9eb86aa/measurements.json b/records/sd1/sweep/6e0a626706f0e9eb86aa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1df1f561abbf8728c6995255ee674880fc94b146
--- /dev/null
+++ b/records/sd1/sweep/6e0a626706f0e9eb86aa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e0a626706f0e9eb86aa",
+ "values": {
+  "std": {
+   "value": 27.55,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02552,
+   "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": 12.887,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.088643,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.22506,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.071875,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.024177,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9486,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4305,
+   "method": "CLIP zero-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/6e0a626706f0e9eb86aa/record.json b/records/sd1/sweep/6e0a626706f0e9eb86aa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fd8db10efaa69833f03a1180ac4a5caefac2d08
--- /dev/null
+++ b/records/sd1/sweep/6e0a626706f0e9eb86aa/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "6e0a626706f0e9eb86aa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:59:16Z",
+ "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": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "heavy"
+  }
+ ],
+ "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",
+   0.8,
+   "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": "f1fa3fb84f6ffe1828f5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e0a626706f0e9eb86aa/workflow.json b/records/sd1/sweep/6e0a626706f0e9eb86aa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..066435c152345350d982170b3b6acf300808e26a
--- /dev/null
+++ b/records/sd1/sweep/6e0a626706f0e9eb86aa/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e0a626706f0e9eb86aa)"
+  },
+  "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\": 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/6e0b657592705b5d20c5/measurements.json b/records/sd1/sweep/6e0b657592705b5d20c5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..34f30abc5213e8aaebe38490d5fb34dbed70550b
--- /dev/null
+++ b/records/sd1/sweep/6e0b657592705b5d20c5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e0b657592705b5d20c5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.349887,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.608089,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.492613,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.419951,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 84.675,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00221,
+   "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": 46.328,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9655,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3816,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e0b657592705b5d20c5/record.json b/records/sd1/sweep/6e0b657592705b5d20c5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6d31b567f7f720ee3b8b85dcb46497bb3476406
--- /dev/null
+++ b/records/sd1/sweep/6e0b657592705b5d20c5/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e0b657592705b5d20c5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:39Z",
+ "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": "output_blocks.7.1.transformer_blocks.0.attn1.to_out.0",
+   "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/883cba3a56fdbea57272765077413746280fc3ff3fd2fb580851ec12130de930.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "883cba3a56fdbea57272765077413746280fc3ff3fd2fb580851ec12130de930",
+  "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": "ff59e52559108eada51a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e0b657592705b5d20c5/workflow.json b/records/sd1/sweep/6e0b657592705b5d20c5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b746c89401f6cf52d846273e16fc1bcbc4be8264
--- /dev/null
+++ b/records/sd1/sweep/6e0b657592705b5d20c5/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 6e0b657592705b5d20c5)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.attn1.to_out.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\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/6e0ee66fc6b07471c3c7/measurements.json b/records/sd1/sweep/6e0ee66fc6b07471c3c7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cce701a0af4b7fad93884a2d61af6dc082eaaa88
--- /dev/null
+++ b/records/sd1/sweep/6e0ee66fc6b07471c3c7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e0ee66fc6b07471c3c7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.186882,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.229913,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.149052,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.038089,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.696,
+   "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": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.971,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9677,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.158,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e0ee66fc6b07471c3c7/record.json b/records/sd1/sweep/6e0ee66fc6b07471c3c7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..01a98cdd26f8f681490a35f7f4001e06d9bf122e
--- /dev/null
+++ b/records/sd1/sweep/6e0ee66fc6b07471c3c7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6e0ee66fc6b07471c3c7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:10Z",
+ "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": "middle_block.1.transformer_blocks.0.attn1.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "mid",
+   "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/c0d7c02ece2c512ec7ee028c6923ee481252d8ef4aef8c2756e63103c39990ed.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c0d7c02ece2c512ec7ee028c6923ee481252d8ef4aef8c2756e63103c39990ed",
+  "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",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "cfc29638ac7eb10070df"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e0ee66fc6b07471c3c7/workflow.json b/records/sd1/sweep/6e0ee66fc6b07471c3c7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..083743eaa62c8871ede29290f6c35387753dceaa
--- /dev/null
+++ b/records/sd1/sweep/6e0ee66fc6b07471c3c7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e0ee66fc6b07471c3c7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/6e15ab15bc6497fc9c62/measurements.json b/records/sd1/sweep/6e15ab15bc6497fc9c62/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c45c68d098a9eab72be2dc42e498df95dbc6b4c
--- /dev/null
+++ b/records/sd1/sweep/6e15ab15bc6497fc9c62/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e15ab15bc6497fc9c62",
+ "values": {
+  "cosine_distance": {
+   "value": 0.051423,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.072969,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.021494,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.011103,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.882,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00407,
+   "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": 12.465,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0224,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1152,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e15ab15bc6497fc9c62/record.json b/records/sd1/sweep/6e15ab15bc6497fc9c62/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d4598faacf40b40ce1141048d0efa938e396cf37
--- /dev/null
+++ b/records/sd1/sweep/6e15ab15bc6497fc9c62/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6e15ab15bc6497fc9c62",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:59Z",
+ "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.10.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/a5065671a9d5da97f698d91a8195318a3ab0c73b469f7579b0a9dc4ae6f4c02c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a5065671a9d5da97f698d91a8195318a3ab0c73b469f7579b0a9dc4ae6f4c02c",
+  "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": "9ee9aab494341ac6ecbd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e15ab15bc6497fc9c62/workflow.json b/records/sd1/sweep/6e15ab15bc6497fc9c62/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..24f1a5c9b9bdd8a53364b93f7ff1aeeacebfe0c1
--- /dev/null
+++ b/records/sd1/sweep/6e15ab15bc6497fc9c62/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e15ab15bc6497fc9c62)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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/6e160e79e1310877d46b/measurements.json b/records/sd1/sweep/6e160e79e1310877d46b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa93c83357f0a1b9df9f5e66817a2f4d4a18c8f4
--- /dev/null
+++ b/records/sd1/sweep/6e160e79e1310877d46b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e160e79e1310877d46b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.030813,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.046367,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.013344,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.006495,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.744,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00591,
+   "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": 7.117,
+   "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": 0.1271,
+   "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/6e160e79e1310877d46b/record.json b/records/sd1/sweep/6e160e79e1310877d46b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..57a2fb1ed0b3afc6c861d65d5e96d7a7923288b7
--- /dev/null
+++ b/records/sd1/sweep/6e160e79e1310877d46b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6e160e79e1310877d46b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.transformer_blocks.0.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9ed2c03838780864f97296de2854e05c1756fa608e3643353f0e03bd2cf39c46.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9ed2c03838780864f97296de2854e05c1756fa608e3643353f0e03bd2cf39c46",
+  "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": "4bd07a34bd3f7a7a7f93"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e160e79e1310877d46b/workflow.json b/records/sd1/sweep/6e160e79e1310877d46b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..26605f1c7da6485d0c6715a7331b2f1b04eaa0d5
--- /dev/null
+++ b/records/sd1/sweep/6e160e79e1310877d46b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e160e79e1310877d46b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.attn2.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/6e1e0af2015c3894e93e/measurements.json b/records/sd1/sweep/6e1e0af2015c3894e93e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..13a7dd16bb56607849f68c54a2d029c0226eae57
--- /dev/null
+++ b/records/sd1/sweep/6e1e0af2015c3894e93e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e1e0af2015c3894e93e",
+ "values": {
+  "std": {
+   "value": 53.415,
+   "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"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.716,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.216575,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.558157,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.803044,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.364775,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6878,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8046,
+   "method": "CLIP zero-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/6e1e0af2015c3894e93e/record.json b/records/sd1/sweep/6e1e0af2015c3894e93e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0941abe3912557d0767f063915db8b0614fc91d
--- /dev/null
+++ b/records/sd1/sweep/6e1e0af2015c3894e93e/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "6e1e0af2015c3894e93e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:22:51Z",
+ "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": "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": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "slight",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.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 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": "3d85824d7a4425835aab"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e1e0af2015c3894e93e/workflow.json b/records/sd1/sweep/6e1e0af2015c3894e93e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3572b0414939e1e654e23ece533dabeb747c433c
--- /dev/null
+++ b/records/sd1/sweep/6e1e0af2015c3894e93e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e1e0af2015c3894e93e)"
+  },
+  "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\": 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/6e2a3904cb7754a9be5d/measurements.json b/records/sd1/sweep/6e2a3904cb7754a9be5d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac223e4bdeff8e94d5457cc8cb5dcf1de56ee282
--- /dev/null
+++ b/records/sd1/sweep/6e2a3904cb7754a9be5d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e2a3904cb7754a9be5d",
+ "values": {
+  "std": {
+   "value": 50.427,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01036,
+   "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": 4.601,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.015145,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.041225,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.019646,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007179,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0105,
+   "method": "CLIP zero-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/6e2a3904cb7754a9be5d/record.json b/records/sd1/sweep/6e2a3904cb7754a9be5d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0d50d192e95fdd42ef5f2a9fa1c3d95674d70c9
--- /dev/null
+++ b/records/sd1/sweep/6e2a3904cb7754a9be5d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6e2a3904cb7754a9be5d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:40:08Z",
+ "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 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": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "mid",
+   "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": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4e440e2afe0744133635"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e2a3904cb7754a9be5d/workflow.json b/records/sd1/sweep/6e2a3904cb7754a9be5d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8404bd80713d8a7313db49a77d05dda268a8495
--- /dev/null
+++ b/records/sd1/sweep/6e2a3904cb7754a9be5d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e2a3904cb7754a9be5d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/6e388f24c90a987383f5/measurements.json b/records/sd1/sweep/6e388f24c90a987383f5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1300baba626f482f59dd25a08c95e30de52ffa05
--- /dev/null
+++ b/records/sd1/sweep/6e388f24c90a987383f5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e388f24c90a987383f5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.288608,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.438346,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.228436,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.106723,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 38.349,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00507,
+   "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": 34.563,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9704,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3654,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e388f24c90a987383f5/record.json b/records/sd1/sweep/6e388f24c90a987383f5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f52deb2a84aa7f14f7726fda86e8147e94f7e01c
--- /dev/null
+++ b/records/sd1/sweep/6e388f24c90a987383f5/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6e388f24c90a987383f5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:14Z",
+ "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": "output_blocks.5.1.transformer_blocks.0.ff.net.0.proj",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.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/4d563fdda85fdf430e3f7a368a1643db2f569daeae65ab437b09327a532c38ab.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4d563fdda85fdf430e3f7a368a1643db2f569daeae65ab437b09327a532c38ab",
+  "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": "afe387c88e2a3711a23c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e388f24c90a987383f5/workflow.json b/records/sd1/sweep/6e388f24c90a987383f5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..35a3578f8d222bb2ccab87b073501d08eb6607fc
--- /dev/null
+++ b/records/sd1/sweep/6e388f24c90a987383f5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e388f24c90a987383f5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.transformer_blocks.0.ff.net.0.proj\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e3ba5f09bf1fccd27c7/measurements.json b/records/sd1/sweep/6e3ba5f09bf1fccd27c7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6b3252d07765002263e0977a6d5a0e3cd6a1c5e
--- /dev/null
+++ b/records/sd1/sweep/6e3ba5f09bf1fccd27c7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e3ba5f09bf1fccd27c7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.055821,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.078358,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.026401,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.014577,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.086,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00523,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.855,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9759,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1639,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e3ba5f09bf1fccd27c7/record.json b/records/sd1/sweep/6e3ba5f09bf1fccd27c7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbdad2d97b1719d5ed77b11bc32499a56326453e
--- /dev/null
+++ b/records/sd1/sweep/6e3ba5f09bf1fccd27c7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6e3ba5f09bf1fccd27c7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:06Z",
+ "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.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e86b9293a3efadb258b4231a4ad00aee0e1b7e6100ad3e856f0530bc8c2ab32f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e86b9293a3efadb258b4231a4ad00aee0e1b7e6100ad3e856f0530bc8c2ab32f",
+  "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": "f270ea1e080620eeb6d8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e3ba5f09bf1fccd27c7/workflow.json b/records/sd1/sweep/6e3ba5f09bf1fccd27c7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..49c794d67d7ea542038a2bc7fdf8f5f2e7a251ad
--- /dev/null
+++ b/records/sd1/sweep/6e3ba5f09bf1fccd27c7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e3ba5f09bf1fccd27c7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.skip_connection\",\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/6e4ee8fefd76221132b7/interpretations.jsonl b/records/sd1/sweep/6e4ee8fefd76221132b7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..3b6ba61e3a9090b87817b202c8b77aac421b4d62
--- /dev/null
+++ b/records/sd1/sweep/6e4ee8fefd76221132b7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6e4ee8fefd76221132b7"}, "kind": "caption", "value": "Figure with sketchy, scribbled face and loose hair on a faded green ground", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "162bdf579f94bc6b"}
+{"target": {"record": "6e4ee8fefd76221132b7"}, "kind": "change", "value": "The face turns into a crude pencil-like scrawl of features over a pale smear while the dark shirt and hair remain.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d189b38ee5bb9d93"}
+{"target": {"record": "6e4ee8fefd76221132b7"}, "kind": "keywords", "value": ["scribbled face", "sketchy features", "distorted portrait", "loose hair", "faded"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6e2fb33c0a1e60da"}
+{"target": {"record": "6e4ee8fefd76221132b7"}, "kind": "effect_tags", "value": ["distorted", "surreal", "washed_out", "fragmented"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "77627f2c4c9848fb"}
diff --git a/records/sd1/sweep/6e4ee8fefd76221132b7/measurements.json b/records/sd1/sweep/6e4ee8fefd76221132b7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bdcc194dbb4f9b9aaf9f74b4410ea7f9a7bd684f
--- /dev/null
+++ b/records/sd1/sweep/6e4ee8fefd76221132b7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e4ee8fefd76221132b7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.225611,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.390653,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.352069,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.277044,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.184,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00428,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.435,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9515,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6663,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e4ee8fefd76221132b7/record.json b/records/sd1/sweep/6e4ee8fefd76221132b7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c69d32aad0daf0ea2dcf5bbcb5fbae04ad155078
--- /dev/null
+++ b/records/sd1/sweep/6e4ee8fefd76221132b7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e4ee8fefd76221132b7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/764ae350a380106df3a2989baa1c7ae0f7f1f91e26e0e3028b71eaeb33c5b664.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "764ae350a380106df3a2989baa1c7ae0f7f1f91e26e0e3028b71eaeb33c5b664",
+  "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": "98d1462049b27812dcca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e4ee8fefd76221132b7/workflow.json b/records/sd1/sweep/6e4ee8fefd76221132b7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a8014bfa5afb94d6ece7888fba56658d394f1af
--- /dev/null
+++ b/records/sd1/sweep/6e4ee8fefd76221132b7/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 6e4ee8fefd76221132b7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.in_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/6e54178be76fe7f61d3d/interpretations.jsonl b/records/sd1/sweep/6e54178be76fe7f61d3d/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..b11b921505b97b776caeb9419e3f6d43396cb3e4
--- /dev/null
+++ b/records/sd1/sweep/6e54178be76fe7f61d3d/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6e54178be76fe7f61d3d"}, "kind": "caption", "value": "Soft-focus close-up of a pale person with a glossy brown bowl haircut", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f9fd71d4e47bba32"}
+{"target": {"record": "6e54178be76fe7f61d3d"}, "kind": "change", "value": "The portrait becomes a closer, slightly blurred frontal face of an androgynous person with a sleek bowl cut and muted cool tones.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a8642297d2e64102"}
+{"target": {"record": "6e54178be76fe7f61d3d"}, "kind": "keywords", "value": ["bowl haircut", "close-up face", "soft focus", "androgynous", "muted tones"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a699ea33d9323a87"}
+{"target": {"record": "6e54178be76fe7f61d3d"}, "kind": "effect_tags", "value": ["subject_kept", "zoom", "blur", "colour_shift", "new_composition"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:25:54Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e066830b62d35437"}
diff --git a/records/sd1/sweep/6e54178be76fe7f61d3d/measurements.json b/records/sd1/sweep/6e54178be76fe7f61d3d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d43b47927238a10ffeba4451b1334dd67e14484e
--- /dev/null
+++ b/records/sd1/sweep/6e54178be76fe7f61d3d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e54178be76fe7f61d3d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.461725,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.523298,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.295023,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.222004,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.471,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00306,
+   "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": 48.657,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8636,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8363,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e54178be76fe7f61d3d/record.json b/records/sd1/sweep/6e54178be76fe7f61d3d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3cb4900c3ded203f22aec93656c98b04d93084ea
--- /dev/null
+++ b/records/sd1/sweep/6e54178be76fe7f61d3d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e54178be76fe7f61d3d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/1f6981dfca274dbea37ddf98eeae8063dee02e1686730b4812b43fbf247c9327.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1f6981dfca274dbea37ddf98eeae8063dee02e1686730b4812b43fbf247c9327",
+  "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": "188c6d9c741053f16781"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e54178be76fe7f61d3d/workflow.json b/records/sd1/sweep/6e54178be76fe7f61d3d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..141d928e04a7a20ea82a1b8d70acabcd42b911f7
--- /dev/null
+++ b/records/sd1/sweep/6e54178be76fe7f61d3d/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 6e54178be76fe7f61d3d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.proj_in\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e5801b4d93c353c7f17/measurements.json b/records/sd1/sweep/6e5801b4d93c353c7f17/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0653cf62083632a408c4cbfd12b37c15310357f
--- /dev/null
+++ b/records/sd1/sweep/6e5801b4d93c353c7f17/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e5801b4d93c353c7f17",
+ "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/6e5801b4d93c353c7f17/record.json b/records/sd1/sweep/6e5801b4d93c353c7f17/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..aef624fd84e812f558e4dbf0c82f8bfe12e6b267
--- /dev/null
+++ b/records/sd1/sweep/6e5801b4d93c353c7f17/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6e5801b4d93c353c7f17",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:19Z",
+ "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": "output_blocks.11.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.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/7d07f402afad73ff88edb07aed31eb0ee56d0637fe915be093353e05fa897a82.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7d07f402afad73ff88edb07aed31eb0ee56d0637fe915be093353e05fa897a82",
+  "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": "e1460edab47968ef69f2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e5801b4d93c353c7f17/workflow.json b/records/sd1/sweep/6e5801b4d93c353c7f17/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..11a168d332247f75dd027f73bfa97106654cdafb
--- /dev/null
+++ b/records/sd1/sweep/6e5801b4d93c353c7f17/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e5801b4d93c353c7f17)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/6e5d253ba290ace3feb6/interpretations.jsonl b/records/sd1/sweep/6e5d253ba290ace3feb6/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..20ea01fed98a5f0e37c5991e6f86e197a07b1cc9
--- /dev/null
+++ b/records/sd1/sweep/6e5d253ba290ace3feb6/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6e5d253ba290ace3feb6"}, "kind": "caption", "value": "Smiling young woman with chin-length bob and red lips on pale green backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:32Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3932b13244055ad3"}
+{"target": {"record": "6e5d253ba290ace3feb6"}, "kind": "change", "value": "The subject gains a fuller bob, brighter skin and red lips with a slight smile, the frame border vanishes and the backdrop becomes lighter and cleaner.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:32Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d2eeb220d721da98"}
+{"target": {"record": "6e5d253ba290ace3feb6"}, "kind": "keywords", "value": ["bob haircut", "red lips", "bright portrait", "borderless", "smile"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:32Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1050b2e300f71d00"}
+{"target": {"record": "6e5d253ba290ace3feb6"}, "kind": "effect_tags", "value": ["subject_kept", "bright", "new_composition"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:14:32Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fbf8f14600a49c8d"}
diff --git a/records/sd1/sweep/6e5d253ba290ace3feb6/measurements.json b/records/sd1/sweep/6e5d253ba290ace3feb6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf669d832a29b27eba3441d496d486f8abf1b83f
--- /dev/null
+++ b/records/sd1/sweep/6e5d253ba290ace3feb6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e5d253ba290ace3feb6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.415303,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.398896,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.132341,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.099575,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.556,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00453,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "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": 48.036,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9935,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3949,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e5d253ba290ace3feb6/record.json b/records/sd1/sweep/6e5d253ba290ace3feb6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6edd70edccf0ae44b8b94cd0a474a8018d8e026f
--- /dev/null
+++ b/records/sd1/sweep/6e5d253ba290ace3feb6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e5d253ba290ace3feb6",
+ "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.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "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/860c51ebb17dde0c60167268acd5dcbb44a537647f8320f4c40cd382dc50db7c.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "860c51ebb17dde0c60167268acd5dcbb44a537647f8320f4c40cd382dc50db7c",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "171190aaa7f7e8ce2e4b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e5d253ba290ace3feb6/workflow.json b/records/sd1/sweep/6e5d253ba290ace3feb6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba910c9c3baa2afc272e783d863d28f95c98f04f
--- /dev/null
+++ b/records/sd1/sweep/6e5d253ba290ace3feb6/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 6e5d253ba290ace3feb6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.out_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e63114a0ca7e141149e/measurements.json b/records/sd1/sweep/6e63114a0ca7e141149e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..263bd4542c023b72f9061d110f457bccf335964c
--- /dev/null
+++ b/records/sd1/sweep/6e63114a0ca7e141149e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e63114a0ca7e141149e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.229837,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.230053,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.132357,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.123239,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.927,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00632,
+   "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": 25.619,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8598,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3482,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e63114a0ca7e141149e/record.json b/records/sd1/sweep/6e63114a0ca7e141149e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f0886690a353c76d6b175829b4dbd6ab9e29e19
--- /dev/null
+++ b/records/sd1/sweep/6e63114a0ca7e141149e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e63114a0ca7e141149e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.mid",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/9f3f2ae54ced65060de04e8868c848b450fcc8fcb7c6dbf606cd53dd42a5c960.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9f3f2ae54ced65060de04e8868c848b450fcc8fcb7c6dbf606cd53dd42a5c960",
+  "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": "ec678a516e0248bee7c0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e63114a0ca7e141149e/workflow.json b/records/sd1/sweep/6e63114a0ca7e141149e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8757ea27b4b3f712196134158f639a8216b818a
--- /dev/null
+++ b/records/sd1/sweep/6e63114a0ca7e141149e/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 6e63114a0ca7e141149e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e6328498a4374a45c2e/measurements.json b/records/sd1/sweep/6e6328498a4374a45c2e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c1f4ee036c99fe7fc9061dc83ef523b471b719e
--- /dev/null
+++ b/records/sd1/sweep/6e6328498a4374a45c2e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e6328498a4374a45c2e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.147144,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.195186,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.10219,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.083232,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.15,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00485,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.355,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9683,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3805,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e6328498a4374a45c2e/record.json b/records/sd1/sweep/6e6328498a4374a45c2e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..df5ad3d3316a5dc7c5b6554afbb5e7c5db49a6f0
--- /dev/null
+++ b/records/sd1/sweep/6e6328498a4374a45c2e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e6328498a4374a45c2e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e5406b997bc76afd71e45c9d001c14727484c6e978e315d27ba0e56d5467c916.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e5406b997bc76afd71e45c9d001c14727484c6e978e315d27ba0e56d5467c916",
+  "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": "f0d5ae7a78273ed1a9fb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e6328498a4374a45c2e/workflow.json b/records/sd1/sweep/6e6328498a4374a45c2e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0e86d5b9f908513eacc1abd868f07faf1426c7f
--- /dev/null
+++ b/records/sd1/sweep/6e6328498a4374a45c2e/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 6e6328498a4374a45c2e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.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/6e6eddabd13f9e74c9fd/measurements.json b/records/sd1/sweep/6e6eddabd13f9e74c9fd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..85a0399f54baafad8d2c65030c37220809d5f661
--- /dev/null
+++ b/records/sd1/sweep/6e6eddabd13f9e74c9fd/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e6eddabd13f9e74c9fd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.236281,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.282945,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.127832,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.131227,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 74.085,
+   "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": 30.712,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9508,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6135,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e6eddabd13f9e74c9fd/record.json b/records/sd1/sweep/6e6eddabd13f9e74c9fd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e7acbfa804a95a977fc405c26af038fb30f5dc4
--- /dev/null
+++ b/records/sd1/sweep/6e6eddabd13f9e74c9fd/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e6eddabd13f9e74c9fd",
+ "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.3.1.proj_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.lo",
+   "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/86167132d98bbf229a1299e7b7181b1357fd272f3f659daeec8a150cff0215e8.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "86167132d98bbf229a1299e7b7181b1357fd272f3f659daeec8a150cff0215e8",
+  "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": "45cf8928de80d6b9d3c6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e6eddabd13f9e74c9fd/workflow.json b/records/sd1/sweep/6e6eddabd13f9e74c9fd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f66cec0e40b901097b10d72ef3d167a6ba317404
--- /dev/null
+++ b/records/sd1/sweep/6e6eddabd13f9e74c9fd/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 6e6eddabd13f9e74c9fd)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.proj_out\",\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/6e7a87e05e1ba8623fff/measurements.json b/records/sd1/sweep/6e7a87e05e1ba8623fff/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0b5c3f849ef077b4d425ee21ecc14fec41e1352
--- /dev/null
+++ b/records/sd1/sweep/6e7a87e05e1ba8623fff/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6e7a87e05e1ba8623fff",
+ "values": {
+  "std": {
+   "value": 92.47,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00273,
+   "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": 82.242,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.761507,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.871817,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.607601,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.3896,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8144,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4446,
+   "method": "CLIP zero-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/6e7a87e05e1ba8623fff/record.json b/records/sd1/sweep/6e7a87e05e1ba8623fff/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a7603d154bd332de68a2396ba0dfdaea5b73dd4
--- /dev/null
+++ b/records/sd1/sweep/6e7a87e05e1ba8623fff/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "6e7a87e05e1ba8623fff",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:59:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a 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": 1.4
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "grow",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "scale",
+   1.4,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.95, 1.05] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "f7b803658e523b75b58e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e7a87e05e1ba8623fff/workflow.json b/records/sd1/sweep/6e7a87e05e1ba8623fff/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0c6cb3f083742df19da0039dc772b01a2ec3e49
--- /dev/null
+++ b/records/sd1/sweep/6e7a87e05e1ba8623fff/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6e7a87e05e1ba8623fff)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 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/6e80bd5f044168c54d7a/measurements.json b/records/sd1/sweep/6e80bd5f044168c54d7a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..29396135888d9b014b553de48ff9260ab3d109a7
--- /dev/null
+++ b/records/sd1/sweep/6e80bd5f044168c54d7a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6e80bd5f044168c54d7a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.492412,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.701559,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.610656,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.40656,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.974,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01034,
+   "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": 40.213,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0334,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3804,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e80bd5f044168c54d7a/record.json b/records/sd1/sweep/6e80bd5f044168c54d7a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e9c0f3b6f331e0dc8be9aa300fb53364bb56687
--- /dev/null
+++ b/records/sd1/sweep/6e80bd5f044168c54d7a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6e80bd5f044168c54d7a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:37Z",
+ "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.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.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/086e2e3c81ddb071067293b202298b9ae65be598a18a30d5f36839ccab40b15a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "086e2e3c81ddb071067293b202298b9ae65be598a18a30d5f36839ccab40b15a",
+  "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": "f3d9edae9fcf9b92ccd5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6e80bd5f044168c54d7a/workflow.json b/records/sd1/sweep/6e80bd5f044168c54d7a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..de33f7f2d37419799d1288f6986374103086b72e
--- /dev/null
+++ b/records/sd1/sweep/6e80bd5f044168c54d7a/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 6e80bd5f044168c54d7a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/6ea1317c9231e253c662/measurements.json b/records/sd1/sweep/6ea1317c9231e253c662/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c20966c6761370c87482b2590b34ddb2261e6c5
--- /dev/null
+++ b/records/sd1/sweep/6ea1317c9231e253c662/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6ea1317c9231e253c662",
+ "values": {
+  "std": {
+   "value": 71.39,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00443,
+   "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": 59.167,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.642236,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.679581,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.899602,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.446137,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.663,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4442,
+   "method": "CLIP zero-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/6ea1317c9231e253c662/record.json b/records/sd1/sweep/6ea1317c9231e253c662/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a36fafc0a83cdb64af27bbeccb5843423228325b
--- /dev/null
+++ b/records/sd1/sweep/6ea1317c9231e253c662/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "6ea1317c9231e253c662",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:31:30Z",
+ "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": 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": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "shrink",
+   "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": [
+   "out.hi",
+   "scale",
+   0.7,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.7,
+    "applied": 0.7,
+    "reason": "outside safe range [0.95, 1.05] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ea1317c9231e253c662/workflow.json b/records/sd1/sweep/6ea1317c9231e253c662/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2143aa557492fc69b7702ff415ff8a430a634dbe
--- /dev/null
+++ b/records/sd1/sweep/6ea1317c9231e253c662/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ea1317c9231e253c662)"
+  },
+  "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\": 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/6ea9f63203d6e2a7c11a/measurements.json b/records/sd1/sweep/6ea9f63203d6e2a7c11a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f94cf4e1cf66374acbe85d649b0e6200c9f24833
--- /dev/null
+++ b/records/sd1/sweep/6ea9f63203d6e2a7c11a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ea9f63203d6e2a7c11a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.236639,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.270375,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.095853,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.112383,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.129,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0042,
+   "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": 25.951,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.877,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3721,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ea9f63203d6e2a7c11a/record.json b/records/sd1/sweep/6ea9f63203d6e2a7c11a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba4974c4823db3cb7d702d6a978d20afe6d4af8f
--- /dev/null
+++ b/records/sd1/sweep/6ea9f63203d6e2a7c11a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6ea9f63203d6e2a7c11a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/124f1d0fff921f5a9e90ac8a2ac90563d23703a03b0dec2d26ce32b08ab233e4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "124f1d0fff921f5a9e90ac8a2ac90563d23703a03b0dec2d26ce32b08ab233e4",
+  "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": "18f900fb54ef7aa56a5c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ea9f63203d6e2a7c11a/workflow.json b/records/sd1/sweep/6ea9f63203d6e2a7c11a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..176ba76ff7ca12c38e08caa667dd00f608cd78fe
--- /dev/null
+++ b/records/sd1/sweep/6ea9f63203d6e2a7c11a/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 6ea9f63203d6e2a7c11a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6eba7d87e03970420846/interpretations.jsonl b/records/sd1/sweep/6eba7d87e03970420846/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..8e0fef3a2528be2aa9c14f11705e508bba35a00b
--- /dev/null
+++ b/records/sd1/sweep/6eba7d87e03970420846/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6eba7d87e03970420846"}, "kind": "caption", "value": "Frontal portrait of a woman with lighter wavy hair and black collared shirt", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2d73ec92d2fa28ce"}
+{"target": {"record": "6eba7d87e03970420846"}, "kind": "change", "value": "Very similar composition with lighter golden-brown hair, slightly younger softer face and a collared shirt replacing the turtleneck.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9541c4fa1fcc64d3"}
+{"target": {"record": "6eba7d87e03970420846"}, "kind": "keywords", "value": ["minor variation", "lighter hair", "collared shirt", "portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d48e8b3a4c6e3216"}
+{"target": {"record": "6eba7d87e03970420846"}, "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": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e9cbb21387fa244e"}
diff --git a/records/sd1/sweep/6eba7d87e03970420846/measurements.json b/records/sd1/sweep/6eba7d87e03970420846/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a79209932987a75ccea8f19c37ca33c1c5f08802
--- /dev/null
+++ b/records/sd1/sweep/6eba7d87e03970420846/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6eba7d87e03970420846",
+ "values": {
+  "cosine_distance": {
+   "value": 0.06523,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.099293,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.085441,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.024251,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.748,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00481,
+   "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": 10.585,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.938,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1613,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6eba7d87e03970420846/record.json b/records/sd1/sweep/6eba7d87e03970420846/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..06124995f43ffc649957e1dcd2c10f68d6795085
--- /dev/null
+++ b/records/sd1/sweep/6eba7d87e03970420846/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6eba7d87e03970420846",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7223e89a25a6801929fb5e1acaf3b84085dffac503faed36c544dc5cd06b4b57.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7223e89a25a6801929fb5e1acaf3b84085dffac503faed36c544dc5cd06b4b57",
+  "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": "a87323ebc0e9b9c114a4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6eba7d87e03970420846/workflow.json b/records/sd1/sweep/6eba7d87e03970420846/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b7c755c01402879a32d22acebc5ab4685c3c20e
--- /dev/null
+++ b/records/sd1/sweep/6eba7d87e03970420846/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6eba7d87e03970420846)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6edc14e02e8a798e057c/measurements.json b/records/sd1/sweep/6edc14e02e8a798e057c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1846b248d147738558fb2e173944c7310d9295af
--- /dev/null
+++ b/records/sd1/sweep/6edc14e02e8a798e057c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6edc14e02e8a798e057c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.187556,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.306315,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.433064,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.282088,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.513,
+   "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": 22.849,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.001,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3573,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6edc14e02e8a798e057c/record.json b/records/sd1/sweep/6edc14e02e8a798e057c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..038de62be4609f847e8e4c61c830a2a6d1e56b20
--- /dev/null
+++ b/records/sd1/sweep/6edc14e02e8a798e057c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6edc14e02e8a798e057c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.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/e2f5ea32e5342ef42c1042ace03cdf4343d2d0798375434500a695345c351a60.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e2f5ea32e5342ef42c1042ace03cdf4343d2d0798375434500a695345c351a60",
+  "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": "bae3a1a0665cdbef79cf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6edc14e02e8a798e057c/workflow.json b/records/sd1/sweep/6edc14e02e8a798e057c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f44d851d99831248997c56e1493124bd199d4c3
--- /dev/null
+++ b/records/sd1/sweep/6edc14e02e8a798e057c/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 6edc14e02e8a798e057c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/6edf77997e136850a593/measurements.json b/records/sd1/sweep/6edf77997e136850a593/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..29b6959a47c7d4ba50eb789b4d76108fae836a08
--- /dev/null
+++ b/records/sd1/sweep/6edf77997e136850a593/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6edf77997e136850a593",
+ "values": {
+  "cosine_distance": {
+   "value": 0.133063,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.215104,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.111478,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044403,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.275,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00331,
+   "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": 21.696,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0541,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1105,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6edf77997e136850a593/record.json b/records/sd1/sweep/6edf77997e136850a593/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e93a311c7efdfa78bdab9968acdd995b53276c2d
--- /dev/null
+++ b/records/sd1/sweep/6edf77997e136850a593/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6edf77997e136850a593",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.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/104afadb7f0462548199efc305043654c8cba6abc88f1383e7c7223ac164c1d2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "104afadb7f0462548199efc305043654c8cba6abc88f1383e7c7223ac164c1d2",
+  "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": "43d9f3541874c02557b7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6edf77997e136850a593/workflow.json b/records/sd1/sweep/6edf77997e136850a593/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c57de210cb2cc7ba3248018c438dd6bcdca00c3
--- /dev/null
+++ b/records/sd1/sweep/6edf77997e136850a593/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6edf77997e136850a593)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/6ee9b3df13bdd809a227/measurements.json b/records/sd1/sweep/6ee9b3df13bdd809a227/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f13c6d9772198c2dbba4e1b51b2c81919946b68
--- /dev/null
+++ b/records/sd1/sweep/6ee9b3df13bdd809a227/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ee9b3df13bdd809a227",
+ "values": {
+  "cosine_distance": {
+   "value": 0.228272,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.49017,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.380781,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.229159,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 42.457,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00472,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.667,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.071,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3596,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ee9b3df13bdd809a227/record.json b/records/sd1/sweep/6ee9b3df13bdd809a227/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..12c84de9682bf729b2d3ad630ee7e6f2e7a20083
--- /dev/null
+++ b/records/sd1/sweep/6ee9b3df13bdd809a227/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6ee9b3df13bdd809a227",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:15Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/13e54f21f6d0acc0498dbdf2da991ea95e0f2ca5cabec3c6daef1d21c878612e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "13e54f21f6d0acc0498dbdf2da991ea95e0f2ca5cabec3c6daef1d21c878612e",
+  "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": "13df484f05bfdf5f1ec1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ee9b3df13bdd809a227/workflow.json b/records/sd1/sweep/6ee9b3df13bdd809a227/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8c0da2e6277beaf18442b570dc63fe487e75ab95
--- /dev/null
+++ b/records/sd1/sweep/6ee9b3df13bdd809a227/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ee9b3df13bdd809a227)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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": [
+    "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/6ef99d4c28df4498a196/measurements.json b/records/sd1/sweep/6ef99d4c28df4498a196/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d019cb1e17b701272b0a3174df2c04821ac9c454
--- /dev/null
+++ b/records/sd1/sweep/6ef99d4c28df4498a196/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6ef99d4c28df4498a196",
+ "values": {
+  "cosine_distance": {
+   "value": 0.019951,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.043059,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.015149,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008856,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.445,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00546,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 6.272,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9975,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1138,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ef99d4c28df4498a196/record.json b/records/sd1/sweep/6ef99d4c28df4498a196/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b12dc89cf0ff5fd095cb3984a6c840642e48296b
--- /dev/null
+++ b/records/sd1/sweep/6ef99d4c28df4498a196/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6ef99d4c28df4498a196",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:25Z",
+ "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.3.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.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/6642f06ad495bde7980b8354b7594d2022476dd3b0d54295e21f02fbdfc66e10.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6642f06ad495bde7980b8354b7594d2022476dd3b0d54295e21f02fbdfc66e10",
+  "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": "8ad5b2d85cb3e92762c9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ef99d4c28df4498a196/workflow.json b/records/sd1/sweep/6ef99d4c28df4498a196/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0feac447e0464c4017410fbee549d5009c96418
--- /dev/null
+++ b/records/sd1/sweep/6ef99d4c28df4498a196/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ef99d4c28df4498a196)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.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/6efd6ff19c98ae692f0d/measurements.json b/records/sd1/sweep/6efd6ff19c98ae692f0d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8a1f6248991c73b889af2d1763412b212495c5c
--- /dev/null
+++ b/records/sd1/sweep/6efd6ff19c98ae692f0d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6efd6ff19c98ae692f0d",
+ "values": {
+  "std": {
+   "value": 45.835,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01343,
+   "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": 15.702,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.122326,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.162231,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.043627,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.029875,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9498,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2021,
+   "method": "CLIP zero-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/6efd6ff19c98ae692f0d/record.json b/records/sd1/sweep/6efd6ff19c98ae692f0d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..384f9de9a991894104e4a34e1225093751271e8e
--- /dev/null
+++ b/records/sd1/sweep/6efd6ff19c98ae692f0d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6efd6ff19c98ae692f0d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:25:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "damp",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "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": "c8a0dbe79624c8739277"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6efd6ff19c98ae692f0d/workflow.json b/records/sd1/sweep/6efd6ff19c98ae692f0d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..50682de5d591f130bbfe896222b0de9c045cd8a1
--- /dev/null
+++ b/records/sd1/sweep/6efd6ff19c98ae692f0d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6efd6ff19c98ae692f0d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 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/6efeb0e6c5d6e36e72cb/measurements.json b/records/sd1/sweep/6efeb0e6c5d6e36e72cb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7673ffd0408ded5903b5ae7afc5f7c8b0ca3dff
--- /dev/null
+++ b/records/sd1/sweep/6efeb0e6c5d6e36e72cb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6efeb0e6c5d6e36e72cb",
+ "values": {
+  "std": {
+   "value": 36.67,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01433,
+   "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": 45.862,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.446388,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.569494,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.565869,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.151933,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0794,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0139,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6efeb0e6c5d6e36e72cb/record.json b/records/sd1/sweep/6efeb0e6c5d6e36e72cb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba47aab0b49aa455739171c023c42680949d8766
--- /dev/null
+++ b/records/sd1/sweep/6efeb0e6c5d6e36e72cb/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6efeb0e6c5d6e36e72cb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:07:25Z",
+ "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 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": 90.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "rotate",
+   90.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "dca8d5d438fd3697d590"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6efeb0e6c5d6e36e72cb/workflow.json b/records/sd1/sweep/6efeb0e6c5d6e36e72cb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c31549c853c92a67c21081a2c4968820abad28c9
--- /dev/null
+++ b/records/sd1/sweep/6efeb0e6c5d6e36e72cb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6efeb0e6c5d6e36e72cb)"
+  },
+  "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}",
+   "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/6eff55eb9d08eaab986e/measurements.json b/records/sd1/sweep/6eff55eb9d08eaab986e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b7615e93a43b90c79057b3178a28d0c559315c6
--- /dev/null
+++ b/records/sd1/sweep/6eff55eb9d08eaab986e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6eff55eb9d08eaab986e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.159284,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.210217,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.097379,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.020244,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.606,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00662,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.253,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9713,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1359,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6eff55eb9d08eaab986e/record.json b/records/sd1/sweep/6eff55eb9d08eaab986e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5bd103e792c9bced95d707b8011f799a1c08a12
--- /dev/null
+++ b/records/sd1/sweep/6eff55eb9d08eaab986e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6eff55eb9d08eaab986e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.1.norm",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "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/115ebddf23c712121be64b7d86896f3ed373484638690fe7f9ec589f462c93c1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "115ebddf23c712121be64b7d86896f3ed373484638690fe7f9ec589f462c93c1",
+  "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": "bcf7d29c1a4485b573ff"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6eff55eb9d08eaab986e/workflow.json b/records/sd1/sweep/6eff55eb9d08eaab986e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ee79fa0c1f2b31c72fad6fe4e24e88696b54365
--- /dev/null
+++ b/records/sd1/sweep/6eff55eb9d08eaab986e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6eff55eb9d08eaab986e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.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": [
+    "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/6f002f33fae2ef0a8225/measurements.json b/records/sd1/sweep/6f002f33fae2ef0a8225/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5002ec6ef073f3f8b19dec61e9789853bb3a2c89
--- /dev/null
+++ b/records/sd1/sweep/6f002f33fae2ef0a8225/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f002f33fae2ef0a8225",
+ "values": {
+  "cosine_distance": {
+   "value": 0.072134,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.090389,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.056389,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023853,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.212,
+   "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": 12.231,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9965,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1504,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f002f33fae2ef0a8225/record.json b/records/sd1/sweep/6f002f33fae2ef0a8225/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..71c9b14a673f6c3c067c99800c26152b953d323b
--- /dev/null
+++ b/records/sd1/sweep/6f002f33fae2ef0a8225/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f002f33fae2ef0a8225",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:59Z",
+ "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.10.0.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": "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/244808f37262fe4fe85cc5f39dd9aab8b6916e319c573c3b69a1e56e030a0a5b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "244808f37262fe4fe85cc5f39dd9aab8b6916e319c573c3b69a1e56e030a0a5b",
+  "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": "63909d1014c5adc3a4b2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f002f33fae2ef0a8225/workflow.json b/records/sd1/sweep/6f002f33fae2ef0a8225/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2fa5cd39b971bb54c8fa9462630862526d8ae0d
--- /dev/null
+++ b/records/sd1/sweep/6f002f33fae2ef0a8225/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f002f33fae2ef0a8225)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.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/6f031ea18120da237b6a/measurements.json b/records/sd1/sweep/6f031ea18120da237b6a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ab74c9bdb06d5d3564138bd3361166a26ce28e0
--- /dev/null
+++ b/records/sd1/sweep/6f031ea18120da237b6a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f031ea18120da237b6a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.150074,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.138909,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.105459,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.150009,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.485,
+   "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": [
+    "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": 15.371,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9653,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.364,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f031ea18120da237b6a/record.json b/records/sd1/sweep/6f031ea18120da237b6a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3eab7c82cb041b666bf06c371570d9a94b60e70f
--- /dev/null
+++ b/records/sd1/sweep/6f031ea18120da237b6a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6f031ea18120da237b6a",
+ "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": "input_blocks.11.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.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/7c89fc0963887ce1ab606a59f4a832cc7f46f1767030bf96054add44c297bf9e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7c89fc0963887ce1ab606a59f4a832cc7f46f1767030bf96054add44c297bf9e",
+  "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": "b8d7beacc6b8365a8c7f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f031ea18120da237b6a/workflow.json b/records/sd1/sweep/6f031ea18120da237b6a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..584cd30d6ab485f6b632849cf546bbad450645e6
--- /dev/null
+++ b/records/sd1/sweep/6f031ea18120da237b6a/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 6f031ea18120da237b6a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    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\": 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/6f09dc0ca58fd8226c3f/measurements.json b/records/sd1/sweep/6f09dc0ca58fd8226c3f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..755eafe9131cfd7de7419d232bba37e0e88515aa
--- /dev/null
+++ b/records/sd1/sweep/6f09dc0ca58fd8226c3f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f09dc0ca58fd8226c3f",
+ "values": {
+  "std": {
+   "value": 66.729,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00284,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.194,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.57423,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.725304,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.891139,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.395443,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.626,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7731,
+   "method": "CLIP zero-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/6f09dc0ca58fd8226c3f/record.json b/records/sd1/sweep/6f09dc0ca58fd8226c3f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f6e415218e077e69e484fb8bf3b6a5d0e4148332
--- /dev/null
+++ b/records/sd1/sweep/6f09dc0ca58fd8226c3f/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "6f09dc0ca58fd8226c3f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:31:50Z",
+ "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": "input_blocks.2.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "scale",
+   1.4,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.95, 1.05] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "36b207c450972bf311eb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f09dc0ca58fd8226c3f/workflow.json b/records/sd1/sweep/6f09dc0ca58fd8226c3f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f92daaf1e75ef536050d3c035b112041309dc4ae
--- /dev/null
+++ b/records/sd1/sweep/6f09dc0ca58fd8226c3f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f09dc0ca58fd8226c3f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f0c70711f6da67fa8e4/measurements.json b/records/sd1/sweep/6f0c70711f6da67fa8e4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..de93fbf2f2ebab31606bf3bfc67336fa62d52901
--- /dev/null
+++ b/records/sd1/sweep/6f0c70711f6da67fa8e4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f0c70711f6da67fa8e4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.405759,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.586714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.514967,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.314718,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 40.881,
+   "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": 42.842,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.99,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2318,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f0c70711f6da67fa8e4/record.json b/records/sd1/sweep/6f0c70711f6da67fa8e4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e5fc39896ac377bbd1e7ba10e2e44513d73956f
--- /dev/null
+++ b/records/sd1/sweep/6f0c70711f6da67fa8e4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6f0c70711f6da67fa8e4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/2722d07de4fc33b5b8da5bed8e0e6336dc83961cb31bd2900ebfc74dbd69482d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2722d07de4fc33b5b8da5bed8e0e6336dc83961cb31bd2900ebfc74dbd69482d",
+  "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": "e358b09b5a66dbcc1401"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f0c70711f6da67fa8e4/workflow.json b/records/sd1/sweep/6f0c70711f6da67fa8e4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9265ad61bb8dbc9a2e3550587b7499e431e6008b
--- /dev/null
+++ b/records/sd1/sweep/6f0c70711f6da67fa8e4/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 6f0c70711f6da67fa8e4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/6f197a708f12c1b79c5a/measurements.json b/records/sd1/sweep/6f197a708f12c1b79c5a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..89daf69b6d69f4071988c31086a4b208a65be9db
--- /dev/null
+++ b/records/sd1/sweep/6f197a708f12c1b79c5a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f197a708f12c1b79c5a",
+ "values": {
+  "std": {
+   "value": 51.315,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01055,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 3.492,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.009525,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.024289,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.009348,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003684,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9996,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0108,
+   "method": "CLIP zero-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/6f197a708f12c1b79c5a/record.json b/records/sd1/sweep/6f197a708f12c1b79c5a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d719567e5bd6941d568a3acc9f424e9052dc111
--- /dev/null
+++ b/records/sd1/sweep/6f197a708f12c1b79c5a/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "6f197a708f12c1b79c5a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:41:09Z",
+ "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": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 1234
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "heavy",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "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": "3309fe98833bfdea9c3a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f197a708f12c1b79c5a/workflow.json b/records/sd1/sweep/6f197a708f12c1b79c5a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6082a9104217ca0907544252a47faf623d982cb6
--- /dev/null
+++ b/records/sd1/sweep/6f197a708f12c1b79c5a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f197a708f12c1b79c5a)"
+  },
+  "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\": 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/6f2eb4ea35958b0542d2/measurements.json b/records/sd1/sweep/6f2eb4ea35958b0542d2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..406a86f097a67fef57d1e004d17be88b39da6997
--- /dev/null
+++ b/records/sd1/sweep/6f2eb4ea35958b0542d2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f2eb4ea35958b0542d2",
+ "values": {
+  "std": {
+   "value": 90.033,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00163,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 91.94,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.897261,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.747515,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 1.005867,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.452009,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7021,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7472,
+   "method": "CLIP zero-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/6f2eb4ea35958b0542d2/record.json b/records/sd1/sweep/6f2eb4ea35958b0542d2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8bc35f6b825363d99aa495a0441b751b9433b5a0
--- /dev/null
+++ b/records/sd1/sweep/6f2eb4ea35958b0542d2/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "6f2eb4ea35958b0542d2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:59:41Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "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,
+   "all"
+  ],
+  "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": "aec932cb25fb1c64485f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f2eb4ea35958b0542d2/workflow.json b/records/sd1/sweep/6f2eb4ea35958b0542d2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c65a6a0a30a18580c3e3412f0ce76d933c5e27d
--- /dev/null
+++ b/records/sd1/sweep/6f2eb4ea35958b0542d2/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f2eb4ea35958b0542d2)"
+  },
+  "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}",
+   "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/6f382bbba4520f337d1a/measurements.json b/records/sd1/sweep/6f382bbba4520f337d1a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4f44a228126d10ba4241eed17ac5d847f22189f
--- /dev/null
+++ b/records/sd1/sweep/6f382bbba4520f337d1a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f382bbba4520f337d1a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.14916,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.327376,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.225253,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.092439,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.926,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00335,
+   "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": 20.176,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0237,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/6f382bbba4520f337d1a/record.json b/records/sd1/sweep/6f382bbba4520f337d1a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0d689e950b7179839c44d81cbe293fcf27b64a80
--- /dev/null
+++ b/records/sd1/sweep/6f382bbba4520f337d1a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f382bbba4520f337d1a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.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/1e9aee96aad4127d6dc714eb92b77adf0ed00d3f31d7bd74e0e95dc096d5fed9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1e9aee96aad4127d6dc714eb92b77adf0ed00d3f31d7bd74e0e95dc096d5fed9",
+  "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": "ee6008e7fdb1c4bbe22e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f382bbba4520f337d1a/workflow.json b/records/sd1/sweep/6f382bbba4520f337d1a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e996d49981abacea9e91abf2b87c82396ce5723
--- /dev/null
+++ b/records/sd1/sweep/6f382bbba4520f337d1a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f382bbba4520f337d1a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.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/6f463adede4104653d98/measurements.json b/records/sd1/sweep/6f463adede4104653d98/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..43a8e634d3f276c3c732b42c8089f4c9de7478fa
--- /dev/null
+++ b/records/sd1/sweep/6f463adede4104653d98/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f463adede4104653d98",
+ "values": {
+  "cosine_distance": {
+   "value": 0.218102,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.349256,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.553824,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.267739,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.238,
+   "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": 25.444,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0102,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5073,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f463adede4104653d98/record.json b/records/sd1/sweep/6f463adede4104653d98/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e04316a8401d80e2fccc51c7beb54199abfff4c2
--- /dev/null
+++ b/records/sd1/sweep/6f463adede4104653d98/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6f463adede4104653d98",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:15Z",
+ "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.6.0.in_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b3b90e563075cd1287905aca8d24a2dc66f955e479b7bbb52e8f9965d60d6615.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b3b90e563075cd1287905aca8d24a2dc66f955e479b7bbb52e8f9965d60d6615",
+  "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": "c5ffe91c7ac281f453d2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f463adede4104653d98/workflow.json b/records/sd1/sweep/6f463adede4104653d98/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f56fe8689a261e9c831fa99418a5af1aca5c998
--- /dev/null
+++ b/records/sd1/sweep/6f463adede4104653d98/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 6f463adede4104653d98)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.in_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f49e8925ff9a630990a/measurements.json b/records/sd1/sweep/6f49e8925ff9a630990a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..992f4e4d759a87b5415dfe9c8acf3e34e68d1d12
--- /dev/null
+++ b/records/sd1/sweep/6f49e8925ff9a630990a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f49e8925ff9a630990a",
+ "values": {
+  "std": {
+   "value": 65.946,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01692,
+   "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": 54.751,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.473389,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.567958,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.310623,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.076413,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0105,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0243,
+   "method": "CLIP zero-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/6f49e8925ff9a630990a/record.json b/records/sd1/sweep/6f49e8925ff9a630990a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a92e21679cfb1f2bcb72a2a1b097e84011aad2d
--- /dev/null
+++ b/records/sd1/sweep/6f49e8925ff9a630990a/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "6f49e8925ff9a630990a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:47:42Z",
+ "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": 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": "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": "a77bd19c2cc87d81e6d0",
+  "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": "fd03c20f337d847081a0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f49e8925ff9a630990a/workflow.json b/records/sd1/sweep/6f49e8925ff9a630990a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1fe6e1c2b54da51e237c09e3a0ed59fcb205a878
--- /dev/null
+++ b/records/sd1/sweep/6f49e8925ff9a630990a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f49e8925ff9a630990a)"
+  },
+  "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 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/6f4f19840739542b5191/measurements.json b/records/sd1/sweep/6f4f19840739542b5191/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e0b9794d5abd41b052bc4ba7ba7d336c81687f6
--- /dev/null
+++ b/records/sd1/sweep/6f4f19840739542b5191/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f4f19840739542b5191",
+ "values": {
+  "cosine_distance": {
+   "value": 0.148618,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.150158,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.086991,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.047772,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.589,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00409,
+   "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": 17.983,
+   "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": 0.2628,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f4f19840739542b5191/record.json b/records/sd1/sweep/6f4f19840739542b5191/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa470fddedf69d396306acc0d6b4b3124d0a776b
--- /dev/null
+++ b/records/sd1/sweep/6f4f19840739542b5191/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6f4f19840739542b5191",
+ "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.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "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/eb62ba846b2a6552d0e25ad6665fc7678a4a5593dc6657b880b8cdb5b90412b7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "eb62ba846b2a6552d0e25ad6665fc7678a4a5593dc6657b880b8cdb5b90412b7",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "10a7720f44e582997b21"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f4f19840739542b5191/workflow.json b/records/sd1/sweep/6f4f19840739542b5191/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8193fda352a6727b43f88f5c8207cea68641db4b
--- /dev/null
+++ b/records/sd1/sweep/6f4f19840739542b5191/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 6f4f19840739542b5191)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f56b1aeb2d9f49bacff/measurements.json b/records/sd1/sweep/6f56b1aeb2d9f49bacff/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1bcf5a2921977a715609f6dd79f70b1bf06551bf
--- /dev/null
+++ b/records/sd1/sweep/6f56b1aeb2d9f49bacff/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f56b1aeb2d9f49bacff",
+ "values": {
+  "cosine_distance": {
+   "value": 0.058066,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.067789,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.044915,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.013848,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.219,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00429,
+   "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": 10.549,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9569,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/6f56b1aeb2d9f49bacff/record.json b/records/sd1/sweep/6f56b1aeb2d9f49bacff/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e39a16e81b2378c6bb1d8509d0e8707b446775c
--- /dev/null
+++ b/records/sd1/sweep/6f56b1aeb2d9f49bacff/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f56b1aeb2d9f49bacff",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:47Z",
+ "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.0.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "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/fbbd5bc0ed880ff4bea79bc06123cde820f1a5882dc97e1c4508a2610e3ee417.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "fbbd5bc0ed880ff4bea79bc06123cde820f1a5882dc97e1c4508a2610e3ee417",
+  "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": "cca296bac8cd52be59f1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f56b1aeb2d9f49bacff/workflow.json b/records/sd1/sweep/6f56b1aeb2d9f49bacff/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2832d7b203a585c2aee27f9ad9ded95462aabc0b
--- /dev/null
+++ b/records/sd1/sweep/6f56b1aeb2d9f49bacff/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f56b1aeb2d9f49bacff)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.skip_connection\",\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/6f592cfa573b74254da6/measurements.json b/records/sd1/sweep/6f592cfa573b74254da6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..212cbfcd9041620359a0c36a9ccb3a6721c87387
--- /dev/null
+++ b/records/sd1/sweep/6f592cfa573b74254da6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f592cfa573b74254da6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.17131,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.164593,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.091692,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.039467,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.7,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00611,
+   "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": 21.239,
+   "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": 0.407,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f592cfa573b74254da6/record.json b/records/sd1/sweep/6f592cfa573b74254da6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..255146ad5efeeea56b592c57693f5f8eb4e46084
--- /dev/null
+++ b/records/sd1/sweep/6f592cfa573b74254da6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6f592cfa573b74254da6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:39Z",
+ "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": "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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/fadee6e054d87af4e7e741a4a08eeee898299639211ef9212f1e53e6f4a1b595.png",
+  "batch_id": "kb_rv_ops_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"
+    ],
+    "previous_id": "97b0b8bda8fff0820b85"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f592cfa573b74254da6/workflow.json b/records/sd1/sweep/6f592cfa573b74254da6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c791ef2bbeea0b70a42900fbadd4f0fc66584ce8
--- /dev/null
+++ b/records/sd1/sweep/6f592cfa573b74254da6/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 6f592cfa573b74254da6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    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": [
+    "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/6f5d76fd5d83676eb846/measurements.json b/records/sd1/sweep/6f5d76fd5d83676eb846/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..28e8936e85715c0a6d7feb4e679e776940399d3b
--- /dev/null
+++ b/records/sd1/sweep/6f5d76fd5d83676eb846/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f5d76fd5d83676eb846",
+ "values": {
+  "cosine_distance": {
+   "value": 0.08507,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.113377,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.125509,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.029056,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.585,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.005,
+   "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": 13.699,
+   "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": 0.146,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f5d76fd5d83676eb846/record.json b/records/sd1/sweep/6f5d76fd5d83676eb846/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..23802acf631ccf4dc303f05aa74c4ed6e82745e7
--- /dev/null
+++ b/records/sd1/sweep/6f5d76fd5d83676eb846/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f5d76fd5d83676eb846",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:07Z",
+ "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": "middle_block.2.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e4c19c0b653e3e3e0154a619ae34421eca08b2e9594156ff90147e5519661a27.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e4c19c0b653e3e3e0154a619ae34421eca08b2e9594156ff90147e5519661a27",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.GroupNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5bb75f5a7f94f3c6c9b4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f5d76fd5d83676eb846/workflow.json b/records/sd1/sweep/6f5d76fd5d83676eb846/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb3ecb4e85fe864e9ae4ab3089af9c381922f294
--- /dev/null
+++ b/records/sd1/sweep/6f5d76fd5d83676eb846/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f5d76fd5d83676eb846)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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": [
+    "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/6f72fbc05e25d8023e2f/measurements.json b/records/sd1/sweep/6f72fbc05e25d8023e2f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..661ac04971526812508f2530062a45ccf8c09896
--- /dev/null
+++ b/records/sd1/sweep/6f72fbc05e25d8023e2f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f72fbc05e25d8023e2f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.258757,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.300486,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.197635,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.047934,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.098,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00415,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.552,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9561,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0873,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f72fbc05e25d8023e2f/record.json b/records/sd1/sweep/6f72fbc05e25d8023e2f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d102264e1ac71eb5fc751445f0e95f7a6d43a0a5
--- /dev/null
+++ b/records/sd1/sweep/6f72fbc05e25d8023e2f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f72fbc05e25d8023e2f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "mid",
+   "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/eb77ec6cc0efa1542482be4598dc9cb0af91329767af9b9d8f83c8f5f1de500d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "eb77ec6cc0efa1542482be4598dc9cb0af91329767af9b9d8f83c8f5f1de500d",
+  "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": "b04b60d3680331f7c9bd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f72fbc05e25d8023e2f/workflow.json b/records/sd1/sweep/6f72fbc05e25d8023e2f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4924ad59d1704fa0fb7cab90748175c39b3956b0
--- /dev/null
+++ b/records/sd1/sweep/6f72fbc05e25d8023e2f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f72fbc05e25d8023e2f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.skip_connection\",\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/6f750d6cc7a21eb63c62/interpretations.jsonl b/records/sd1/sweep/6f750d6cc7a21eb63c62/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..af45de7b4cd8f9fdbfd037ad129214accb1482d7
--- /dev/null
+++ b/records/sd1/sweep/6f750d6cc7a21eb63c62/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "6f750d6cc7a21eb63c62"}, "kind": "caption", "value": "Illustrated portrait of a woman with straight dark hair, big brown eyes, black collared jacket", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "88703ef330565690"}
+{"target": {"record": "6f750d6cc7a21eb63c62"}, "kind": "change", "value": "The face becomes more illustrated and anime-like, with larger brown eyes, sleeker darker hair with flyaway strands, a black collared jacket, and a lighter grey backdrop.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e893dd536fe8eb00"}
+{"target": {"record": "6f750d6cc7a21eb63c62"}, "kind": "keywords", "value": ["illustrated portrait", "anime eyes", "dark straight hair", "black jacket", "grey backdrop"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b5e475e7af578874"}
+{"target": {"record": "6f750d6cc7a21eb63c62"}, "kind": "effect_tags", "value": ["style_change", "subject_kept", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:19:14Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "33f0ccb50fd42acf"}
diff --git a/records/sd1/sweep/6f750d6cc7a21eb63c62/measurements.json b/records/sd1/sweep/6f750d6cc7a21eb63c62/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..12f9ac1afaea7440002a3b0234bcbbeef985058c
--- /dev/null
+++ b/records/sd1/sweep/6f750d6cc7a21eb63c62/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f750d6cc7a21eb63c62",
+ "values": {
+  "cosine_distance": {
+   "value": 0.286188,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.365787,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.207058,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.061891,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.062,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00689,
+   "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": 33.734,
+   "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": 0.0975,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f750d6cc7a21eb63c62/record.json b/records/sd1/sweep/6f750d6cc7a21eb63c62/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..770d179572a256f7376ef14d299e328d7139f77f
--- /dev/null
+++ b/records/sd1/sweep/6f750d6cc7a21eb63c62/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f750d6cc7a21eb63c62",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:47Z",
+ "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.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "kind": "skip",
+   "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/7c4031e002352d036b9fcd162931c86c4f28fd51759afce454839fab70392a9f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7c4031e002352d036b9fcd162931c86c4f28fd51759afce454839fab70392a9f",
+  "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": "f857af6716fdd059cf43"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f750d6cc7a21eb63c62/workflow.json b/records/sd1/sweep/6f750d6cc7a21eb63c62/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..073041ba7233afd0c84913e97d43b1bbb2951a87
--- /dev/null
+++ b/records/sd1/sweep/6f750d6cc7a21eb63c62/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f750d6cc7a21eb63c62)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.skip_connection\",\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/6f790a2e3706fcc8dd74/measurements.json b/records/sd1/sweep/6f790a2e3706fcc8dd74/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d44dd0252d560cd7d0bb808f6b5b368da226652f
--- /dev/null
+++ b/records/sd1/sweep/6f790a2e3706fcc8dd74/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f790a2e3706fcc8dd74",
+ "values": {
+  "std": {
+   "value": 52.159,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01034,
+   "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": 29.914,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.200096,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.248277,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.12586,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.033734,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0236,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0282,
+   "method": "CLIP zero-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/6f790a2e3706fcc8dd74/record.json b/records/sd1/sweep/6f790a2e3706fcc8dd74/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0844c22f988a06185a3dcd6e25abaab667d51f38
--- /dev/null
+++ b/records/sd1/sweep/6f790a2e3706fcc8dd74/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6f790a2e3706fcc8dd74",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:47:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": 0.5
+   },
+   "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": "27f5baccf7c0d4f1de23",
+  "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": "f27989f9665b15a59db6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f790a2e3706fcc8dd74/workflow.json b/records/sd1/sweep/6f790a2e3706fcc8dd74/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab67e7fed4cc2fa2ce57891ac2aaabdef0b96951
--- /dev/null
+++ b/records/sd1/sweep/6f790a2e3706fcc8dd74/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f790a2e3706fcc8dd74)"
+  },
+  "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 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/6f79af0118e0becc2f15/measurements.json b/records/sd1/sweep/6f79af0118e0becc2f15/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbebf150cac34ca4775860b9590260e2de48fb53
--- /dev/null
+++ b/records/sd1/sweep/6f79af0118e0becc2f15/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f79af0118e0becc2f15",
+ "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/6f79af0118e0becc2f15/record.json b/records/sd1/sweep/6f79af0118e0becc2f15/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b2921564318857ca6cf52ff1d7c5ad19fead448
--- /dev/null
+++ b/records/sd1/sweep/6f79af0118e0becc2f15/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f79af0118e0becc2f15",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:02Z",
+ "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.2.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "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/3d8db81a7f05f5fb437d047c5150292ddcb1987ccf26a8888788f2426f50ab77.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3d8db81a7f05f5fb437d047c5150292ddcb1987ccf26a8888788f2426f50ab77",
+  "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": "d2e24c74520efc256199"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f79af0118e0becc2f15/workflow.json b/records/sd1/sweep/6f79af0118e0becc2f15/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..541ff8a1802716c113a28d58dc548dcc6ae6f28d
--- /dev/null
+++ b/records/sd1/sweep/6f79af0118e0becc2f15/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f79af0118e0becc2f15)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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/6f7a8e04a310e71335d8/measurements.json b/records/sd1/sweep/6f7a8e04a310e71335d8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9a6f8664a274c295c12bf26f0fe8725ace6b3d3
--- /dev/null
+++ b/records/sd1/sweep/6f7a8e04a310e71335d8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f7a8e04a310e71335d8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.341846,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.432631,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.242341,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.131064,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 41.212,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01143,
+   "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": 39.164,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9258,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0352,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f7a8e04a310e71335d8/record.json b/records/sd1/sweep/6f7a8e04a310e71335d8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cbb769d058d95431c8dc3a405bbb4a3d4821614
--- /dev/null
+++ b/records/sd1/sweep/6f7a8e04a310e71335d8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f7a8e04a310e71335d8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "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/22c26732da36c0d7a5495a7ffe1e6e1369d69cc337033a678ecadca4fabd31fa.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "22c26732da36c0d7a5495a7ffe1e6e1369d69cc337033a678ecadca4fabd31fa",
+  "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": "2a3c5d5516dc5ac7bded"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f7a8e04a310e71335d8/workflow.json b/records/sd1/sweep/6f7a8e04a310e71335d8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cb9a4d6dcaaa7574f99a965f6aaf18b8b7d0665
--- /dev/null
+++ b/records/sd1/sweep/6f7a8e04a310e71335d8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f7a8e04a310e71335d8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f82e895501bd01ade93/measurements.json b/records/sd1/sweep/6f82e895501bd01ade93/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e0a7033a1a1c34eded109a62d36d48a745d867f
--- /dev/null
+++ b/records/sd1/sweep/6f82e895501bd01ade93/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f82e895501bd01ade93",
+ "values": {
+  "std": {
+   "value": 37.658,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01576,
+   "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": 6.047,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.022713,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.069939,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.016643,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003194,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.011,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0138,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f82e895501bd01ade93/record.json b/records/sd1/sweep/6f82e895501bd01ade93/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8933a83acaf8e32a89da4ff455cc6cd87ba61c64
--- /dev/null
+++ b/records/sd1/sweep/6f82e895501bd01ade93/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "6f82e895501bd01ade93",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:10:22Z",
+ "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.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": "de1eccace874429a61dc",
+  "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": "ab5122f3fdea67301ab4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f82e895501bd01ade93/workflow.json b/records/sd1/sweep/6f82e895501bd01ade93/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea3a7ad122bd2a99dace2be045392932224c4acb
--- /dev/null
+++ b/records/sd1/sweep/6f82e895501bd01ade93/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f82e895501bd01ade93)"
+  },
+  "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 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/6f8a653e1943c09ac2fb/measurements.json b/records/sd1/sweep/6f8a653e1943c09ac2fb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..20ad5d1406e9cbff15aeefd7b3b5a12907a64d08
--- /dev/null
+++ b/records/sd1/sweep/6f8a653e1943c09ac2fb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f8a653e1943c09ac2fb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.180696,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.320827,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.230518,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.126329,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.013,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00438,
+   "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": 22.896,
+   "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": 0.0386,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f8a653e1943c09ac2fb/record.json b/records/sd1/sweep/6f8a653e1943c09ac2fb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3792d677bd89cd805eee5a6515678a700b8d655d
--- /dev/null
+++ b/records/sd1/sweep/6f8a653e1943c09ac2fb/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f8a653e1943c09ac2fb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/3fd915209f812491be6ecad244028829571b30decc4ac1dcd234d5907346445b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3fd915209f812491be6ecad244028829571b30decc4ac1dcd234d5907346445b",
+  "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": "ab083e41f892bae5242d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f8a653e1943c09ac2fb/workflow.json b/records/sd1/sweep/6f8a653e1943c09ac2fb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..650ef1209ff4ce6921ee1090b2dbe37ccc509bab
--- /dev/null
+++ b/records/sd1/sweep/6f8a653e1943c09ac2fb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f8a653e1943c09ac2fb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.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/6f905e9de4b78a4e3f74/measurements.json b/records/sd1/sweep/6f905e9de4b78a4e3f74/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb2c57b132ed69e2ed595ce6ab71a40e7e1f13ed
--- /dev/null
+++ b/records/sd1/sweep/6f905e9de4b78a4e3f74/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6f905e9de4b78a4e3f74",
+ "values": {
+  "cosine_distance": {
+   "value": 0.054617,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.100292,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.054983,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.017568,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.885,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00736,
+   "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": 13.754,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9804,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0909,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f905e9de4b78a4e3f74/record.json b/records/sd1/sweep/6f905e9de4b78a4e3f74/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2cc84fabeab5922a84d90ef01cdc305ddbba4275
--- /dev/null
+++ b/records/sd1/sweep/6f905e9de4b78a4e3f74/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6f905e9de4b78a4e3f74",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:56Z",
+ "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": 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_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "cross_attn",
+   "module_type": "Dropout",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/524ad6260d10e4cafaafc3ae8d61cd89cc5700649dafee690f71b83958a1e43f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "524ad6260d10e4cafaafc3ae8d61cd89cc5700649dafee690f71b83958a1e43f",
+  "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": "adacfc1cb18b429dd47d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f905e9de4b78a4e3f74/workflow.json b/records/sd1/sweep/6f905e9de4b78a4e3f74/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..eabb3c0e0ee8e493c0e7c636327426f32f133507
--- /dev/null
+++ b/records/sd1/sweep/6f905e9de4b78a4e3f74/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f905e9de4b78a4e3f74)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.transformer_blocks.0.attn2.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f906ca82e1ed0943ad0/measurements.json b/records/sd1/sweep/6f906ca82e1ed0943ad0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..46ce4bacb5fa65aef9e5f4f4f5787bc483459707
--- /dev/null
+++ b/records/sd1/sweep/6f906ca82e1ed0943ad0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f906ca82e1ed0943ad0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.436241,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.470822,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.33323,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.243491,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.43,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0055,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 50.305,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0312,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0109,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f906ca82e1ed0943ad0/record.json b/records/sd1/sweep/6f906ca82e1ed0943ad0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..17fd7d7578506f0eb12c19209becb0eadbd2c997
--- /dev/null
+++ b/records/sd1/sweep/6f906ca82e1ed0943ad0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6f906ca82e1ed0943ad0",
+ "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.4.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/05b7a9bd6fe2040e9465b1fb5e4db112421fee9da24edf8347b53a140234d157.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "05b7a9bd6fe2040e9465b1fb5e4db112421fee9da24edf8347b53a140234d157",
+  "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": "fd338e13f039e97b8e3d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f906ca82e1ed0943ad0/workflow.json b/records/sd1/sweep/6f906ca82e1ed0943ad0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2624a752a55aa5a2e93acfffa47d3a79c00a2c3e
--- /dev/null
+++ b/records/sd1/sweep/6f906ca82e1ed0943ad0/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 6f906ca82e1ed0943ad0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.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": [
+    "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/6f9ffcccfbbd79272980/measurements.json b/records/sd1/sweep/6f9ffcccfbbd79272980/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a4007983a258e4a329369a1294261c53876e0e4
--- /dev/null
+++ b/records/sd1/sweep/6f9ffcccfbbd79272980/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6f9ffcccfbbd79272980",
+ "values": {
+  "std": {
+   "value": 69.879,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00458,
+   "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": 61.364,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.642999,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.688431,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.770196,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.39927,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7687,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5213,
+   "method": "CLIP zero-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/6f9ffcccfbbd79272980/record.json b/records/sd1/sweep/6f9ffcccfbbd79272980/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3344d952f122b33047c4c23afb3af2b606c80c54
--- /dev/null
+++ b/records/sd1/sweep/6f9ffcccfbbd79272980/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "6f9ffcccfbbd79272980",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:23:59Z",
+ "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": "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": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "shrink",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "scale",
+   0.7,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.7,
+    "applied": 0.7,
+    "reason": "outside safe range [0.95, 1.05] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6f9ffcccfbbd79272980/workflow.json b/records/sd1/sweep/6f9ffcccfbbd79272980/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7226c0d3514f894ad39edaaae1cc7d877ad0c2d3
--- /dev/null
+++ b/records/sd1/sweep/6f9ffcccfbbd79272980/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6f9ffcccfbbd79272980)"
+  },
+  "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\": 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/6fb8e195b17c8e28a526/measurements.json b/records/sd1/sweep/6fb8e195b17c8e28a526/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6739f7bcd2cf654703f675be6e442b09c138cce
--- /dev/null
+++ b/records/sd1/sweep/6fb8e195b17c8e28a526/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6fb8e195b17c8e28a526",
+ "values": {
+  "cosine_distance": {
+   "value": 0.442474,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.590471,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.492962,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.260594,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.899,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00723,
+   "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": 45.525,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0675,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0215,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fb8e195b17c8e28a526/record.json b/records/sd1/sweep/6fb8e195b17c8e28a526/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a45a61e7a4301ff547b95451e45fc4045d82be86
--- /dev/null
+++ b/records/sd1/sweep/6fb8e195b17c8e28a526/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6fb8e195b17c8e28a526",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:55Z",
+ "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.10.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "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/7f609b941c8ffe0593e2f5fa05aaf2cc0fa1edb8e8d77291dbac04c6c7cb2199.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7f609b941c8ffe0593e2f5fa05aaf2cc0fa1edb8e8d77291dbac04c6c7cb2199",
+  "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": "c391e81bef6572762f3c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fb8e195b17c8e28a526/workflow.json b/records/sd1/sweep/6fb8e195b17c8e28a526/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a00a70789d7be14de1c0953877ec040c96193ec
--- /dev/null
+++ b/records/sd1/sweep/6fb8e195b17c8e28a526/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 6fb8e195b17c8e28a526)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.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/6fbc995bf1f8d5b6469a/measurements.json b/records/sd1/sweep/6fbc995bf1f8d5b6469a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2f13134053f84f7e419984ec27246611fe1f962
--- /dev/null
+++ b/records/sd1/sweep/6fbc995bf1f8d5b6469a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6fbc995bf1f8d5b6469a",
+ "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/6fbc995bf1f8d5b6469a/record.json b/records/sd1/sweep/6fbc995bf1f8d5b6469a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c21a6a0361462854f8909b8d98247baa05d4e0bf
--- /dev/null
+++ b/records/sd1/sweep/6fbc995bf1f8d5b6469a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6fbc995bf1f8d5b6469a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:51Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "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/12fc9d50b10d0a94878de93381bece8c8eea3e6b4227fd6e7338d9778663b789.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "12fc9d50b10d0a94878de93381bece8c8eea3e6b4227fd6e7338d9778663b789",
+  "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": "b15d305cc77f93d9aaaf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fbc995bf1f8d5b6469a/workflow.json b/records/sd1/sweep/6fbc995bf1f8d5b6469a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b2b95ff954bb22fb2143b9959294621a0cc14d8
--- /dev/null
+++ b/records/sd1/sweep/6fbc995bf1f8d5b6469a/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 6fbc995bf1f8d5b6469a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.h_upd\",\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/6fdef978bbc3fbcda301/measurements.json b/records/sd1/sweep/6fdef978bbc3fbcda301/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5dd6f1159126153ac87e9966d4c7d115373d3ae
--- /dev/null
+++ b/records/sd1/sweep/6fdef978bbc3fbcda301/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6fdef978bbc3fbcda301",
+ "values": {
+  "std": {
+   "value": 40.528,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0235,
+   "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": 38.812,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.344778,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.507378,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.163127,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.054136,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9631,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2583,
+   "method": "CLIP zero-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/6fdef978bbc3fbcda301/record.json b/records/sd1/sweep/6fdef978bbc3fbcda301/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a2f2d7ee873e67bed0bd3e5b92a2116c8ffc35b
--- /dev/null
+++ b/records/sd1/sweep/6fdef978bbc3fbcda301/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "6fdef978bbc3fbcda301",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:51:11Z",
+ "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": 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.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": "1bbda6ca1f7efc631dfb",
+  "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": "3f3200ec2ea57178cb75"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fdef978bbc3fbcda301/workflow.json b/records/sd1/sweep/6fdef978bbc3fbcda301/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..71c0d3ad1fdfd2950775687c1b1865e6901a4750
--- /dev/null
+++ b/records/sd1/sweep/6fdef978bbc3fbcda301/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6fdef978bbc3fbcda301)"
+  },
+  "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 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/6fe4d980b6dfe90310d2/measurements.json b/records/sd1/sweep/6fe4d980b6dfe90310d2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ffa8db6eb4a776e58129c6f74c4fb7408a35ef4b
--- /dev/null
+++ b/records/sd1/sweep/6fe4d980b6dfe90310d2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6fe4d980b6dfe90310d2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.239009,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.43576,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.453721,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.335753,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.225,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00407,
+   "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": 29.102,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9886,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.894,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fe4d980b6dfe90310d2/record.json b/records/sd1/sweep/6fe4d980b6dfe90310d2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d48f6424a424b25f5fd9a5da6c8ccfe85c87c01
--- /dev/null
+++ b/records/sd1/sweep/6fe4d980b6dfe90310d2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "6fe4d980b6dfe90310d2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.in_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/a9ecaa6f355c08d8aefc9f1a02e1a80a351edadcfa588b08a565532c10c2f635.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a9ecaa6f355c08d8aefc9f1a02e1a80a351edadcfa588b08a565532c10c2f635",
+  "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": "11abf93d726481bb72a9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fe4d980b6dfe90310d2/workflow.json b/records/sd1/sweep/6fe4d980b6dfe90310d2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f043153490ef0084f23cf33701938f19499913a5
--- /dev/null
+++ b/records/sd1/sweep/6fe4d980b6dfe90310d2/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 6fe4d980b6dfe90310d2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.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/6fea6ee8f21428832c55/measurements.json b/records/sd1/sweep/6fea6ee8f21428832c55/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..99d3a2df1c275db1abe5cb3a620b6a3bc82df71a
--- /dev/null
+++ b/records/sd1/sweep/6fea6ee8f21428832c55/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6fea6ee8f21428832c55",
+ "values": {
+  "cosine_distance": {
+   "value": 0.251692,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.478609,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.293163,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.152427,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.243,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00272,
+   "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": 29.62,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0278,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5622,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fea6ee8f21428832c55/record.json b/records/sd1/sweep/6fea6ee8f21428832c55/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..01e7570d294ca2042c93db6e8d42c348eb7eaf05
--- /dev/null
+++ b/records/sd1/sweep/6fea6ee8f21428832c55/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6fea6ee8f21428832c55",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:25Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/444b3bf584b18055dd345e2a047ad6141a0f0f3188accfa741b7818f09f5829c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "444b3bf584b18055dd345e2a047ad6141a0f0f3188accfa741b7818f09f5829c",
+  "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": "10315af706bc5b377643"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fea6ee8f21428832c55/workflow.json b/records/sd1/sweep/6fea6ee8f21428832c55/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..890f5fd6ff03958ab8f3bb46ea9cc91bd44251a8
--- /dev/null
+++ b/records/sd1/sweep/6fea6ee8f21428832c55/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6fea6ee8f21428832c55)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fea6f67f6ef0e762015/measurements.json b/records/sd1/sweep/6fea6f67f6ef0e762015/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..121e9340cafa7e54c5e9f1792c93b99108de26bd
--- /dev/null
+++ b/records/sd1/sweep/6fea6f67f6ef0e762015/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6fea6f67f6ef0e762015",
+ "values": {
+  "std": {
+   "value": 27.354,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03524,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.011,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.27961,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.454931,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.172399,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.032264,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9925,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3212,
+   "method": "CLIP zero-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/6fea6f67f6ef0e762015/record.json b/records/sd1/sweep/6fea6f67f6ef0e762015/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e630faaaf4aa6f9e848ca502dbf8e54573febeb
--- /dev/null
+++ b/records/sd1/sweep/6fea6f67f6ef0e762015/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "6fea6f67f6ef0e762015",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:24:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "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": "81ef55543b3b3c080780",
+  "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": "ab1e18f4f80dc66677bf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6fea6f67f6ef0e762015/workflow.json b/records/sd1/sweep/6fea6f67f6ef0e762015/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbaa20128a27c894f0f4ca7cbadc6a3bc77edf1e
--- /dev/null
+++ b/records/sd1/sweep/6fea6f67f6ef0e762015/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6fea6f67f6ef0e762015)"
+  },
+  "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 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/6ffc1ea0d2298945e0f2/measurements.json b/records/sd1/sweep/6ffc1ea0d2298945e0f2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2a03d2ee840b25d28349ba88ead0fcedf6fb25e
--- /dev/null
+++ b/records/sd1/sweep/6ffc1ea0d2298945e0f2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "6ffc1ea0d2298945e0f2",
+ "values": {
+  "std": {
+   "value": 54.267,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00867,
+   "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": 17.967,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.120654,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.160163,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.053303,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.020878,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0215,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2139,
+   "method": "CLIP zero-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/6ffc1ea0d2298945e0f2/record.json b/records/sd1/sweep/6ffc1ea0d2298945e0f2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6339291852941bc3aa3b1275788b7b582e239f8d
--- /dev/null
+++ b/records/sd1/sweep/6ffc1ea0d2298945e0f2/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "6ffc1ea0d2298945e0f2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:57:09Z",
+ "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": "output_blocks.10.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 1234
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "heavy"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_noise",
+   0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "cb93e3e92320c5269fa6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ffc1ea0d2298945e0f2/workflow.json b/records/sd1/sweep/6ffc1ea0d2298945e0f2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d0b3b7f39b30f88f6767555c8447379b4fd5a05f
--- /dev/null
+++ b/records/sd1/sweep/6ffc1ea0d2298945e0f2/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ffc1ea0d2298945e0f2)"
+  },
+  "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}",
+   "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/6ffd504dd10c35e64272/measurements.json b/records/sd1/sweep/6ffd504dd10c35e64272/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f1dc744126d13089411b64cf5ae3987989b1e2c
--- /dev/null
+++ b/records/sd1/sweep/6ffd504dd10c35e64272/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "6ffd504dd10c35e64272",
+ "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/6ffd504dd10c35e64272/record.json b/records/sd1/sweep/6ffd504dd10c35e64272/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..631710b9a3a44fda204ca5e37cae2a2fa2719737
--- /dev/null
+++ b/records/sd1/sweep/6ffd504dd10c35e64272/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "6ffd504dd10c35e64272",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/f5335406d8e7f1bb92d2986c24b14afbbbae89c16ad289df3b244f6a26cf2cf8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "f5335406d8e7f1bb92d2986c24b14afbbbae89c16ad289df3b244f6a26cf2cf8",
+  "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": "6a703084e3f91aed9b11"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/6ffd504dd10c35e64272/workflow.json b/records/sd1/sweep/6ffd504dd10c35e64272/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0110a8dbf5f574b3834e146586a5266db588eba
--- /dev/null
+++ b/records/sd1/sweep/6ffd504dd10c35e64272/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 6ffd504dd10c35e64272)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.h_upd\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/7013b34633276d3a91ee/measurements.json b/records/sd1/sweep/7013b34633276d3a91ee/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7837f592989dac745b24ecf0051aff764e5dca27
--- /dev/null
+++ b/records/sd1/sweep/7013b34633276d3a91ee/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7013b34633276d3a91ee",
+ "values": {
+  "cosine_distance": {
+   "value": 0.4399,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.530521,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.285708,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.201234,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 79.806,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00595,
+   "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": 55.23,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0425,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/7013b34633276d3a91ee/record.json b/records/sd1/sweep/7013b34633276d3a91ee/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c796b3b94b11e7931d2c7b08a0ff2f434e7add5c
--- /dev/null
+++ b/records/sd1/sweep/7013b34633276d3a91ee/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7013b34633276d3a91ee",
+ "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.2.0.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/560558e25b643eeb5dcc9e86e6e9f4ec93586be32b1166b9ffaf386df3b891e7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "560558e25b643eeb5dcc9e86e6e9f4ec93586be32b1166b9ffaf386df3b891e7",
+  "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": "960b1c63b9a054a6135e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7013b34633276d3a91ee/workflow.json b/records/sd1/sweep/7013b34633276d3a91ee/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ef3a79a357eb9515a75ba2ddb38fd7e5b15b597
--- /dev/null
+++ b/records/sd1/sweep/7013b34633276d3a91ee/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 7013b34633276d3a91ee)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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/701d6249910f55ec79da/measurements.json b/records/sd1/sweep/701d6249910f55ec79da/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfa60422091d54e4b02976ddf9d41cc7a825dc67
--- /dev/null
+++ b/records/sd1/sweep/701d6249910f55ec79da/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "701d6249910f55ec79da",
+ "values": {
+  "cosine_distance": {
+   "value": 0.06661,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.096271,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.07178,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023486,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.484,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00662,
+   "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": 12.197,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0517,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1283,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/701d6249910f55ec79da/record.json b/records/sd1/sweep/701d6249910f55ec79da/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9017674e1996b44de538a5a03454a7d9fee74f44
--- /dev/null
+++ b/records/sd1/sweep/701d6249910f55ec79da/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "701d6249910f55ec79da",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:31Z",
+ "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.6.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/55be150b578d27199cfdec1a7217f91bbb7e6dd34827174c8fd505548664c53a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "55be150b578d27199cfdec1a7217f91bbb7e6dd34827174c8fd505548664c53a",
+  "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": "07eb8da71e90a746f7fc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/701d6249910f55ec79da/workflow.json b/records/sd1/sweep/701d6249910f55ec79da/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ad90768d31fc52c21a6cf9046274a456041318ea
--- /dev/null
+++ b/records/sd1/sweep/701d6249910f55ec79da/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 701d6249910f55ec79da)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_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/701d79c7e2ca45170a46/measurements.json b/records/sd1/sweep/701d79c7e2ca45170a46/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa8679e36ac79a7077337ddbe925e5e1593c921a
--- /dev/null
+++ b/records/sd1/sweep/701d79c7e2ca45170a46/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "701d79c7e2ca45170a46",
+ "values": {
+  "cosine_distance": {
+   "value": 0.035454,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.064018,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.045442,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010714,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.825,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0051,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.306,
+   "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": 0.1177,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/701d79c7e2ca45170a46/record.json b/records/sd1/sweep/701d79c7e2ca45170a46/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..733a07fc4889c0ba0731fd00db8cda62b41b82cb
--- /dev/null
+++ b/records/sd1/sweep/701d79c7e2ca45170a46/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "701d79c7e2ca45170a46",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:47Z",
+ "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.5.1.proj_out",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.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/b8b34ae2004a541b32cababb14f3ef1aad26c2f48a9946aae34fb829c2c847b7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "b8b34ae2004a541b32cababb14f3ef1aad26c2f48a9946aae34fb829c2c847b7",
+  "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": "a13c586a300925c9ed05"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/701d79c7e2ca45170a46/workflow.json b/records/sd1/sweep/701d79c7e2ca45170a46/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..65859fcfa7c94f13b1a41af3ded030f85b308db1
--- /dev/null
+++ b/records/sd1/sweep/701d79c7e2ca45170a46/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 701d79c7e2ca45170a46)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.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/7027f95c382fccead7b6/measurements.json b/records/sd1/sweep/7027f95c382fccead7b6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..40b10626a233c93e3fbaec1449b6701ea2d13cce
--- /dev/null
+++ b/records/sd1/sweep/7027f95c382fccead7b6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7027f95c382fccead7b6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.40703,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.582465,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.517257,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.31667,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 40.93,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00451,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 42.895,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9778,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3019,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7027f95c382fccead7b6/record.json b/records/sd1/sweep/7027f95c382fccead7b6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8965541870ad8c47e73456dd9501cd9810d15b4
--- /dev/null
+++ b/records/sd1/sweep/7027f95c382fccead7b6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7027f95c382fccead7b6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "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/54267e192c7dbe4811e7b21e6653ed2795d01b207ff73af9a5e9ea33bbd5786c.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "54267e192c7dbe4811e7b21e6653ed2795d01b207ff73af9a5e9ea33bbd5786c",
+  "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": "cd56933d44ba00d49918"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7027f95c382fccead7b6/workflow.json b/records/sd1/sweep/7027f95c382fccead7b6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..622578e7e4d896ae47636aae1881a3a2e30478f6
--- /dev/null
+++ b/records/sd1/sweep/7027f95c382fccead7b6/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 7027f95c382fccead7b6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.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/70408947cbacd16e5cb3/interpretations.jsonl b/records/sd1/sweep/70408947cbacd16e5cb3/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..49eddbbe7868b357dde28c5057c1ccced521ef46
--- /dev/null
+++ b/records/sd1/sweep/70408947cbacd16e5cb3/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "70408947cbacd16e5cb3"}, "kind": "caption", "value": "Blurry soft-focus painted portrait of a woman with wavy brown hair", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:48:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "62be8070490b5111"}
+{"target": {"record": "70408947cbacd16e5cb3"}, "kind": "change", "value": "The crisp painted portrait is smeared into a strong blur that keeps the composition and colours but loses all fine detail.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:48:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c4d0f420b1e73e9e"}
+{"target": {"record": "70408947cbacd16e5cb3"}, "kind": "keywords", "value": ["blur", "soft focus", "painted portrait", "smeared detail"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:48:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "149b4d7b1b1692bb"}
+{"target": {"record": "70408947cbacd16e5cb3"}, "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": "2026-10-05T21:48:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2473c5ed8d9a148c"}
diff --git a/records/sd1/sweep/70408947cbacd16e5cb3/measurements.json b/records/sd1/sweep/70408947cbacd16e5cb3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..15cb38507741726f6b8955b8aa8f3088378b8413
--- /dev/null
+++ b/records/sd1/sweep/70408947cbacd16e5cb3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70408947cbacd16e5cb3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.187945,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.451246,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.222271,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.135527,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.555,
+   "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": 22.486,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0124,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7543,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70408947cbacd16e5cb3/record.json b/records/sd1/sweep/70408947cbacd16e5cb3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..633a2fd6da9af22ba251ed4ed43e15990bf6d641
--- /dev/null
+++ b/records/sd1/sweep/70408947cbacd16e5cb3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "70408947cbacd16e5cb3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:00Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/3732d434bde0ad4a45dbc7d6d734b84d9ae0b861bb4f0e8d5d1e7db892428e88.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3732d434bde0ad4a45dbc7d6d734b84d9ae0b861bb4f0e8d5d1e7db892428e88",
+  "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": "480115b68c203a6d14ab"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70408947cbacd16e5cb3/workflow.json b/records/sd1/sweep/70408947cbacd16e5cb3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9eff634879eaf555fa555e0eddc7290825e0be9c
--- /dev/null
+++ b/records/sd1/sweep/70408947cbacd16e5cb3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70408947cbacd16e5cb3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/70448302efa6d6aebade/measurements.json b/records/sd1/sweep/70448302efa6d6aebade/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..edda1838fee93c1ceaf8b65c8924fa87270ebe2f
--- /dev/null
+++ b/records/sd1/sweep/70448302efa6d6aebade/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70448302efa6d6aebade",
+ "values": {
+  "cosine_distance": {
+   "value": 0.230649,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.326409,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.112458,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.079916,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.682,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00546,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.409,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9516,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0694,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70448302efa6d6aebade/record.json b/records/sd1/sweep/70448302efa6d6aebade/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..155a4ee063bda7abd986fff82fb14c7682bb25e1
--- /dev/null
+++ b/records/sd1/sweep/70448302efa6d6aebade/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "70448302efa6d6aebade",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:36Z",
+ "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.11.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/41953a3fe01c9dc087152663389c842ee8476849f9c15398819478bf86ab72a2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "41953a3fe01c9dc087152663389c842ee8476849f9c15398819478bf86ab72a2",
+  "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": "f3de17c48a0ca3d13090"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70448302efa6d6aebade/workflow.json b/records/sd1/sweep/70448302efa6d6aebade/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb0d0254ba0c7ed8a5fa6ddda6f5f084833fd826
--- /dev/null
+++ b/records/sd1/sweep/70448302efa6d6aebade/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70448302efa6d6aebade)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.norm2\",\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/70495c4a1f5a0b9f4ab5/measurements.json b/records/sd1/sweep/70495c4a1f5a0b9f4ab5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ac8c92ab172290232411962f10b066c046ac881
--- /dev/null
+++ b/records/sd1/sweep/70495c4a1f5a0b9f4ab5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70495c4a1f5a0b9f4ab5",
+ "values": {
+  "std": {
+   "value": 41.039,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01585,
+   "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": 16.096,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.101594,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.314024,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.503214,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.207713,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9078,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0634,
+   "method": "CLIP zero-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/70495c4a1f5a0b9f4ab5/record.json b/records/sd1/sweep/70495c4a1f5a0b9f4ab5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eda8ebe6f4bfd1b20625c1c02309c4e488da0ef6
--- /dev/null
+++ b/records/sd1/sweep/70495c4a1f5a0b9f4ab5/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "70495c4a1f5a0b9f4ab5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:57:19Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.mid",
+   "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": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "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": "a7a0f0d58700f02b0c2a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70495c4a1f5a0b9f4ab5/workflow.json b/records/sd1/sweep/70495c4a1f5a0b9f4ab5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b0163be8537227a03e7a475b1b8d8ab7d07d793a
--- /dev/null
+++ b/records/sd1/sweep/70495c4a1f5a0b9f4ab5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70495c4a1f5a0b9f4ab5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"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/705b7c9c20f04dec0662/measurements.json b/records/sd1/sweep/705b7c9c20f04dec0662/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..236c93e7160623790f85a17797dd11c8a474c86b
--- /dev/null
+++ b/records/sd1/sweep/705b7c9c20f04dec0662/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "705b7c9c20f04dec0662",
+ "values": {
+  "cosine_distance": {
+   "value": 0.15144,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.24202,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.19318,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.042714,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.566,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00342,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.631,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0683,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1196,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/705b7c9c20f04dec0662/record.json b/records/sd1/sweep/705b7c9c20f04dec0662/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c87cc789b48e7f6d522e5e6e123fcbe79bebb55b
--- /dev/null
+++ b/records/sd1/sweep/705b7c9c20f04dec0662/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "705b7c9c20f04dec0662",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:09Z",
+ "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": "output_blocks.11.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.hi",
+   "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/34f3068111fb452c2cab2d2123a170fc8372ce97a0e22526e4933ebc6d145b71.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "34f3068111fb452c2cab2d2123a170fc8372ce97a0e22526e4933ebc6d145b71",
+  "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": "317eddcd586cdfd8d411"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/705b7c9c20f04dec0662/workflow.json b/records/sd1/sweep/705b7c9c20f04dec0662/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f139326cb945d0dc0b54eed017ebf0a262a82cb5
--- /dev/null
+++ b/records/sd1/sweep/705b7c9c20f04dec0662/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 705b7c9c20f04dec0662)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.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/70605bb8eb99348c7f0b/measurements.json b/records/sd1/sweep/70605bb8eb99348c7f0b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..927f43610e0a7dd502b8312ae022adaeee13cc3d
--- /dev/null
+++ b/records/sd1/sweep/70605bb8eb99348c7f0b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70605bb8eb99348c7f0b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.048214,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.062605,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.02969,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008941,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.541,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00583,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.897,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9936,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1594,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70605bb8eb99348c7f0b/record.json b/records/sd1/sweep/70605bb8eb99348c7f0b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd651bc7d4bb234d7b2a440903801c8a6dc9718f
--- /dev/null
+++ b/records/sd1/sweep/70605bb8eb99348c7f0b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "70605bb8eb99348c7f0b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:16Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ecc1fc4e0926a7dbe792bc6a1ee899e2b0d3f199e095308089900c5c5521ab15.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ecc1fc4e0926a7dbe792bc6a1ee899e2b0d3f199e095308089900c5c5521ab15",
+  "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": "8eb68327ffe1909cac38"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70605bb8eb99348c7f0b/workflow.json b/records/sd1/sweep/70605bb8eb99348c7f0b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..92a031198a6e39b25085a5a56fd6958a4ec5a7e2
--- /dev/null
+++ b/records/sd1/sweep/70605bb8eb99348c7f0b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70605bb8eb99348c7f0b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.norm2\",\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/7061ae6b4f7b589a769e/measurements.json b/records/sd1/sweep/7061ae6b4f7b589a769e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0482836e4f49f04fdf7e5ae7bb3c68f83f243df
--- /dev/null
+++ b/records/sd1/sweep/7061ae6b4f7b589a769e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7061ae6b4f7b589a769e",
+ "values": {
+  "std": {
+   "value": 46.425,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01784,
+   "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": 27.476,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.206947,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.276199,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.137461,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.033716,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0601,
+   "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"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/7061ae6b4f7b589a769e/record.json b/records/sd1/sweep/7061ae6b4f7b589a769e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e425794a142ef3de2537f0aaaf541575dcbb802
--- /dev/null
+++ b/records/sd1/sweep/7061ae6b4f7b589a769e/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7061ae6b4f7b589a769e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:21:53Z",
+ "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": "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": 0.5
+   },
+   "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": "13216623a742ccb32c5e",
+  "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": "e44d72ac5b7b29c4405e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7061ae6b4f7b589a769e/workflow.json b/records/sd1/sweep/7061ae6b4f7b589a769e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3432463d47b80b86e96950128ff253ec28677cc9
--- /dev/null
+++ b/records/sd1/sweep/7061ae6b4f7b589a769e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7061ae6b4f7b589a769e)"
+  },
+  "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 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/7063ec5937ac7667ddf0/measurements.json b/records/sd1/sweep/7063ec5937ac7667ddf0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..af7993181b96031a7d7d2fb5fef05dd757de9a25
--- /dev/null
+++ b/records/sd1/sweep/7063ec5937ac7667ddf0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7063ec5937ac7667ddf0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.046745,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.080004,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.050932,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023181,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.92,
+   "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": 11.148,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.2475,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7063ec5937ac7667ddf0/record.json b/records/sd1/sweep/7063ec5937ac7667ddf0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a5a0a2c5288a55d5a01cde6acd8a8a3de543929
--- /dev/null
+++ b/records/sd1/sweep/7063ec5937ac7667ddf0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7063ec5937ac7667ddf0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1.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/023ac9990c342af393c94d570f095388c25a9b42c21fe4e06cb40b80634e2bcb.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "023ac9990c342af393c94d570f095388c25a9b42c21fe4e06cb40b80634e2bcb",
+  "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": "07537fc5e80266b8bf0d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7063ec5937ac7667ddf0/workflow.json b/records/sd1/sweep/7063ec5937ac7667ddf0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a952eee61a66299b59d027879e7e86ec011c8b19
--- /dev/null
+++ b/records/sd1/sweep/7063ec5937ac7667ddf0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7063ec5937ac7667ddf0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.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/7066c07f1554b793c074/measurements.json b/records/sd1/sweep/7066c07f1554b793c074/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..05d84843ba016a3d144757c4010df87ef6ec2ccf
--- /dev/null
+++ b/records/sd1/sweep/7066c07f1554b793c074/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7066c07f1554b793c074",
+ "values": {
+  "cosine_distance": {
+   "value": 0.118194,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.152808,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.075586,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040298,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.071,
+   "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": [
+    "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": 19.7,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.034,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0762,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7066c07f1554b793c074/record.json b/records/sd1/sweep/7066c07f1554b793c074/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d3ffebbf0825e608136974ac23e4e15e8bcf5b6b
--- /dev/null
+++ b/records/sd1/sweep/7066c07f1554b793c074/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7066c07f1554b793c074",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:07Z",
+ "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.10.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.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/05403bf88792726008cdabb5ec5d1144ab988d0b8f0bd7f4f72a92b1a767a037.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "05403bf88792726008cdabb5ec5d1144ab988d0b8f0bd7f4f72a92b1a767a037",
+  "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": "334104dacff3c8c418c5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7066c07f1554b793c074/workflow.json b/records/sd1/sweep/7066c07f1554b793c074/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7395c0b82ebaff83b4bec3a6fedbeaf917761790
--- /dev/null
+++ b/records/sd1/sweep/7066c07f1554b793c074/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7066c07f1554b793c074)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.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/7070d43d640df6df7942/interpretations.jsonl b/records/sd1/sweep/7070d43d640df6df7942/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..651b502ae5c986f45dda188b91f36cff44fd9b92
--- /dev/null
+++ b/records/sd1/sweep/7070d43d640df6df7942/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7070d43d640df6df7942"}, "kind": "caption", "value": "White lighthouse on a rocky islet at sunset with golden sun and clouds", "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": "212c75233818fe26"}
+{"target": {"record": "7070d43d640df6df7942"}, "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": "14f705e9176c2007"}
+{"target": {"record": "7070d43d640df6df7942"}, "kind": "keywords", "value": ["lighthouse", "sunset", "rocky islet", "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": "f3a1ca9bd82b488f"}
+{"target": {"record": "7070d43d640df6df7942"}, "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": "491905d0a56a78bd"}
diff --git a/records/sd1/sweep/7070d43d640df6df7942/measurements.json b/records/sd1/sweep/7070d43d640df6df7942/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7cd1a6c38919605bc6e83434e2658d31f26fd89
--- /dev/null
+++ b/records/sd1/sweep/7070d43d640df6df7942/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7070d43d640df6df7942",
+ "values": {
+  "std": {
+   "value": 37.698,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01469,
+   "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": 0.697,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000349,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.001819,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.00069,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000207,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0016,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0144,
+   "method": "CLIP zero-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/7070d43d640df6df7942/record.json b/records/sd1/sweep/7070d43d640df6df7942/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5dd2925c63c2d80c43583655c91fa69cca7c9539
--- /dev/null
+++ b/records/sd1/sweep/7070d43d640df6df7942/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7070d43d640df6df7942",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:08:38Z",
+ "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": 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": 0.667,
+   "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": [
+   "mid",
+   "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": "7bc231285a03393ce202"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7070d43d640df6df7942/workflow.json b/records/sd1/sweep/7070d43d640df6df7942/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bd3d29d99e581fad51425ff45b655aa84a96632
--- /dev/null
+++ b/records/sd1/sweep/7070d43d640df6df7942/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7070d43d640df6df7942)"
+  },
+  "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\": 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/7090576eceb9fb17e371/measurements.json b/records/sd1/sweep/7090576eceb9fb17e371/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e6784b476f4cd83b68b5f30f24591d03beaaf94
--- /dev/null
+++ b/records/sd1/sweep/7090576eceb9fb17e371/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7090576eceb9fb17e371",
+ "values": {
+  "std": {
+   "value": 50.915,
+   "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": 0.519,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000149,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.000754,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000779,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000182,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9986,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0159,
+   "method": "CLIP zero-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/7090576eceb9fb17e371/record.json b/records/sd1/sweep/7090576eceb9fb17e371/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..90b37c0e55ae2c62cbd5f7890ad9c6a4f7e4d7f6
--- /dev/null
+++ b/records/sd1/sweep/7090576eceb9fb17e371/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7090576eceb9fb17e371",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:41:43Z",
+ "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 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_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "mid",
+   "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": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "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": "55aab3bfe75d3181e0f8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7090576eceb9fb17e371/workflow.json b/records/sd1/sweep/7090576eceb9fb17e371/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6621f029e528d9cf284d10676c4f4371e8e932b
--- /dev/null
+++ b/records/sd1/sweep/7090576eceb9fb17e371/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7090576eceb9fb17e371)"
+  },
+  "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\": 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/7093127ac020d22e3e08/measurements.json b/records/sd1/sweep/7093127ac020d22e3e08/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a47afdd3c3ceeaa0130d9c173b4f5db2aa88ed2
--- /dev/null
+++ b/records/sd1/sweep/7093127ac020d22e3e08/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7093127ac020d22e3e08",
+ "values": {
+  "std": {
+   "value": 50.24,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0121,
+   "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": 17.689,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.148622,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.240629,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.228838,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.059988,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0058,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.034,
+   "method": "CLIP zero-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/7093127ac020d22e3e08/record.json b/records/sd1/sweep/7093127ac020d22e3e08/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..11b7aae837dbab1f54d069a99e14286c11b3b327
--- /dev/null
+++ b/records/sd1/sweep/7093127ac020d22e3e08/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "7093127ac020d22e3e08",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:22:03Z",
+ "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.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "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": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "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/7093127ac020d22e3e08/workflow.json b/records/sd1/sweep/7093127ac020d22e3e08/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..00628c5d4c82905c6246fb505b5ed7a79d0c1386
--- /dev/null
+++ b/records/sd1/sweep/7093127ac020d22e3e08/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7093127ac020d22e3e08)"
+  },
+  "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\": 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/70971a88f765046da7da/interpretations.jsonl b/records/sd1/sweep/70971a88f765046da7da/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2bf5e712ce245e996a71768b527a03975840e9c3
--- /dev/null
+++ b/records/sd1/sweep/70971a88f765046da7da/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "70971a88f765046da7da"}, "kind": "caption", "value": "Blurry blotchy camouflage pattern in black, cream and muted teal", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "251595bdf8c13634"}
+{"target": {"record": "70971a88f765046da7da"}, "kind": "change", "value": "The still life collapses into a soft, high-contrast blob texture with no recognisable flowers or fruit.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a968e3f7a19d569a"}
+{"target": {"record": "70971a88f765046da7da"}, "kind": "keywords", "value": ["camouflage blobs", "collapsed still life", "blotchy texture", "abstract noise"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cfae7915479fb26b"}
+{"target": {"record": "70971a88f765046da7da"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "texture", "blur", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2ce2b25ca467c866"}
diff --git a/records/sd1/sweep/70971a88f765046da7da/measurements.json b/records/sd1/sweep/70971a88f765046da7da/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9cef0e78b80b7b17ebf6d59a6cedc228c0c986ed
--- /dev/null
+++ b/records/sd1/sweep/70971a88f765046da7da/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70971a88f765046da7da",
+ "values": {
+  "std": {
+   "value": 77.921,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00223,
+   "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": 76.634,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.822705,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.715921,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 1.00488,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.441523,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.5752,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8621,
+   "method": "CLIP zero-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/70971a88f765046da7da/record.json b/records/sd1/sweep/70971a88f765046da7da/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d4c0cdf492b9efeac6d6f22835bd94a7d5a6c6e
--- /dev/null
+++ b/records/sd1/sweep/70971a88f765046da7da/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "70971a88f765046da7da",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:22:36Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "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": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "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": "444772bcae3d300d3428"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70971a88f765046da7da/workflow.json b/records/sd1/sweep/70971a88f765046da7da/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2fe08a0289cc17784bf4ccbdfccb6861644c33fb
--- /dev/null
+++ b/records/sd1/sweep/70971a88f765046da7da/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70971a88f765046da7da)"
+  },
+  "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 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/709b724715d21b9974a8/interpretations.jsonl b/records/sd1/sweep/709b724715d21b9974a8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c3000ae257767c771acc822c66c1308ef4b421b8
--- /dev/null
+++ b/records/sd1/sweep/709b724715d21b9974a8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "709b724715d21b9974a8"}, "kind": "caption", "value": "Smooth, painted-looking face with dark bob on mint-green background, rounded arch frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4ec2f76740f6aa4d"}
+{"target": {"record": "709b724715d21b9974a8"}, "kind": "change", "value": "The photo becomes a soft, airbrushed, painting-like frontal face with a slightly flattened, uncanny look, set against mint green inside an arched frame.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "feb8a80b8d77a518"}
+{"target": {"record": "709b724715d21b9974a8"}, "kind": "keywords", "value": ["airbrushed face", "mint green", "arched frame", "painted portrait", "uncanny"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ecd45b023bbd77a3"}
+{"target": {"record": "709b724715d21b9974a8"}, "kind": "effect_tags", "value": ["painterly", "style_change", "colour_shift", "zoom", "surreal", "subject_kept", "blur"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:20:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9d0572e90aecaf25"}
diff --git a/records/sd1/sweep/709b724715d21b9974a8/measurements.json b/records/sd1/sweep/709b724715d21b9974a8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aad2057bd68b6a13c094d43fdb1276f8d0c358d3
--- /dev/null
+++ b/records/sd1/sweep/709b724715d21b9974a8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "709b724715d21b9974a8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.484765,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.500982,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.345816,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.226699,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 36.242,
+   "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": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.518,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9526,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.495,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/709b724715d21b9974a8/record.json b/records/sd1/sweep/709b724715d21b9974a8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7f3960b8d8468ed4a568022f55436b0b67f29896
--- /dev/null
+++ b/records/sd1/sweep/709b724715d21b9974a8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "709b724715d21b9974a8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:22Z",
+ "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.4.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "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/00ea0e857fe35db6853f37625c8be84c8e5424c0f94e11aebf78a7c3d0e05d76.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "00ea0e857fe35db6853f37625c8be84c8e5424c0f94e11aebf78a7c3d0e05d76",
+  "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": "a5eb9369700f0b035dca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/709b724715d21b9974a8/workflow.json b/records/sd1/sweep/709b724715d21b9974a8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5c0498833402727b9954ed888962d3c253ee913
--- /dev/null
+++ b/records/sd1/sweep/709b724715d21b9974a8/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 709b724715d21b9974a8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.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": [
+    "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/709d927b9c34ef32584c/measurements.json b/records/sd1/sweep/709d927b9c34ef32584c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea60bec4f1e3a1f1ae704d018a5766d845efe4e8
--- /dev/null
+++ b/records/sd1/sweep/709d927b9c34ef32584c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "709d927b9c34ef32584c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.015756,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.025045,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.011731,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004455,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.753,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00572,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 5.282,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9896,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1502,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/709d927b9c34ef32584c/record.json b/records/sd1/sweep/709d927b9c34ef32584c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..248886a78f3d3678339191aa476a1d06e08fb486
--- /dev/null
+++ b/records/sd1/sweep/709d927b9c34ef32584c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "709d927b9c34ef32584c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:14Z",
+ "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": "output_blocks.10.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7333dafc62a6916d67790b193a1bab91cbdfa2276c62970389578a6fb77ac3de.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7333dafc62a6916d67790b193a1bab91cbdfa2276c62970389578a6fb77ac3de",
+  "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": "39cef53e6f1cbdbb58d8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/709d927b9c34ef32584c/workflow.json b/records/sd1/sweep/709d927b9c34ef32584c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6d4c0bfed23c1f4ecda4823589cfebcda440c72
--- /dev/null
+++ b/records/sd1/sweep/709d927b9c34ef32584c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 709d927b9c34ef32584c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 1.5\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/70a3ef3d9fb33b7ec843/measurements.json b/records/sd1/sweep/70a3ef3d9fb33b7ec843/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a38d832b989dabbcbd23a15ded1e565b7a50c076
--- /dev/null
+++ b/records/sd1/sweep/70a3ef3d9fb33b7ec843/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70a3ef3d9fb33b7ec843",
+ "values": {
+  "std": {
+   "value": 25.928,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02325,
+   "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": 13.946,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.079067,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.280884,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.475584,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.092678,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.923,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2674,
+   "method": "CLIP zero-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/70a3ef3d9fb33b7ec843/record.json b/records/sd1/sweep/70a3ef3d9fb33b7ec843/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..44aa6348cc4be074c53741b1121445b8e8049476
--- /dev/null
+++ b/records/sd1/sweep/70a3ef3d9fb33b7ec843/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "70a3ef3d9fb33b7ec843",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:07:12Z",
+ "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": "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": 30.0
+   },
+   "group": "out.mid",
+   "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": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "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": "98ef7ac6a6395649d05d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70a3ef3d9fb33b7ec843/workflow.json b/records/sd1/sweep/70a3ef3d9fb33b7ec843/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4dc67a24032aaba8347c33852dd8935e3456632
--- /dev/null
+++ b/records/sd1/sweep/70a3ef3d9fb33b7ec843/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70a3ef3d9fb33b7ec843)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"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/70ab7203a9e277bb54a1/measurements.json b/records/sd1/sweep/70ab7203a9e277bb54a1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d17909b8a9399645ca6d1ca1451b35e0befaf153
--- /dev/null
+++ b/records/sd1/sweep/70ab7203a9e277bb54a1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70ab7203a9e277bb54a1",
+ "values": {
+  "std": {
+   "value": 44.26,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01697,
+   "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": 4.866,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.008343,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.033119,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.013428,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003996,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9744,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.088,
+   "method": "CLIP zero-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/70ab7203a9e277bb54a1/record.json b/records/sd1/sweep/70ab7203a9e277bb54a1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e5d1acd7b5c1862e87b0dc95ca8577ba2f97fd08
--- /dev/null
+++ b/records/sd1/sweep/70ab7203a9e277bb54a1/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "70ab7203a9e277bb54a1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:54:59Z",
+ "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": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "beefcde17175fafb8286"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70ab7203a9e277bb54a1/workflow.json b/records/sd1/sweep/70ab7203a9e277bb54a1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..acab91dcf86f216ceb6e22bcce049d9c1d0d835d
--- /dev/null
+++ b/records/sd1/sweep/70ab7203a9e277bb54a1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70ab7203a9e277bb54a1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/70b555c3a464325f17b2/interpretations.jsonl b/records/sd1/sweep/70b555c3a464325f17b2/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..87e143e325b04c498753f7786556eeb4a0ccf535
--- /dev/null
+++ b/records/sd1/sweep/70b555c3a464325f17b2/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "70b555c3a464325f17b2"}, "kind": "caption", "value": "Long brown hair veiling a face, framed by dark brushy painted edges", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c1d112980c229bb4"}
+{"target": {"record": "70b555c3a464325f17b2"}, "kind": "change", "value": "The face is hidden behind streaming brown hair inside a dark, brush-stroked border, giving a painterly, half-abstract image.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ce08f79e46b930d0"}
+{"target": {"record": "70b555c3a464325f17b2"}, "kind": "keywords", "value": ["hair over face", "painted border", "hidden face", "brushy edges", "moody"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cfd9bdb29e68b243"}
+{"target": {"record": "70b555c3a464325f17b2"}, "kind": "effect_tags", "value": ["painterly", "framing", "subject_lost", "texture", "surreal"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "733c71d6efb37e3a"}
diff --git a/records/sd1/sweep/70b555c3a464325f17b2/measurements.json b/records/sd1/sweep/70b555c3a464325f17b2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fd7f33ace4f16b6bf16dbecb715e5cda1a0a3c7
--- /dev/null
+++ b/records/sd1/sweep/70b555c3a464325f17b2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70b555c3a464325f17b2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.289351,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.490703,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.475118,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.326657,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.734,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00788,
+   "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": 35.692,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9602,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0733,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70b555c3a464325f17b2/record.json b/records/sd1/sweep/70b555c3a464325f17b2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..729beacd2648958b13eebc85df9b03aafaf9fcf4
--- /dev/null
+++ b/records/sd1/sweep/70b555c3a464325f17b2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "70b555c3a464325f17b2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "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/1e90a5fd58d3242f7bcbbc4c3d6873cd3dcc6ebc25bad0f6c3571600fe1a7aa4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1e90a5fd58d3242f7bcbbc4c3d6873cd3dcc6ebc25bad0f6c3571600fe1a7aa4",
+  "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": "a7362ec04e8d77a3e43e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70b555c3a464325f17b2/workflow.json b/records/sd1/sweep/70b555c3a464325f17b2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..894eaffdc8df34c0dbc4ed2c4529c827888a1043
--- /dev/null
+++ b/records/sd1/sweep/70b555c3a464325f17b2/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 70b555c3a464325f17b2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.in_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/70c23eb81a129d1f0bed/measurements.json b/records/sd1/sweep/70c23eb81a129d1f0bed/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2db1c1a1cc6fe2b65e93a8c441124edf9373a8fa
--- /dev/null
+++ b/records/sd1/sweep/70c23eb81a129d1f0bed/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70c23eb81a129d1f0bed",
+ "values": {
+  "cosine_distance": {
+   "value": 0.096952,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.159212,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.115986,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.094288,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.852,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00336,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.274,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0126,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4409,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70c23eb81a129d1f0bed/record.json b/records/sd1/sweep/70c23eb81a129d1f0bed/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b85e45c3c4bc4bf08c12dca6013980a7533ec2bc
--- /dev/null
+++ b/records/sd1/sweep/70c23eb81a129d1f0bed/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "70c23eb81a129d1f0bed",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5d13ebc0d912b8c4a28793f63e2a8c7df4a397f8150af3ed87a50688a72dd478.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5d13ebc0d912b8c4a28793f63e2a8c7df4a397f8150af3ed87a50688a72dd478",
+  "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": "d113d26f690664df06ac"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70c23eb81a129d1f0bed/workflow.json b/records/sd1/sweep/70c23eb81a129d1f0bed/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2da4fb7c333d3437582b5770e88e0c1719448598
--- /dev/null
+++ b/records/sd1/sweep/70c23eb81a129d1f0bed/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 70c23eb81a129d1f0bed)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70c3ecbbf4f3747e434d/measurements.json b/records/sd1/sweep/70c3ecbbf4f3747e434d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b63bff4f1048d06308b3a642a6274ab25988ab3
--- /dev/null
+++ b/records/sd1/sweep/70c3ecbbf4f3747e434d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70c3ecbbf4f3747e434d",
+ "values": {
+  "std": {
+   "value": 36.249,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01309,
+   "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": 34.31,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.404336,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.564517,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.251344,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.139992,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0289,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.004,
+   "method": "CLIP zero-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/70c3ecbbf4f3747e434d/record.json b/records/sd1/sweep/70c3ecbbf4f3747e434d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..54024fc9e52ad477858cda68e37449948d29992e
--- /dev/null
+++ b/records/sd1/sweep/70c3ecbbf4f3747e434d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "70c3ecbbf4f3747e434d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:49:48Z",
+ "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": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "shrink",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "scale",
+   0.7,
+   "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": "2705879aacdf77ead5a2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70c3ecbbf4f3747e434d/workflow.json b/records/sd1/sweep/70c3ecbbf4f3747e434d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5ebe7f47528f82e330c19e902e7411cc5c232e7
--- /dev/null
+++ b/records/sd1/sweep/70c3ecbbf4f3747e434d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70c3ecbbf4f3747e434d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\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 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/70d0dad81b4cee97fef6/measurements.json b/records/sd1/sweep/70d0dad81b4cee97fef6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..abafd56b3a6a2d921a6e8a8cab421fc3073c65f2
--- /dev/null
+++ b/records/sd1/sweep/70d0dad81b4cee97fef6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70d0dad81b4cee97fef6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.641678,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.754687,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.585565,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.347711,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 20.567,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00933,
+   "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": 57.44,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9935,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1131,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70d0dad81b4cee97fef6/record.json b/records/sd1/sweep/70d0dad81b4cee97fef6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..91d9273078e04ceb72b114bad253b8a64e8fd928
--- /dev/null
+++ b/records/sd1/sweep/70d0dad81b4cee97fef6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "70d0dad81b4cee97fef6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:53Z",
+ "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.1.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "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/5fa7fbeb4e69b73d686fbe617be2286267ab4b47349138759a06f7d5b04aee41.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5fa7fbeb4e69b73d686fbe617be2286267ab4b47349138759a06f7d5b04aee41",
+  "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": "268c00d929f1afbd9049"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70d0dad81b4cee97fef6/workflow.json b/records/sd1/sweep/70d0dad81b4cee97fef6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..12787603230b228e82ae51d45ad2ff83dd94e023
--- /dev/null
+++ b/records/sd1/sweep/70d0dad81b4cee97fef6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70d0dad81b4cee97fef6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.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/70d55575f96bd82bc899/measurements.json b/records/sd1/sweep/70d55575f96bd82bc899/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5e1d22918b778234d2639636b748c7ffb365812
--- /dev/null
+++ b/records/sd1/sweep/70d55575f96bd82bc899/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70d55575f96bd82bc899",
+ "values": {
+  "cosine_distance": {
+   "value": 0.448223,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.494855,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.361574,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.226618,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 77.912,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00253,
+   "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": 51.392,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9505,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4525,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70d55575f96bd82bc899/record.json b/records/sd1/sweep/70d55575f96bd82bc899/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4276b867fe641741b6621283b7930d6cc2beeb83
--- /dev/null
+++ b/records/sd1/sweep/70d55575f96bd82bc899/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "70d55575f96bd82bc899",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.ff.net.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "ff",
+   "module_type": "Dropout",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/8273bb6ae9418280eb16d2ab909d459bae0f6eaa3ee160bd096b7c1bfa3c7350.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8273bb6ae9418280eb16d2ab909d459bae0f6eaa3ee160bd096b7c1bfa3c7350",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.FeedForward.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "cd0f5943fb2cb11ff52b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70d55575f96bd82bc899/workflow.json b/records/sd1/sweep/70d55575f96bd82bc899/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..87a4f2e21e24fc4da454fb141f27874c7def1c53
--- /dev/null
+++ b/records/sd1/sweep/70d55575f96bd82bc899/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 70d55575f96bd82bc899)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.ff.net.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/70db5d0750f1bc2f197b/measurements.json b/records/sd1/sweep/70db5d0750f1bc2f197b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..429df05bb7acdd87b055d15a6aff5069050ce127
--- /dev/null
+++ b/records/sd1/sweep/70db5d0750f1bc2f197b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70db5d0750f1bc2f197b",
+ "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/70db5d0750f1bc2f197b/record.json b/records/sd1/sweep/70db5d0750f1bc2f197b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0e13cf2930f532b593355ea37906c84d438aecd
--- /dev/null
+++ b/records/sd1/sweep/70db5d0750f1bc2f197b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "70db5d0750f1bc2f197b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:54Z",
+ "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.7.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.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/26c1228eb06fd01cf7efe835739e6bf142bb028526267fb57d9d6569e8a1b926.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "26c1228eb06fd01cf7efe835739e6bf142bb028526267fb57d9d6569e8a1b926",
+  "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": "9b288e4b5052670b1902"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70db5d0750f1bc2f197b/workflow.json b/records/sd1/sweep/70db5d0750f1bc2f197b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..da8a24266bee2a8ef18df941b994a940de71dc0a
--- /dev/null
+++ b/records/sd1/sweep/70db5d0750f1bc2f197b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70db5d0750f1bc2f197b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.h_upd\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70e59b7431636ae3034f/measurements.json b/records/sd1/sweep/70e59b7431636ae3034f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3219ee8d9f1be1e4b2f82bf7d510deeb5ddf7209
--- /dev/null
+++ b/records/sd1/sweep/70e59b7431636ae3034f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70e59b7431636ae3034f",
+ "values": {
+  "std": {
+   "value": 30.976,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02016,
+   "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": 13.103,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.032652,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.095211,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.043076,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.020469,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9884,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1966,
+   "method": "CLIP zero-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/70e59b7431636ae3034f/record.json b/records/sd1/sweep/70e59b7431636ae3034f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4239cdf28d6a1e67e82c046ce68cca1d07fbb215
--- /dev/null
+++ b/records/sd1/sweep/70e59b7431636ae3034f/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "70e59b7431636ae3034f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:12:00Z",
+ "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 city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.hi",
+   "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": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_scalar",
+   -0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70e59b7431636ae3034f/workflow.json b/records/sd1/sweep/70e59b7431636ae3034f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6754808ca3e17e9a95b2eadc66023d1d7cf21ec3
--- /dev/null
+++ b/records/sd1/sweep/70e59b7431636ae3034f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70e59b7431636ae3034f)"
+  },
+  "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\": 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/70e7ed895e98a7d01d57/interpretations.jsonl b/records/sd1/sweep/70e7ed895e98a7d01d57/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c0a5dd60470c7ba125361b1cf31569896531fcc5
--- /dev/null
+++ b/records/sd1/sweep/70e7ed895e98a7d01d57/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "70e7ed895e98a7d01d57"}, "kind": "caption", "value": "Young woman with ponytail looking over shoulder at viewer, aged photo border", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e67139d9068d899d"}
+{"target": {"record": "70e7ed895e98a7d01d57"}, "kind": "change", "value": "The subject turns away and looks back over her shoulder with a ponytail, inside a similar aged photo frame and palette.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3dcb304799fb0cb1"}
+{"target": {"record": "70e7ed895e98a7d01d57"}, "kind": "keywords", "value": ["over the shoulder", "ponytail", "vintage frame", "portrait pose", "muted tones"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "63d2f9935893d305"}
+{"target": {"record": "70e7ed895e98a7d01d57"}, "kind": "effect_tags", "value": ["subject_kept", "new_composition", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f6ff71d80e831e28"}
diff --git a/records/sd1/sweep/70e7ed895e98a7d01d57/measurements.json b/records/sd1/sweep/70e7ed895e98a7d01d57/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d366179b0748a75c69b0bbf1ea00d8059c1ab88
--- /dev/null
+++ b/records/sd1/sweep/70e7ed895e98a7d01d57/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "70e7ed895e98a7d01d57",
+ "values": {
+  "cosine_distance": {
+   "value": 0.334867,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.349153,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.147058,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.133042,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.181,
+   "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": 35.817,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0031,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.243,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70e7ed895e98a7d01d57/record.json b/records/sd1/sweep/70e7ed895e98a7d01d57/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f9d519201f1bfcc6573048a1b104e2632d205a9
--- /dev/null
+++ b/records/sd1/sweep/70e7ed895e98a7d01d57/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "70e7ed895e98a7d01d57",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:43Z",
+ "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": "output_blocks.0.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "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/aca29b1c24d6f86444cb98e2a6646103322efa42ab50e34749362025deafb0d0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "aca29b1c24d6f86444cb98e2a6646103322efa42ab50e34749362025deafb0d0",
+  "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": "05e0745f2bb919d9e1aa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70e7ed895e98a7d01d57/workflow.json b/records/sd1/sweep/70e7ed895e98a7d01d57/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a91ac60bf282d91129ba39977b04994a3ff3ab1
--- /dev/null
+++ b/records/sd1/sweep/70e7ed895e98a7d01d57/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 70e7ed895e98a7d01d57)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.in_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/70ec05ccdbff0816e637/measurements.json b/records/sd1/sweep/70ec05ccdbff0816e637/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e802153ce6db24a9400a689b469dca46b6d5ed25
--- /dev/null
+++ b/records/sd1/sweep/70ec05ccdbff0816e637/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70ec05ccdbff0816e637",
+ "values": {
+  "std": {
+   "value": 58.676,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00903,
+   "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": 14.096,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.089119,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.096716,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.026719,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.009495,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9913,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2994,
+   "method": "CLIP zero-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/70ec05ccdbff0816e637/record.json b/records/sd1/sweep/70ec05ccdbff0816e637/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c44ee18c9d2cbec2bf70196c3a03958526aeb28
--- /dev/null
+++ b/records/sd1/sweep/70ec05ccdbff0816e637/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "70ec05ccdbff0816e637",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:39:23Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "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",
+   "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": "2a8d43323cc5fd1b8ef1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70ec05ccdbff0816e637/workflow.json b/records/sd1/sweep/70ec05ccdbff0816e637/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0275f215157e3383b859422f1a488a55b7d3d2fc
--- /dev/null
+++ b/records/sd1/sweep/70ec05ccdbff0816e637/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70ec05ccdbff0816e637)"
+  },
+  "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}",
+   "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/70f1e83ff915e2e3f5f0/measurements.json b/records/sd1/sweep/70f1e83ff915e2e3f5f0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d0dfdec2398426aacaabe5b494be54a3086dcc1
--- /dev/null
+++ b/records/sd1/sweep/70f1e83ff915e2e3f5f0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "70f1e83ff915e2e3f5f0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.239834,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.307713,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.139325,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.050155,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.544,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00943,
+   "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": 27.826,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9031,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70f1e83ff915e2e3f5f0/record.json b/records/sd1/sweep/70f1e83ff915e2e3f5f0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..398bc6f15ca2fa5778cbf5609420dd1d19931921
--- /dev/null
+++ b/records/sd1/sweep/70f1e83ff915e2e3f5f0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "70f1e83ff915e2e3f5f0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:23Z",
+ "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.8.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.mid",
+   "kind": "self_attn",
+   "module_type": "Dropout",
+   "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/654815393d43e1a4f553d8adee8d77fc6bf838e52f2bc0b446c5d7e4647453cb.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "654815393d43e1a4f553d8adee8d77fc6bf838e52f2bc0b446c5d7e4647453cb",
+  "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": "b8f5471cbc122ca33f15"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/70f1e83ff915e2e3f5f0/workflow.json b/records/sd1/sweep/70f1e83ff915e2e3f5f0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..280b611133208caec7911674e9f8c9eb3cb2e05f
--- /dev/null
+++ b/records/sd1/sweep/70f1e83ff915e2e3f5f0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 70f1e83ff915e2e3f5f0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.transformer_blocks.0.attn1.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71022e935c3bdfced4cb/interpretations.jsonl b/records/sd1/sweep/71022e935c3bdfced4cb/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..55b2f90f33a53e8f1d449e14ea4a776f27882632
--- /dev/null
+++ b/records/sd1/sweep/71022e935c3bdfced4cb/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "71022e935c3bdfced4cb"}, "kind": "caption", "value": "Close-up of a young woman with long brown hair and red lips, blurred background", "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": "d3fbcce084812ddb"}
+{"target": {"record": "71022e935c3bdfced4cb"}, "kind": "change", "value": "The subject becomes a closer, softer-lit young woman with longer hair and red lips, and the background turns to shallow-focus blur.", "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": "51e9cd10a6a55aac"}
+{"target": {"record": "71022e935c3bdfced4cb"}, "kind": "keywords", "value": ["close-up portrait", "red lips", "shallow depth of field", "long hair", "soft focus"], "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": "5b05e032eac6dbb5"}
+{"target": {"record": "71022e935c3bdfced4cb"}, "kind": "effect_tags", "value": ["zoom", "blur", "subject_kept", "new_composition"], "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": "93836b2ce50b8f2c"}
diff --git a/records/sd1/sweep/71022e935c3bdfced4cb/measurements.json b/records/sd1/sweep/71022e935c3bdfced4cb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f0c681a4530566ed0160b003e24dc18bce14bb2
--- /dev/null
+++ b/records/sd1/sweep/71022e935c3bdfced4cb/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71022e935c3bdfced4cb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.253556,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.35998,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.342491,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.139806,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.667,
+   "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": 28.275,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9829,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4715,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71022e935c3bdfced4cb/record.json b/records/sd1/sweep/71022e935c3bdfced4cb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9324af61a06ce594f396629c20c6f56b10715d96
--- /dev/null
+++ b/records/sd1/sweep/71022e935c3bdfced4cb/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71022e935c3bdfced4cb",
+ "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": 60
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f2077e5c01ebeffb37b1b79a0d3804516cef658b6e26a959f734819f90f90da7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f2077e5c01ebeffb37b1b79a0d3804516cef658b6e26a959f734819f90f90da7",
+  "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": "60bef189f2a6b3a942e7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71022e935c3bdfced4cb/workflow.json b/records/sd1/sweep/71022e935c3bdfced4cb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd5c63899b398fd1e9cdaf4b3dcabfa89d8e201c
--- /dev/null
+++ b/records/sd1/sweep/71022e935c3bdfced4cb/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 71022e935c3bdfced4cb)"
+  },
+  "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\": 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/71029697400db1e8284c/measurements.json b/records/sd1/sweep/71029697400db1e8284c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5be2486e43443c1a740cb4ca7f48317c5d10654
--- /dev/null
+++ b/records/sd1/sweep/71029697400db1e8284c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71029697400db1e8284c",
+ "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/71029697400db1e8284c/record.json b/records/sd1/sweep/71029697400db1e8284c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f3d99a2207977a44887fedcd76912088534e71f
--- /dev/null
+++ b/records/sd1/sweep/71029697400db1e8284c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71029697400db1e8284c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:06Z",
+ "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.3.0.h_upd",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b7a999cf18d1a8d62a09c7a954a50e58299d2452a257776d7c97ca676780fa35.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b7a999cf18d1a8d62a09c7a954a50e58299d2452a257776d7c97ca676780fa35",
+  "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": "e187616c3c8104da4b22"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71029697400db1e8284c/workflow.json b/records/sd1/sweep/71029697400db1e8284c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b3ae731038e7b34a5e08ad8a35f727087065678
--- /dev/null
+++ b/records/sd1/sweep/71029697400db1e8284c/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 71029697400db1e8284c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.h_upd\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71060966b1ffe379ceab/measurements.json b/records/sd1/sweep/71060966b1ffe379ceab/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b6a97709d2ec2a720ef08c57339871dd85792c16
--- /dev/null
+++ b/records/sd1/sweep/71060966b1ffe379ceab/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71060966b1ffe379ceab",
+ "values": {
+  "cosine_distance": {
+   "value": 0.460005,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.499808,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.283716,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.098984,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 39.729,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00421,
+   "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": 43.124,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9593,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0726,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71060966b1ffe379ceab/record.json b/records/sd1/sweep/71060966b1ffe379ceab/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6f23bea807e036d71581cc04f73daf999347d9e8
--- /dev/null
+++ b/records/sd1/sweep/71060966b1ffe379ceab/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71060966b1ffe379ceab",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:22Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9365ba3a2dd2821d8cd314e26fe9ab758e6a127734da79ac35bca299a3f4141f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9365ba3a2dd2821d8cd314e26fe9ab758e6a127734da79ac35bca299a3f4141f",
+  "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": "99bec1827d21b0459271"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71060966b1ffe379ceab/workflow.json b/records/sd1/sweep/71060966b1ffe379ceab/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a7988fc49bebc4d1a76b0a58829a6e1779199ece
--- /dev/null
+++ b/records/sd1/sweep/71060966b1ffe379ceab/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71060966b1ffe379ceab)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/71191a8f69507aa58d6f/measurements.json b/records/sd1/sweep/71191a8f69507aa58d6f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b344bbdf7bc2c368b565a9b397d063728550261
--- /dev/null
+++ b/records/sd1/sweep/71191a8f69507aa58d6f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71191a8f69507aa58d6f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.420341,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.477242,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.657085,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.281221,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.24,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00547,
+   "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": 46.974,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9084,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3378,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71191a8f69507aa58d6f/record.json b/records/sd1/sweep/71191a8f69507aa58d6f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c6d36096d4cc791adf355fa303c33fde3a5d048
--- /dev/null
+++ b/records/sd1/sweep/71191a8f69507aa58d6f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71191a8f69507aa58d6f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:11Z",
+ "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.2.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/dd6123575065ae8b1af3d7ae1b0939b0863c3fcb5c467153fa480ba31bafd3b0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "dd6123575065ae8b1af3d7ae1b0939b0863c3fcb5c467153fa480ba31bafd3b0",
+  "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": "92513bdd002d65a0c2c4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71191a8f69507aa58d6f/workflow.json b/records/sd1/sweep/71191a8f69507aa58d6f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..41f1a5f62b7f8820b1df0112b512ddcfb709704f
--- /dev/null
+++ b/records/sd1/sweep/71191a8f69507aa58d6f/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 71191a8f69507aa58d6f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.in_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/711eb9da540a01c81534/measurements.json b/records/sd1/sweep/711eb9da540a01c81534/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..af1b67254be00cb77a24977b5a24bcc278d6234a
--- /dev/null
+++ b/records/sd1/sweep/711eb9da540a01c81534/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "711eb9da540a01c81534",
+ "values": {
+  "std": {
+   "value": 47.818,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0023,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "clip_degenerate"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 60.066,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.67494,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.824659,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.954663,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.414878,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6368,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9407,
+   "method": "CLIP zero-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/711eb9da540a01c81534/record.json b/records/sd1/sweep/711eb9da540a01c81534/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4ee826aa49f0ef3d104115fed947e87ed1f11b8c
--- /dev/null
+++ b/records/sd1/sweep/711eb9da540a01c81534/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "711eb9da540a01c81534",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:58:25Z",
+ "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 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": 2.0
+   },
+   "group": "in.hi",
+   "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": [
+   "in.hi",
+   "multiply",
+   2.0,
+   "all"
+  ],
+  "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": "70594c2ccd97fa38cb07"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/711eb9da540a01c81534/workflow.json b/records/sd1/sweep/711eb9da540a01c81534/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..60e4ad3768d7e82723b338b0c2c196057a3d6ddf
--- /dev/null
+++ b/records/sd1/sweep/711eb9da540a01c81534/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 711eb9da540a01c81534)"
+  },
+  "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}",
+   "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/7129ff9ae7aa4a636098/measurements.json b/records/sd1/sweep/7129ff9ae7aa4a636098/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cf066831ccf247817ab68db4d35acf002bd289b9
--- /dev/null
+++ b/records/sd1/sweep/7129ff9ae7aa4a636098/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7129ff9ae7aa4a636098",
+ "values": {
+  "cosine_distance": {
+   "value": 0.139208,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.173442,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.137324,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.139315,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.11,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00419,
+   "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": 17.136,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9506,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4506,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7129ff9ae7aa4a636098/record.json b/records/sd1/sweep/7129ff9ae7aa4a636098/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ab6b4c9c78a9c894ecbc383dbb85f80b627b7a4
--- /dev/null
+++ b/records/sd1/sweep/7129ff9ae7aa4a636098/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7129ff9ae7aa4a636098",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:50Z",
+ "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": "output_blocks.9.1.transformer_blocks.0.attn1.to_out.0",
+   "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": "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/c02a4e01d1ef6542c30292e7062e8b899d1e8908a433684eb74e334e97c46b3b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c02a4e01d1ef6542c30292e7062e8b899d1e8908a433684eb74e334e97c46b3b",
+  "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": "dccb7f64564193c12e3f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7129ff9ae7aa4a636098/workflow.json b/records/sd1/sweep/7129ff9ae7aa4a636098/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..388c306a5282e504927fe7a52fdc4d6066e70c41
--- /dev/null
+++ b/records/sd1/sweep/7129ff9ae7aa4a636098/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 7129ff9ae7aa4a636098)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.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/7133e19b0acc9d212ca7/measurements.json b/records/sd1/sweep/7133e19b0acc9d212ca7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..91c840c5e319214305a69a9f4cc6a36bdfa8118e
--- /dev/null
+++ b/records/sd1/sweep/7133e19b0acc9d212ca7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7133e19b0acc9d212ca7",
+ "values": {
+  "std": {
+   "value": 58.854,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00974,
+   "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": 4.273,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.009134,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.013635,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.01033,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002787,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0008,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3605,
+   "method": "CLIP zero-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/7133e19b0acc9d212ca7/record.json b/records/sd1/sweep/7133e19b0acc9d212ca7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a39b8017c136157102eb1551888ec1fbbd0633d
--- /dev/null
+++ b/records/sd1/sweep/7133e19b0acc9d212ca7/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7133e19b0acc9d212ca7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:36:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "in.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": [
+   "in.mid",
+   "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/7133e19b0acc9d212ca7/workflow.json b/records/sd1/sweep/7133e19b0acc9d212ca7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2205d341f3f830193b7649a81b02b5fa59cec293
--- /dev/null
+++ b/records/sd1/sweep/7133e19b0acc9d212ca7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7133e19b0acc9d212ca7)"
+  },
+  "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\": 0.2,\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/713f3390b5cc3f45ba61/interpretations.jsonl b/records/sd1/sweep/713f3390b5cc3f45ba61/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..edc8f9ac5cb90b649c5902402121c2be8a650da3
--- /dev/null
+++ b/records/sd1/sweep/713f3390b5cc3f45ba61/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "713f3390b5cc3f45ba61"}, "kind": "caption", "value": "Warm blotchy texture of cream, ochre, green and black patches", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7447779a459a7099"}
+{"target": {"record": "713f3390b5cc3f45ba61"}, "kind": "change", "value": "The still life collapses into a warm-toned camouflage of blurred blobs, keeping only a faint echo of its palette.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a06188c68988f908"}
+{"target": {"record": "713f3390b5cc3f45ba61"}, "kind": "keywords", "value": ["warm blotches", "camouflage texture", "collapsed still life", "abstract noise"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "da2a2c4a0bfc3e80"}
+{"target": {"record": "713f3390b5cc3f45ba61"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "texture", "blur", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:51:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "271c1b2e60be8399"}
diff --git a/records/sd1/sweep/713f3390b5cc3f45ba61/measurements.json b/records/sd1/sweep/713f3390b5cc3f45ba61/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea7e1a23109e548e1046026c2aed75d8ff378c85
--- /dev/null
+++ b/records/sd1/sweep/713f3390b5cc3f45ba61/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "713f3390b5cc3f45ba61",
+ "values": {
+  "std": {
+   "value": 75.642,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00347,
+   "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": 71.873,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.702416,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.738209,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.989307,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.450441,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6174,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8502,
+   "method": "CLIP zero-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/713f3390b5cc3f45ba61/record.json b/records/sd1/sweep/713f3390b5cc3f45ba61/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d9a52394b9da15d509eab97d74f5fa870a8e74a
--- /dev/null
+++ b/records/sd1/sweep/713f3390b5cc3f45ba61/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "713f3390b5cc3f45ba61",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:24:13Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.hi",
+   "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.hi",
+   "scale",
+   1.4,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.95, 1.05] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0318238066ec862b7d16"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/713f3390b5cc3f45ba61/workflow.json b/records/sd1/sweep/713f3390b5cc3f45ba61/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e85c1ccb2ceb786b1e49cab74d45bdcc1fca383a
--- /dev/null
+++ b/records/sd1/sweep/713f3390b5cc3f45ba61/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 713f3390b5cc3f45ba61)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ]\n}",
+   "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/714169e100e849a237c8/measurements.json b/records/sd1/sweep/714169e100e849a237c8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae7be07843d3a965f69b1a683487055eaa94cfae
--- /dev/null
+++ b/records/sd1/sweep/714169e100e849a237c8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "714169e100e849a237c8",
+ "values": {
+  "std": {
+   "value": 38.335,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00635,
+   "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": 33.817,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.398092,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.538999,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.585363,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.093917,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0102,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0676,
+   "method": "CLIP zero-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/714169e100e849a237c8/record.json b/records/sd1/sweep/714169e100e849a237c8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0471c7db25fdc915bcd6c0ea0dbc0bb02c11fae3
--- /dev/null
+++ b/records/sd1/sweep/714169e100e849a237c8/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "714169e100e849a237c8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:44Z",
+ "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 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": 0.7
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "shrink",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.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/714169e100e849a237c8/workflow.json b/records/sd1/sweep/714169e100e849a237c8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7213c53b85b45bf53b218f3c748f51fad2dbf6c
--- /dev/null
+++ b/records/sd1/sweep/714169e100e849a237c8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 714169e100e849a237c8)"
+  },
+  "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\": 0.7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71430be7a1b7bedbc362/measurements.json b/records/sd1/sweep/71430be7a1b7bedbc362/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c030f81a2db602d9a71429f52ab1f9d49fe38938
--- /dev/null
+++ b/records/sd1/sweep/71430be7a1b7bedbc362/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71430be7a1b7bedbc362",
+ "values": {
+  "std": {
+   "value": 27.963,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01001,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.003,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.60527,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.762247,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.84183,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.305999,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8123,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/71430be7a1b7bedbc362/record.json b/records/sd1/sweep/71430be7a1b7bedbc362/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0253c5cf13fc61ff2b0e80efe9fdb6609941fda0
--- /dev/null
+++ b/records/sd1/sweep/71430be7a1b7bedbc362/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "71430be7a1b7bedbc362",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:03:59Z",
+ "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": "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": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.mid",
+   "kind": "res_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": [
+   "in.mid",
+   "scale",
+   1.4,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.7, 1.3] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2c286982bc870a385e5a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71430be7a1b7bedbc362/workflow.json b/records/sd1/sweep/71430be7a1b7bedbc362/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0c2bbc206af5a1f8af492a7c66fce6c1a124eb32
--- /dev/null
+++ b/records/sd1/sweep/71430be7a1b7bedbc362/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71430be7a1b7bedbc362)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ]\n}",
+   "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/71470dd252966636150e/measurements.json b/records/sd1/sweep/71470dd252966636150e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e2d7cd36069d6577548a5433448c63e1138cb9f
--- /dev/null
+++ b/records/sd1/sweep/71470dd252966636150e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71470dd252966636150e",
+ "values": {
+  "std": {
+   "value": 39.013,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01639,
+   "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": 12.114,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.089685,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.167731,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.042304,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.018813,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.005,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0243,
+   "method": "CLIP zero-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/71470dd252966636150e/record.json b/records/sd1/sweep/71470dd252966636150e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..47d151cf3f04073907fd18a82036b946671848ee
--- /dev/null
+++ b/records/sd1/sweep/71470dd252966636150e/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "71470dd252966636150e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:06:46Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "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": "f99e313fa94b4365c483"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71470dd252966636150e/workflow.json b/records/sd1/sweep/71470dd252966636150e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..43b2dc3b0fac6c92d45c7344aeed565271afe0d7
--- /dev/null
+++ b/records/sd1/sweep/71470dd252966636150e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71470dd252966636150e)"
+  },
+  "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}",
+   "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/7151d9433ef00c2b3cda/measurements.json b/records/sd1/sweep/7151d9433ef00c2b3cda/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e1436b8e17b927ea0d31f7743427996d594f729
--- /dev/null
+++ b/records/sd1/sweep/7151d9433ef00c2b3cda/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7151d9433ef00c2b3cda",
+ "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/7151d9433ef00c2b3cda/record.json b/records/sd1/sweep/7151d9433ef00c2b3cda/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..24bd6e3e37b05af2984dabacb5bffc4a6482ec73
--- /dev/null
+++ b/records/sd1/sweep/7151d9433ef00c2b3cda/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7151d9433ef00c2b3cda",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:48Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/42bed510ef71fe4dfcd29fae1956fc095e9b44c5932bde9d89357b6376268943.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "42bed510ef71fe4dfcd29fae1956fc095e9b44c5932bde9d89357b6376268943",
+  "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": "7767299879b4db8fba4e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7151d9433ef00c2b3cda/workflow.json b/records/sd1/sweep/7151d9433ef00c2b3cda/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f04d6d9c62a35f7cf2a799516ec506b3c04c4c43
--- /dev/null
+++ b/records/sd1/sweep/7151d9433ef00c2b3cda/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 7151d9433ef00c2b3cda)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.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": [
+    "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/71520b7f2b74367db816/measurements.json b/records/sd1/sweep/71520b7f2b74367db816/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6ef87afe8e1d45446521432776d93d2c5ce21a0
--- /dev/null
+++ b/records/sd1/sweep/71520b7f2b74367db816/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71520b7f2b74367db816",
+ "values": {
+  "cosine_distance": {
+   "value": 0.647212,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.602743,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.792283,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.404497,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 26.247,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03448,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 51.16,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8331,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1587,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71520b7f2b74367db816/record.json b/records/sd1/sweep/71520b7f2b74367db816/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c643e79197b518975d7714dbccba9260a462a07d
--- /dev/null
+++ b/records/sd1/sweep/71520b7f2b74367db816/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71520b7f2b74367db816",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.6.0.op",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "kind": "downsample",
+   "module_type": "Conv2d",
+   "container": "Downsample",
+   "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/7bdf514775089a07ba13a531cd4d86ae4d3e5cdcc91eb5336a13dc5203ea641b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7bdf514775089a07ba13a531cd4d86ae4d3e5cdcc91eb5336a13dc5203ea641b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.Downsample.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "6984b9393ebc2e34449a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71520b7f2b74367db816/workflow.json b/records/sd1/sweep/71520b7f2b74367db816/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fcf651bc82e31513b51397f3f4cae9361f1d543
--- /dev/null
+++ b/records/sd1/sweep/71520b7f2b74367db816/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 71520b7f2b74367db816)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.6.0.op\",\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/7179aaa6d28804253b4e/measurements.json b/records/sd1/sweep/7179aaa6d28804253b4e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..17b1939b074aa38d8cb4d449493b7a27978247aa
--- /dev/null
+++ b/records/sd1/sweep/7179aaa6d28804253b4e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7179aaa6d28804253b4e",
+ "values": {
+  "std": {
+   "value": 29.354,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02894,
+   "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": 5.162,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013429,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.056589,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.029275,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007164,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9913,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2469,
+   "method": "CLIP zero-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/7179aaa6d28804253b4e/record.json b/records/sd1/sweep/7179aaa6d28804253b4e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..87ef7c59edd00fa3c6a784957facb66dd2725872
--- /dev/null
+++ b/records/sd1/sweep/7179aaa6d28804253b4e/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "7179aaa6d28804253b4e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:51:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "in.hi",
+   "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": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_noise",
+   0.2,
+   "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": "30b65488cde6d2cd74fe"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7179aaa6d28804253b4e/workflow.json b/records/sd1/sweep/7179aaa6d28804253b4e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cdd44f14941fa4c35b860acb0aca4fe4d0d957e
--- /dev/null
+++ b/records/sd1/sweep/7179aaa6d28804253b4e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7179aaa6d28804253b4e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a city street at night with neon signs, digital illustration"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 7,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71933eb438f8500d105e/measurements.json b/records/sd1/sweep/71933eb438f8500d105e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..74fe2e895c45d997b32f89cc48d76d3293ce026c
--- /dev/null
+++ b/records/sd1/sweep/71933eb438f8500d105e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71933eb438f8500d105e",
+ "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/71933eb438f8500d105e/record.json b/records/sd1/sweep/71933eb438f8500d105e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e84f807b875346646df4f1f2f7a6476d33437a0
--- /dev/null
+++ b/records/sd1/sweep/71933eb438f8500d105e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71933eb438f8500d105e",
+ "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.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.hi",
+   "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/1af32e59eff4bba18941e0d9d9ede28a6008b3d0119b65948f4df7e855ba1a06.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1af32e59eff4bba18941e0d9d9ede28a6008b3d0119b65948f4df7e855ba1a06",
+  "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": "f2740f1fbd6dc5f867d7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71933eb438f8500d105e/workflow.json b/records/sd1/sweep/71933eb438f8500d105e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..16465209ef8d9d1175844b9f7d467492dd884afe
--- /dev/null
+++ b/records/sd1/sweep/71933eb438f8500d105e/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 71933eb438f8500d105e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/7198dd9e359b42e8c5a8/interpretations.jsonl b/records/sd1/sweep/7198dd9e359b42e8c5a8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..1f896312b39b9e5da57e047968b560a500ff0963
--- /dev/null
+++ b/records/sd1/sweep/7198dd9e359b42e8c5a8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7198dd9e359b42e8c5a8"}, "kind": "caption", "value": "Blurry abstract smear of red, orange, pink, white and dark green patches", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4b5a846f7560f87d"}
+{"target": {"record": "7198dd9e359b42e8c5a8"}, "kind": "change", "value": "The image turns into a soft, painterly abstract of warm reds and pinks with dark streaks, like an out-of-focus floral or flame, losing the subject.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "44fcd8e69a7ad815"}
+{"target": {"record": "7198dd9e359b42e8c5a8"}, "kind": "keywords", "value": ["warm abstract", "red smear", "painterly blur", "floral haze", "lost subject"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "88bfbfad54d11438"}
+{"target": {"record": "7198dd9e359b42e8c5a8"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "painterly", "saturated", "blur", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:18:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6b49cb55db67d33b"}
diff --git a/records/sd1/sweep/7198dd9e359b42e8c5a8/measurements.json b/records/sd1/sweep/7198dd9e359b42e8c5a8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b09987f28861ee2789f120dc1cd103208b34c83
--- /dev/null
+++ b/records/sd1/sweep/7198dd9e359b42e8c5a8/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7198dd9e359b42e8c5a8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.788768,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.766676,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.892899,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.444986,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 85.677,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0053,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 87.626,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8895,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5599,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7198dd9e359b42e8c5a8/record.json b/records/sd1/sweep/7198dd9e359b42e8c5a8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..908d7a0424f4573bfa6938645f56599f7552e75c
--- /dev/null
+++ b/records/sd1/sweep/7198dd9e359b42e8c5a8/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "7198dd9e359b42e8c5a8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:25Z",
+ "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.9.0.skip_connection",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.hi",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/54de6af791c65f946b40c0e0e57a273e53d47522b6818541033e0259428520fd.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "54de6af791c65f946b40c0e0e57a273e53d47522b6818541033e0259428520fd",
+  "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": "a30ade964ea5e67714cb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7198dd9e359b42e8c5a8/workflow.json b/records/sd1/sweep/7198dd9e359b42e8c5a8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bac3e96b61cfd64f41ced3361a02bdfe08dad0dd
--- /dev/null
+++ b/records/sd1/sweep/7198dd9e359b42e8c5a8/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 7198dd9e359b42e8c5a8)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.skip_connection\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/719d47eee18d785c9537/measurements.json b/records/sd1/sweep/719d47eee18d785c9537/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b27b3d18fa62d6c4ccdcb04a863fd1e685d7b4b9
--- /dev/null
+++ b/records/sd1/sweep/719d47eee18d785c9537/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "719d47eee18d785c9537",
+ "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/719d47eee18d785c9537/record.json b/records/sd1/sweep/719d47eee18d785c9537/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f87374391a833cfea778ed6c32ff0c0356ccd21
--- /dev/null
+++ b/records/sd1/sweep/719d47eee18d785c9537/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "719d47eee18d785c9537",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/72fc3452f0cc4312f9a52f4e7759ff7eb4383502e27c513e4f9521263c715225.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "72fc3452f0cc4312f9a52f4e7759ff7eb4383502e27c513e4f9521263c715225",
+  "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": "5c0e4c82a3ef913a78a4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/719d47eee18d785c9537/workflow.json b/records/sd1/sweep/719d47eee18d785c9537/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0888843a58ed2c1fe0888dc5882598a905075930
--- /dev/null
+++ b/records/sd1/sweep/719d47eee18d785c9537/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 719d47eee18d785c9537)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.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": [
+    "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/71a0b17d70988c0b66e3/measurements.json b/records/sd1/sweep/71a0b17d70988c0b66e3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..908e941dd509c78f00a1e14fe95d1d9ab2a10621
--- /dev/null
+++ b/records/sd1/sweep/71a0b17d70988c0b66e3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71a0b17d70988c0b66e3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.307371,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.329744,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.128504,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.056476,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.77,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00593,
+   "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": 35.142,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9551,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1439,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71a0b17d70988c0b66e3/record.json b/records/sd1/sweep/71a0b17d70988c0b66e3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e2b026e3281381e54adfda65072c5f654515c70
--- /dev/null
+++ b/records/sd1/sweep/71a0b17d70988c0b66e3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71a0b17d70988c0b66e3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:59Z",
+ "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": "input_blocks.4.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.mid",
+   "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/d2d3019e592e8ad3c5fbf209484676ac3cf010fb51b8fa9b2f97e2b1acc8edb1.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d2d3019e592e8ad3c5fbf209484676ac3cf010fb51b8fa9b2f97e2b1acc8edb1",
+  "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": "dee3c560d416c119ad01"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71a0b17d70988c0b66e3/workflow.json b/records/sd1/sweep/71a0b17d70988c0b66e3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..28c04529caaf7fd9ec9957a35357eb4b998339f7
--- /dev/null
+++ b/records/sd1/sweep/71a0b17d70988c0b66e3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71a0b17d70988c0b66e3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.proj_in\",\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/71a752dbeca05056caf0/measurements.json b/records/sd1/sweep/71a752dbeca05056caf0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0113b2b8d73690edda0a50a17f250967523a7cf1
--- /dev/null
+++ b/records/sd1/sweep/71a752dbeca05056caf0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71a752dbeca05056caf0",
+ "values": {
+  "std": {
+   "value": 50.329,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01115,
+   "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": 12.964,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.115924,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.219456,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.233671,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036546,
+   "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": 0.0046,
+   "method": "CLIP zero-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/71a752dbeca05056caf0/record.json b/records/sd1/sweep/71a752dbeca05056caf0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0501aff1ad00c51d1274e5d07018f4911293a29b
--- /dev/null
+++ b/records/sd1/sweep/71a752dbeca05056caf0/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "71a752dbeca05056caf0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:35:15Z",
+ "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 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": 1234
+   },
+   "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": "27f5baccf7c0d4f1de23",
+  "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/71a752dbeca05056caf0/workflow.json b/records/sd1/sweep/71a752dbeca05056caf0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c76afb02a7d18865484b6ddee68a75bbec50a89
--- /dev/null
+++ b/records/sd1/sweep/71a752dbeca05056caf0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71a752dbeca05056caf0)"
+  },
+  "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\": 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/71a7a79b7a71898ec6fd/measurements.json b/records/sd1/sweep/71a7a79b7a71898ec6fd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5eb038e102e14ab762cf105f957d6a2066deff91
--- /dev/null
+++ b/records/sd1/sweep/71a7a79b7a71898ec6fd/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71a7a79b7a71898ec6fd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.197769,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.258991,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.119173,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044618,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.963,
+   "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": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.059,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9494,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1076,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71a7a79b7a71898ec6fd/record.json b/records/sd1/sweep/71a7a79b7a71898ec6fd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e79cd7c1b310d1bd4130a6dc9b001b99047f18c
--- /dev/null
+++ b/records/sd1/sweep/71a7a79b7a71898ec6fd/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71a7a79b7a71898ec6fd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/777954ee6cfb1680b5cbd8fec04643bbda300fe890f0a8694d6d1c25aa12907c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "777954ee6cfb1680b5cbd8fec04643bbda300fe890f0a8694d6d1c25aa12907c",
+  "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": "07886457c23a66b7c507"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71a7a79b7a71898ec6fd/workflow.json b/records/sd1/sweep/71a7a79b7a71898ec6fd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bff3661ad84b7cf9998444ccafbb9c7ca8624d6f
--- /dev/null
+++ b/records/sd1/sweep/71a7a79b7a71898ec6fd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71a7a79b7a71898ec6fd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.out_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71aaf97104bdc9e388a7/interpretations.jsonl b/records/sd1/sweep/71aaf97104bdc9e388a7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c3e77a87c66f9507f292980bb1134502b0057084
--- /dev/null
+++ b/records/sd1/sweep/71aaf97104bdc9e388a7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "71aaf97104bdc9e388a7"}, "kind": "caption", "value": "Near-monochrome portrait of woman with messy short dark hair on flat beige background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:15:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b9d1ab3bc1b91ceb"}
+{"target": {"record": "71aaf97104bdc9e388a7"}, "kind": "change", "value": "Colour drains to a nearly greyscale, faded look, the plate border dissolves into a flat beige ground and the hair becomes wispy.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:15:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6f1730b9bf46bc90"}
+{"target": {"record": "71aaf97104bdc9e388a7"}, "kind": "keywords", "value": ["desaturated", "monochrome", "beige background", "messy hair", "borderless"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:15:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5dce4f9245bab2b4"}
+{"target": {"record": "71aaf97104bdc9e388a7"}, "kind": "effect_tags", "value": ["desaturated", "washed_out", "subject_kept", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:15:22Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9d852f1f37f39f58"}
diff --git a/records/sd1/sweep/71aaf97104bdc9e388a7/measurements.json b/records/sd1/sweep/71aaf97104bdc9e388a7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..551e6f08f3bc6fcf16ef5b01119c689d19da1ea6
--- /dev/null
+++ b/records/sd1/sweep/71aaf97104bdc9e388a7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71aaf97104bdc9e388a7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.248948,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.295083,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.107637,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.143461,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.354,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00749,
+   "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": 31.374,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9739,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.251,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71aaf97104bdc9e388a7/record.json b/records/sd1/sweep/71aaf97104bdc9e388a7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf3dc50b1e816b98809a865713f906909dbe021a
--- /dev/null
+++ b/records/sd1/sweep/71aaf97104bdc9e388a7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71aaf97104bdc9e388a7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:53Z",
+ "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": "middle_block.0.in_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "mid",
+   "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/9eacee14cd632b976eff89366ed4122dfd2f5fd13debdcf2e54fc2bbf77198cb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9eacee14cd632b976eff89366ed4122dfd2f5fd13debdcf2e54fc2bbf77198cb",
+  "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": "9361c0318d77823fedd3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71aaf97104bdc9e388a7/workflow.json b/records/sd1/sweep/71aaf97104bdc9e388a7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c0c6aa8d558cb556a5addd7bc255c579514f1858
--- /dev/null
+++ b/records/sd1/sweep/71aaf97104bdc9e388a7/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 71aaf97104bdc9e388a7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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/71b5042b73c4f47b3df3/measurements.json b/records/sd1/sweep/71b5042b73c4f47b3df3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b4159b82b661846fb2e5ff40c193b471ee5d503
--- /dev/null
+++ b/records/sd1/sweep/71b5042b73c4f47b3df3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71b5042b73c4f47b3df3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.365068,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.391807,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.23147,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.075588,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 76.907,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00732,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 47.701,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9623,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0525,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71b5042b73c4f47b3df3/record.json b/records/sd1/sweep/71b5042b73c4f47b3df3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..682788a1bebcbed098763bfcaa2d71d9c0c75531
--- /dev/null
+++ b/records/sd1/sweep/71b5042b73c4f47b3df3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71b5042b73c4f47b3df3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:31Z",
+ "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.1.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9501c0b4e795cd8d9e30487f9af384aec22f9cdd9bdab1ef5c72e1ba72bbd3f5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9501c0b4e795cd8d9e30487f9af384aec22f9cdd9bdab1ef5c72e1ba72bbd3f5",
+  "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": "3f91b3a09ab02200479b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71b5042b73c4f47b3df3/workflow.json b/records/sd1/sweep/71b5042b73c4f47b3df3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e93dd21afc08528c4bdebc80eb9cac5ed507edcf
--- /dev/null
+++ b/records/sd1/sweep/71b5042b73c4f47b3df3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71b5042b73c4f47b3df3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71b9322cff88615a71d1/measurements.json b/records/sd1/sweep/71b9322cff88615a71d1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb8ccad9553dbc96a4168c66b64063ac32881fd0
--- /dev/null
+++ b/records/sd1/sweep/71b9322cff88615a71d1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71b9322cff88615a71d1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.287888,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.35642,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131494,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.113289,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.17,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00374,
+   "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": 29.213,
+   "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": 0.1169,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71b9322cff88615a71d1/record.json b/records/sd1/sweep/71b9322cff88615a71d1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..19031971067665077658a14b3c9b2d0137977128
--- /dev/null
+++ b/records/sd1/sweep/71b9322cff88615a71d1/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71b9322cff88615a71d1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:40Z",
+ "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.8.1.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/2b079c07ddae0ae8abc731058ce96a6a39a3c0d39364b2c1692618b42a7f9640.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2b079c07ddae0ae8abc731058ce96a6a39a3c0d39364b2c1692618b42a7f9640",
+  "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": "19c1815b43e8e6c6dd08"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71b9322cff88615a71d1/workflow.json b/records/sd1/sweep/71b9322cff88615a71d1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0d59ab213ebe657463ff06611da479dffeeb3ce
--- /dev/null
+++ b/records/sd1/sweep/71b9322cff88615a71d1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71b9322cff88615a71d1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.norm3\",\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/71ba725ec0e12ffddd0d/measurements.json b/records/sd1/sweep/71ba725ec0e12ffddd0d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7df7e24832da287f0beed52bbc78f81545200cb
--- /dev/null
+++ b/records/sd1/sweep/71ba725ec0e12ffddd0d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71ba725ec0e12ffddd0d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.513122,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.592016,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.524052,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.235178,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 38.809,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0023,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 47.207,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0817,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1784,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71ba725ec0e12ffddd0d/record.json b/records/sd1/sweep/71ba725ec0e12ffddd0d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ca15a15d316e371c29d431544a77badf12f2921
--- /dev/null
+++ b/records/sd1/sweep/71ba725ec0e12ffddd0d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71ba725ec0e12ffddd0d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/3ec80223badd317ec14586efee6caa76f88b4dda5d9e6fe42ac08818dff496fa.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3ec80223badd317ec14586efee6caa76f88b4dda5d9e6fe42ac08818dff496fa",
+  "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": "99e0c4899bdafddcca2a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71ba725ec0e12ffddd0d/workflow.json b/records/sd1/sweep/71ba725ec0e12ffddd0d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2b41882e61361b77a1c3e418324f771201d7551
--- /dev/null
+++ b/records/sd1/sweep/71ba725ec0e12ffddd0d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71ba725ec0e12ffddd0d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.in_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bb5012164a18cdc389/interpretations.jsonl b/records/sd1/sweep/71bb5012164a18cdc389/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..9977408ff81b81d483e6d0e2bd5a542cc66b98e8
--- /dev/null
+++ b/records/sd1/sweep/71bb5012164a18cdc389/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "71bb5012164a18cdc389"}, "kind": "caption", "value": "Same portrait of short-haired woman, but the centre of her face is smeared and partly erased.", "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": "89c412b17a34758c"}
+{"target": {"record": "71bb5012164a18cdc389"}, "kind": "change", "value": "Composition, hair, clothing and grey backdrop stay the same, but the facial features dissolve into a patchy, ghostly smear, as if the face were rubbed out.", "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": "2bb9063d9a5a5616"}
+{"target": {"record": "71bb5012164a18cdc389"}, "kind": "keywords", "value": ["erased face", "smeared features", "ghostly portrait", "facial dissolution", "patchy skin"], "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": "4d28e20e231a7539"}
+{"target": {"record": "71bb5012164a18cdc389"}, "kind": "effect_tags", "value": ["subject_kept", "distorted", "fragmented", "surreal"], "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": "738c73c0e6595c19"}
diff --git a/records/sd1/sweep/71bb5012164a18cdc389/measurements.json b/records/sd1/sweep/71bb5012164a18cdc389/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5cebac8da36da302bc67a1ea67c93ba8749a880c
--- /dev/null
+++ b/records/sd1/sweep/71bb5012164a18cdc389/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71bb5012164a18cdc389",
+ "values": {
+  "cosine_distance": {
+   "value": 0.087969,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.175308,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.232301,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.160817,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.908,
+   "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": [
+    "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": 15.241,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9397,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6632,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bb5012164a18cdc389/record.json b/records/sd1/sweep/71bb5012164a18cdc389/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d5a255825147d56b354ee9037718664b46ae08e
--- /dev/null
+++ b/records/sd1/sweep/71bb5012164a18cdc389/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71bb5012164a18cdc389",
+ "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.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.hi",
+   "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/4ecbfe68e412cfb22eee25471774d07cc8456366d5c369024f9130faf12587b9.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "4ecbfe68e412cfb22eee25471774d07cc8456366d5c369024f9130faf12587b9",
+  "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": "08db4cdd393f3e4b42aa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bb5012164a18cdc389/workflow.json b/records/sd1/sweep/71bb5012164a18cdc389/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8907e5b4c470bc83c2d8f99f457c1084a7715ae1
--- /dev/null
+++ b/records/sd1/sweep/71bb5012164a18cdc389/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 71bb5012164a18cdc389)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.in_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bc1d3a1e929a71d165/measurements.json b/records/sd1/sweep/71bc1d3a1e929a71d165/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d3b9895be775ee202288702f4ee1d33d75945ab
--- /dev/null
+++ b/records/sd1/sweep/71bc1d3a1e929a71d165/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71bc1d3a1e929a71d165",
+ "values": {
+  "cosine_distance": {
+   "value": 0.203117,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.294951,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.12352,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.06753,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 43.824,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00572,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 27.298,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9462,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0909,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bc1d3a1e929a71d165/record.json b/records/sd1/sweep/71bc1d3a1e929a71d165/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4816a187b96a77d0995e34a8559a6ad442437811
--- /dev/null
+++ b/records/sd1/sweep/71bc1d3a1e929a71d165/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71bc1d3a1e929a71d165",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:31Z",
+ "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.6.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/91c8dbd23aaba374238e415669c8ab67dd45d041459efa5c4be816342a2fb1fd.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "91c8dbd23aaba374238e415669c8ab67dd45d041459efa5c4be816342a2fb1fd",
+  "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": "b1d58d8505726ba5c327"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bc1d3a1e929a71d165/workflow.json b/records/sd1/sweep/71bc1d3a1e929a71d165/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e2648ccff1b054a32a48ac086b8d065b2f06319a
--- /dev/null
+++ b/records/sd1/sweep/71bc1d3a1e929a71d165/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71bc1d3a1e929a71d165)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bfa7fc20d9d615bfb2/measurements.json b/records/sd1/sweep/71bfa7fc20d9d615bfb2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e9ae676a9d989789b8ac610bd95aea66c021d9b
--- /dev/null
+++ b/records/sd1/sweep/71bfa7fc20d9d615bfb2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71bfa7fc20d9d615bfb2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.065293,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.076405,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.057081,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.018265,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.492,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00579,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.73,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9926,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2625,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bfa7fc20d9d615bfb2/record.json b/records/sd1/sweep/71bfa7fc20d9d615bfb2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..923bdfccb2532e69e42d2b3842873a5e1f8345c0
--- /dev/null
+++ b/records/sd1/sweep/71bfa7fc20d9d615bfb2/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71bfa7fc20d9d615bfb2",
+ "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_k",
+   "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/0a9086f70366ab0e55a2e44508f60d0d7f68a0cf4e6572377af1865a7b575b3c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0a9086f70366ab0e55a2e44508f60d0d7f68a0cf4e6572377af1865a7b575b3c",
+  "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": "a6274ab77e67c01f36cd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71bfa7fc20d9d615bfb2/workflow.json b/records/sd1/sweep/71bfa7fc20d9d615bfb2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae5fdf85a13e3b6ee84bbb90df6903f362f4f9b6
--- /dev/null
+++ b/records/sd1/sweep/71bfa7fc20d9d615bfb2/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71bfa7fc20d9d615bfb2)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.attn2.to_k\",\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/71c034971a6ea01116d1/measurements.json b/records/sd1/sweep/71c034971a6ea01116d1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9ab2608fefbc150d1efcf5ec1ba1c3e32e85740
--- /dev/null
+++ b/records/sd1/sweep/71c034971a6ea01116d1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71c034971a6ea01116d1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.127086,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.162337,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.155962,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.038665,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.524,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00523,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.743,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0376,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/71c034971a6ea01116d1/record.json b/records/sd1/sweep/71c034971a6ea01116d1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7fbf2e7422c47fee00fcbf47583ae695d63c855
--- /dev/null
+++ b/records/sd1/sweep/71c034971a6ea01116d1/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71c034971a6ea01116d1",
+ "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": 1.5
+   },
+   "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/79024c81e724c618e7292495b338099e3734d589e39c2a6c0f80821f4df8ecf8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "79024c81e724c618e7292495b338099e3734d589e39c2a6c0f80821f4df8ecf8",
+  "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": "f4b8b768a6c2c2db9be0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71c034971a6ea01116d1/workflow.json b/records/sd1/sweep/71c034971a6ea01116d1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..97cde296a371212a20cbfb8315bb8cb16f049edd
--- /dev/null
+++ b/records/sd1/sweep/71c034971a6ea01116d1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71c034971a6ea01116d1)"
+  },
+  "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\": 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/71c6a089510ada38474e/measurements.json b/records/sd1/sweep/71c6a089510ada38474e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..079542e000f842173025f9fe2b3c6b23635ccd98
--- /dev/null
+++ b/records/sd1/sweep/71c6a089510ada38474e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71c6a089510ada38474e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.276071,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.306593,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.174726,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.082922,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.482,
+   "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": 33.173,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0446,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3457,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71c6a089510ada38474e/record.json b/records/sd1/sweep/71c6a089510ada38474e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..56e572004cdb6f9bb63ead0e6deec30d9943a632
--- /dev/null
+++ b/records/sd1/sweep/71c6a089510ada38474e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71c6a089510ada38474e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:40Z",
+ "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": "input_blocks.5.0.out_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.mid",
+   "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/a10bd9e97e19fc7abaf059975fd88d0392e07b12c369e2dcc21efb37f05961e0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a10bd9e97e19fc7abaf059975fd88d0392e07b12c369e2dcc21efb37f05961e0",
+  "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": "dba152fb1fd966171cbb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71c6a089510ada38474e/workflow.json b/records/sd1/sweep/71c6a089510ada38474e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a78c32723f5f2da0b52926d4a4d561c4e85bf31
--- /dev/null
+++ b/records/sd1/sweep/71c6a089510ada38474e/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 71c6a089510ada38474e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_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/71c72442d1009e2dac97/measurements.json b/records/sd1/sweep/71c72442d1009e2dac97/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4698f6dd7be673afa3d78943651d96c4c7fe8cc
--- /dev/null
+++ b/records/sd1/sweep/71c72442d1009e2dac97/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71c72442d1009e2dac97",
+ "values": {
+  "cosine_distance": {
+   "value": 0.219198,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.250011,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.098798,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036625,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 48.554,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00435,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.775,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.971,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1431,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71c72442d1009e2dac97/record.json b/records/sd1/sweep/71c72442d1009e2dac97/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..70443b3371852f62d50b6abfb416f319895cd9ea
--- /dev/null
+++ b/records/sd1/sweep/71c72442d1009e2dac97/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71c72442d1009e2dac97",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:41Z",
+ "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": 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.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.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/59b5467eaefc3a53742b581e4ebaa4eec337bc6b053109e29d670a8357456e73.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "59b5467eaefc3a53742b581e4ebaa4eec337bc6b053109e29d670a8357456e73",
+  "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": "10cbd0f6ec6f054f866e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71c72442d1009e2dac97/workflow.json b/records/sd1/sweep/71c72442d1009e2dac97/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1dbecbbac08068dfb8ac2afc790dc00aceb2405c
--- /dev/null
+++ b/records/sd1/sweep/71c72442d1009e2dac97/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71c72442d1009e2dac97)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.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/71cfc42591aab45cafcc/measurements.json b/records/sd1/sweep/71cfc42591aab45cafcc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5dacaf1c72e3f8d5eae169222b15c022b9c0a637
--- /dev/null
+++ b/records/sd1/sweep/71cfc42591aab45cafcc/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71cfc42591aab45cafcc",
+ "values": {
+  "cosine_distance": {
+   "value": 0.095085,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.115639,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.128589,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.034225,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.551,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00461,
+   "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": 13.94,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.1429,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71cfc42591aab45cafcc/record.json b/records/sd1/sweep/71cfc42591aab45cafcc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4404234514137639865994396369ad85a8dbe62d
--- /dev/null
+++ b/records/sd1/sweep/71cfc42591aab45cafcc/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71cfc42591aab45cafcc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:07Z",
+ "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": "middle_block.2.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/41ac2cebada6a3f0359101d923100a1d47cf222be74adee8f8af4d2c8a055159.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "41ac2cebada6a3f0359101d923100a1d47cf222be74adee8f8af4d2c8a055159",
+  "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": "f4703b6cd3dbf2959838"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71cfc42591aab45cafcc/workflow.json b/records/sd1/sweep/71cfc42591aab45cafcc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5b172527aecd1e30147fdf5318909cc3ce3eb7e
--- /dev/null
+++ b/records/sd1/sweep/71cfc42591aab45cafcc/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71cfc42591aab45cafcc)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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/71dc6ebf8593a2d4b663/measurements.json b/records/sd1/sweep/71dc6ebf8593a2d4b663/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c8db836fcf7c19b10df26e0be0763e941c804d4
--- /dev/null
+++ b/records/sd1/sweep/71dc6ebf8593a2d4b663/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71dc6ebf8593a2d4b663",
+ "values": {
+  "cosine_distance": {
+   "value": 0.12206,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.173675,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.100729,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.067555,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.102,
+   "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": 17.284,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9672,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6855,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71dc6ebf8593a2d4b663/record.json b/records/sd1/sweep/71dc6ebf8593a2d4b663/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..63c9b1fee9994d825eb6575bac86214b4e529ad6
--- /dev/null
+++ b/records/sd1/sweep/71dc6ebf8593a2d4b663/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71dc6ebf8593a2d4b663",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.in_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.mid",
+   "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/768a0a9db6291ebdcd81f5a496df1f57932d9ed791beb9b0c468481e0514b225.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "768a0a9db6291ebdcd81f5a496df1f57932d9ed791beb9b0c468481e0514b225",
+  "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": "317d7c1f8f8293d18e6a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71dc6ebf8593a2d4b663/workflow.json b/records/sd1/sweep/71dc6ebf8593a2d4b663/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..55934e130bcabaab68d24a62f4a2ab0ed4f71503
--- /dev/null
+++ b/records/sd1/sweep/71dc6ebf8593a2d4b663/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 71dc6ebf8593a2d4b663)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.in_layers.2\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 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/71de26a2cde7c146934c/measurements.json b/records/sd1/sweep/71de26a2cde7c146934c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3609270739b331acc54db55ba14492f857b324f6
--- /dev/null
+++ b/records/sd1/sweep/71de26a2cde7c146934c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71de26a2cde7c146934c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.306009,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.314155,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.167878,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.100882,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.079,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00564,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 32.279,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0419,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2554,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71de26a2cde7c146934c/record.json b/records/sd1/sweep/71de26a2cde7c146934c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..577868db9867bd694ffb0217063b4973a4006758
--- /dev/null
+++ b/records/sd1/sweep/71de26a2cde7c146934c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71de26a2cde7c146934c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.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/f015f9f39d636ce3d97f48e86253144a97ae2339636df8a04c0d4007f51d12b5.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f015f9f39d636ce3d97f48e86253144a97ae2339636df8a04c0d4007f51d12b5",
+  "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": "c8cfb0015f6638677f13"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71de26a2cde7c146934c/workflow.json b/records/sd1/sweep/71de26a2cde7c146934c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7e37904e82cc26415ffecb47d2bb2c641249abd
--- /dev/null
+++ b/records/sd1/sweep/71de26a2cde7c146934c/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 71de26a2cde7c146934c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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/71e43f983c41558872aa/measurements.json b/records/sd1/sweep/71e43f983c41558872aa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e42e7dcfaace677c0a441eb3aa1114a43d3bbb7
--- /dev/null
+++ b/records/sd1/sweep/71e43f983c41558872aa/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71e43f983c41558872aa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.123218,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.145886,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.068052,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.147971,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.639,
+   "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": 15.221,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.942,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4063,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71e43f983c41558872aa/record.json b/records/sd1/sweep/71e43f983c41558872aa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2a9be3096e867cf40505924d2fd367e9f381a27
--- /dev/null
+++ b/records/sd1/sweep/71e43f983c41558872aa/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71e43f983c41558872aa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:35Z",
+ "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.11.0.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/7e5e83778281db656863d974078dd5a78b316b18c6922f0dfae00083e6fb525a.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7e5e83778281db656863d974078dd5a78b316b18c6922f0dfae00083e6fb525a",
+  "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": "be4d5b139fcbce508e50"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71e43f983c41558872aa/workflow.json b/records/sd1/sweep/71e43f983c41558872aa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..75bbb17dfe74ef1b3a11d05cad2a3a94233b1832
--- /dev/null
+++ b/records/sd1/sweep/71e43f983c41558872aa/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 71e43f983c41558872aa)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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/71e9b69f821ef3764a3c/interpretations.jsonl b/records/sd1/sweep/71e9b69f821ef3764a3c/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..23f0f428b4c53e55eabb0a026b5b0ec76147f107
--- /dev/null
+++ b/records/sd1/sweep/71e9b69f821ef3764a3c/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "71e9b69f821ef3764a3c"}, "kind": "caption", "value": "Grainy sepia portrait with a heavily shadowed, boyish face and messy hair on beige paper", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:16:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "82c42b577dd21132"}
+{"target": {"record": "71e9b69f821ef3764a3c"}, "kind": "change", "value": "The image turns into a grainy, low-fidelity print with harsh shadows across a boyish, distorted face, messy hair and a flat beige paper background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:16:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "13cda110b2dff6ac"}
+{"target": {"record": "71e9b69f821ef3764a3c"}, "kind": "keywords", "value": ["grainy print", "harsh shadows", "boyish face", "beige paper", "low fidelity"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:16:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9f3c47da42bb1553"}
+{"target": {"record": "71e9b69f821ef3764a3c"}, "kind": "effect_tags", "value": ["high_contrast", "texture", "distorted", "desaturated", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:16:05Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "01bb7b78950b4f48"}
diff --git a/records/sd1/sweep/71e9b69f821ef3764a3c/measurements.json b/records/sd1/sweep/71e9b69f821ef3764a3c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..723a8491e57df08e9e7180385b1f5b5e3ca5b813
--- /dev/null
+++ b/records/sd1/sweep/71e9b69f821ef3764a3c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71e9b69f821ef3764a3c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.292364,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.378985,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.303926,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.233104,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.291,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00512,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 35.886,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9821,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2583,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71e9b69f821ef3764a3c/record.json b/records/sd1/sweep/71e9b69f821ef3764a3c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45c81385c313b70eb61223327532049c1e899163
--- /dev/null
+++ b/records/sd1/sweep/71e9b69f821ef3764a3c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71e9b69f821ef3764a3c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:49Z",
+ "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.9.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/bca7a32cda641fc8e5193cc7e63c6484b5975eac2e479cc33a8285d44a62e5e1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "bca7a32cda641fc8e5193cc7e63c6484b5975eac2e479cc33a8285d44a62e5e1",
+  "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": "f1530058c17b9c3ba80e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71e9b69f821ef3764a3c/workflow.json b/records/sd1/sweep/71e9b69f821ef3764a3c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..94b05b2f1f6a8e3b6696682b3a229fb32d19b920
--- /dev/null
+++ b/records/sd1/sweep/71e9b69f821ef3764a3c/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 71e9b69f821ef3764a3c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.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": [
+    "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/71f34cda7a2725fbcce2/measurements.json b/records/sd1/sweep/71f34cda7a2725fbcce2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..51f6750797be8b67c8d64b70b2a59df72b67e945
--- /dev/null
+++ b/records/sd1/sweep/71f34cda7a2725fbcce2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "71f34cda7a2725fbcce2",
+ "values": {
+  "std": {
+   "value": 23.034,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.04424,
+   "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": 31.04,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.364625,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.548639,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.397761,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.072398,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0586,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2557,
+   "method": "CLIP zero-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/71f34cda7a2725fbcce2/record.json b/records/sd1/sweep/71f34cda7a2725fbcce2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b1d46bc42f5c5458c3d5f980e83f6b5e7daf29b3
--- /dev/null
+++ b/records/sd1/sweep/71f34cda7a2725fbcce2/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "71f34cda7a2725fbcce2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:25:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.lo",
+   "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": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "rotate",
+   180.0,
+   "middle"
+  ],
+  "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": "9e93837450a32d4f32d4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71f34cda7a2725fbcce2/workflow.json b/records/sd1/sweep/71f34cda7a2725fbcce2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e9154872654a96b1968cf8faab3355b688a5580
--- /dev/null
+++ b/records/sd1/sweep/71f34cda7a2725fbcce2/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71f34cda7a2725fbcce2)"
+  },
+  "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 \"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/71f683eb119fdf0c2b9e/measurements.json b/records/sd1/sweep/71f683eb119fdf0c2b9e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4224f6303790798fa8c3b48868cbf7130bf84dc
--- /dev/null
+++ b/records/sd1/sweep/71f683eb119fdf0c2b9e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71f683eb119fdf0c2b9e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.439033,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.596667,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.841072,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.319931,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.238,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00245,
+   "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": 45.488,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0039,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5095,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71f683eb119fdf0c2b9e/record.json b/records/sd1/sweep/71f683eb119fdf0c2b9e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ddab7d5a043c1006b2f393c6a88eaeff4f3404b
--- /dev/null
+++ b/records/sd1/sweep/71f683eb119fdf0c2b9e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "71f683eb119fdf0c2b9e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:19Z",
+ "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": "output_blocks.7.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.mid",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ce94a57616ffc9cfc76be6f51958016cd0f6d05bdb4ba11f903feac67c5f758f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ce94a57616ffc9cfc76be6f51958016cd0f6d05bdb4ba11f903feac67c5f758f",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "fe6a3583abd297f0f7b6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71f683eb119fdf0c2b9e/workflow.json b/records/sd1/sweep/71f683eb119fdf0c2b9e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4ed3cfee02174d11a1d7129c8cae3fbbdab4d90
--- /dev/null
+++ b/records/sd1/sweep/71f683eb119fdf0c2b9e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 71f683eb119fdf0c2b9e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/71f70031e8998fa61a65/measurements.json b/records/sd1/sweep/71f70031e8998fa61a65/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e0a7fe058979dd000838261b987f10755760251
--- /dev/null
+++ b/records/sd1/sweep/71f70031e8998fa61a65/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "71f70031e8998fa61a65",
+ "values": {
+  "cosine_distance": {
+   "value": 0.646881,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.575211,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.445221,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.202892,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.912,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00507,
+   "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": 55.016,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9523,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9248,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71f70031e8998fa61a65/record.json b/records/sd1/sweep/71f70031e8998fa61a65/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f3f4dfd949b2765a3d994e68f2a8c3351fb9a91
--- /dev/null
+++ b/records/sd1/sweep/71f70031e8998fa61a65/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "71f70031e8998fa61a65",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:56Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/b6da04f8b7332031a9edbb4ce1040b43f0c0b2cf30b8adf7f3377c9bf4ede96e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b6da04f8b7332031a9edbb4ce1040b43f0c0b2cf30b8adf7f3377c9bf4ede96e",
+  "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": "fe19f46b37968540a642"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/71f70031e8998fa61a65/workflow.json b/records/sd1/sweep/71f70031e8998fa61a65/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa2c84b48ca7ac1c4e66684f7b78c491453e81c3
--- /dev/null
+++ b/records/sd1/sweep/71f70031e8998fa61a65/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 71f70031e8998fa61a65)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.norm1\",\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/7200c124fec9e0d5c97d/measurements.json b/records/sd1/sweep/7200c124fec9e0d5c97d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a49856687da45e4702b3d8439dd0e89c969f2c34
--- /dev/null
+++ b/records/sd1/sweep/7200c124fec9e0d5c97d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7200c124fec9e0d5c97d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.138699,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.155623,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.095004,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036352,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.526,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00555,
+   "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": 0.9435,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/7200c124fec9e0d5c97d/record.json b/records/sd1/sweep/7200c124fec9e0d5c97d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7b9e1fc38d7ab5957f13ad8f2e061bdbe114955d
--- /dev/null
+++ b/records/sd1/sweep/7200c124fec9e0d5c97d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7200c124fec9e0d5c97d",
+ "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.transformer_blocks.0.norm3",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/41bb87cb0c7f43fdbef2d4dc975784e1f5ad5751ff286bfde9c4037e18f211f4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "41bb87cb0c7f43fdbef2d4dc975784e1f5ad5751ff286bfde9c4037e18f211f4",
+  "type_path": "UNetModel.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0dccef45b50f4219a89d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7200c124fec9e0d5c97d/workflow.json b/records/sd1/sweep/7200c124fec9e0d5c97d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c4b4a9bc6553ed29d404fe48d6404397bb1885a
--- /dev/null
+++ b/records/sd1/sweep/7200c124fec9e0d5c97d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7200c124fec9e0d5c97d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.norm3\",\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/720350c0a5677b8d7543/measurements.json b/records/sd1/sweep/720350c0a5677b8d7543/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..92724abfed5a530d505fef689f6ba80c9931d62b
--- /dev/null
+++ b/records/sd1/sweep/720350c0a5677b8d7543/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "720350c0a5677b8d7543",
+ "values": {
+  "cosine_distance": {
+   "value": 0.193212,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.223529,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.238529,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.245767,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.988,
+   "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": 23.339,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9978,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1253,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/720350c0a5677b8d7543/record.json b/records/sd1/sweep/720350c0a5677b8d7543/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebbd937f9bf08c5ea87fe93116170046fa4c2174
--- /dev/null
+++ b/records/sd1/sweep/720350c0a5677b8d7543/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "720350c0a5677b8d7543",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:05Z",
+ "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": "middle_block.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "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/b8c80704fb32c70bac87f6d953e99d98852c4f6eae69fa4871824e57d0ab40ce.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b8c80704fb32c70bac87f6d953e99d98852c4f6eae69fa4871824e57d0ab40ce",
+  "type_path": "UNetModel.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "c7152f18edc9fd0c0e2d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/720350c0a5677b8d7543/workflow.json b/records/sd1/sweep/720350c0a5677b8d7543/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e4ef3050bf4beb1c4800eda3437b0abc303c882
--- /dev/null
+++ b/records/sd1/sweep/720350c0a5677b8d7543/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 720350c0a5677b8d7543)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.norm2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\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/720359581e80966a0cfa/interpretations.jsonl b/records/sd1/sweep/720359581e80966a0cfa/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ec490c7282ab8db766bb9fd15f8db1232654f0c4
--- /dev/null
+++ b/records/sd1/sweep/720359581e80966a0cfa/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "720359581e80966a0cfa"}, "kind": "caption", "value": "Stylised portrait of a woman with dark bob, big eyes, on mottled teal-cream 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": "1151f6de0ab644b4"}
+{"target": {"record": "720359581e80966a0cfa"}, "kind": "change", "value": "The face becomes cuter with larger eyes and fuller cheeks, the hair a glossy dark bob, and the background a mottled teal and cream texture.", "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": "5ddbd92f373d3715"}
+{"target": {"record": "720359581e80966a0cfa"}, "kind": "keywords", "value": ["big eyes", "glossy bob", "teal texture", "stylised portrait"], "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": "5cf2f0a1694f036c"}
+{"target": {"record": "720359581e80966a0cfa"}, "kind": "effect_tags", "value": ["style_change", "texture", "colour_shift", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:01Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b9810b9f2f936d95"}
diff --git a/records/sd1/sweep/720359581e80966a0cfa/measurements.json b/records/sd1/sweep/720359581e80966a0cfa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0dca52c4ff62c599902d39569ce91a1de47fe8c
--- /dev/null
+++ b/records/sd1/sweep/720359581e80966a0cfa/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "720359581e80966a0cfa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.347743,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.403588,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.195593,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.066702,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.815,
+   "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": [
+    "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": 44.587,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9549,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1176,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/720359581e80966a0cfa/record.json b/records/sd1/sweep/720359581e80966a0cfa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e15ce858dded0a9dc7e0bea5835ebfbbb0673565
--- /dev/null
+++ b/records/sd1/sweep/720359581e80966a0cfa/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "720359581e80966a0cfa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9cf9d3787be55fa5e3c5e58678cea7cf0f56437c775c77d22e520d612fc2cdf7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9cf9d3787be55fa5e3c5e58678cea7cf0f56437c775c77d22e520d612fc2cdf7",
+  "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": "eabbe9ae464fdc5d0553"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/720359581e80966a0cfa/workflow.json b/records/sd1/sweep/720359581e80966a0cfa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..67f1acbc4ace1d7e6581a692ca15ee23ce48eaf3
--- /dev/null
+++ b/records/sd1/sweep/720359581e80966a0cfa/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 720359581e80966a0cfa)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72052c7a937402cb6cce/interpretations.jsonl b/records/sd1/sweep/72052c7a937402cb6cce/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..72915beb2595844055f94cdfe8ab2c69e401c4fd
--- /dev/null
+++ b/records/sd1/sweep/72052c7a937402cb6cce/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72052c7a937402cb6cce"}, "kind": "caption", "value": "Painterly portrait of a smiling woman with a voluminous brown quiff and black collar", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "826147f2077eb56b"}
+{"target": {"record": "72052c7a937402cb6cce"}, "kind": "change", "value": "The photo turns into a soft painted illustration with a larger head, fuller, wavier hair, a slight smile and warmer skin tones, and the frame is gone.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5de8c941238db08a"}
+{"target": {"record": "72052c7a937402cb6cce"}, "kind": "keywords", "value": ["painted portrait", "quiff hairstyle", "illustration", "warm skin"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6bc705abdf108632"}
+{"target": {"record": "72052c7a937402cb6cce"}, "kind": "effect_tags", "value": ["subject_kept", "painterly", "style_change", "zoom"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fd3e60faaaf332e7"}
diff --git a/records/sd1/sweep/72052c7a937402cb6cce/measurements.json b/records/sd1/sweep/72052c7a937402cb6cce/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..464da16f552b447631ed3ac724c7fbc5de8492ca
--- /dev/null
+++ b/records/sd1/sweep/72052c7a937402cb6cce/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72052c7a937402cb6cce",
+ "values": {
+  "cosine_distance": {
+   "value": 0.45711,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.474844,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.316878,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.266407,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.354,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00319,
+   "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": 52.055,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9347,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3982,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72052c7a937402cb6cce/record.json b/records/sd1/sweep/72052c7a937402cb6cce/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..51d2d25284559fb00a7d6e5dfefec192ea379165
--- /dev/null
+++ b/records/sd1/sweep/72052c7a937402cb6cce/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72052c7a937402cb6cce",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:30Z",
+ "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": "input_blocks.9.0.op",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.lo",
+   "kind": "downsample",
+   "module_type": "Conv2d",
+   "container": "Downsample",
+   "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/a754a9323d2fab972bf0da8981ee05cb7a9d7530b74516c17611f235db38230b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a754a9323d2fab972bf0da8981ee05cb7a9d7530b74516c17611f235db38230b",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.Downsample.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "763e3157ebd5b2ede826"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72052c7a937402cb6cce/workflow.json b/records/sd1/sweep/72052c7a937402cb6cce/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fff294dab65e1b6eb83a39f102442db51232520
--- /dev/null
+++ b/records/sd1/sweep/72052c7a937402cb6cce/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 72052c7a937402cb6cce)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.9.0.op\",\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/7205941fbacacc13fcc9/interpretations.jsonl b/records/sd1/sweep/7205941fbacacc13fcc9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7303e34e5bdd734360d116792e3f24bd9d2d7287
--- /dev/null
+++ b/records/sd1/sweep/7205941fbacacc13fcc9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7205941fbacacc13fcc9"}, "kind": "caption", "value": "Portrait of woman with fuller dark wavy hair on a teal-grey background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ef86332c817e4a5c"}
+{"target": {"record": "7205941fbacacc13fcc9"}, "kind": "change", "value": "The woman stays similar but the hair becomes fuller and darker, and the background shifts to a cooler teal-grey with a lighter frame edge.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9eb47cf725087f54"}
+{"target": {"record": "7205941fbacacc13fcc9"}, "kind": "keywords", "value": ["fuller hair", "teal background", "cool tones", "portrait variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "66a4e4fa63185f31"}
+{"target": {"record": "7205941fbacacc13fcc9"}, "kind": "effect_tags", "value": ["subject_kept", "subtle", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:23:08Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "44caabb17ab8bd3a"}
diff --git a/records/sd1/sweep/7205941fbacacc13fcc9/measurements.json b/records/sd1/sweep/7205941fbacacc13fcc9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..798e25e4a2c788d3e56269e44c47fdafc1c34889
--- /dev/null
+++ b/records/sd1/sweep/7205941fbacacc13fcc9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7205941fbacacc13fcc9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.093103,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.125863,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.034612,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.014542,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.138,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00683,
+   "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": 15.125,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0341,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1194,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7205941fbacacc13fcc9/record.json b/records/sd1/sweep/7205941fbacacc13fcc9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd28555c4dbb49d8d34afed9bfd2cae6c91a49f1
--- /dev/null
+++ b/records/sd1/sweep/7205941fbacacc13fcc9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7205941fbacacc13fcc9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/036f6ca9f262f81c40d4dc0fe5621e21f4527de5b4c78a3d4a89bf8ecd5d897f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "036f6ca9f262f81c40d4dc0fe5621e21f4527de5b4c78a3d4a89bf8ecd5d897f",
+  "type_path": "UNetModel.TimestepEmbedSequential.SpatialTransformer.ModuleList.BasicTransformerBlock.LayerNorm",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "97ba08864f8a699ce92e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7205941fbacacc13fcc9/workflow.json b/records/sd1/sweep/7205941fbacacc13fcc9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2932a1f56651128bf6ebefd9deda237af7ca9d2
--- /dev/null
+++ b/records/sd1/sweep/7205941fbacacc13fcc9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7205941fbacacc13fcc9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.norm1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7205a5b4096fc8aa770f/measurements.json b/records/sd1/sweep/7205a5b4096fc8aa770f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7d883181a7dcd23c340e2df3a03c6714ff99e5a4
--- /dev/null
+++ b/records/sd1/sweep/7205a5b4096fc8aa770f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7205a5b4096fc8aa770f",
+ "values": {
+  "std": {
+   "value": 41.314,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01281,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.951,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.285085,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.418391,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.343217,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.054638,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.0076,
+   "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/7205a5b4096fc8aa770f/record.json b/records/sd1/sweep/7205a5b4096fc8aa770f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..03376d1ac65ea07afea89c6ccf465869a0832197
--- /dev/null
+++ b/records/sd1/sweep/7205a5b4096fc8aa770f/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "7205a5b4096fc8aa770f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:40:32Z",
+ "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": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "mid",
+   "kind": "attn_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": [
+   "mid",
+   "scale",
+   0.7,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7205a5b4096fc8aa770f/workflow.json b/records/sd1/sweep/7205a5b4096fc8aa770f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..80e3cb185e3e57bfefaa15c738e63fcfdcc19cd2
--- /dev/null
+++ b/records/sd1/sweep/7205a5b4096fc8aa770f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7205a5b4096fc8aa770f)"
+  },
+  "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\": 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/7212c156ed55f97134e5/measurements.json b/records/sd1/sweep/7212c156ed55f97134e5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5082d7570b6b74fa9ea2cffcab0ac1bd4d92e030
--- /dev/null
+++ b/records/sd1/sweep/7212c156ed55f97134e5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7212c156ed55f97134e5",
+ "values": {
+  "std": {
+   "value": 47.063,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01342,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.295,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.00519,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.015505,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.002243,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002007,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0312,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.228,
+   "method": "CLIP zero-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/7212c156ed55f97134e5/record.json b/records/sd1/sweep/7212c156ed55f97134e5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8fe79b7684678adf06ba8d599942bbb92ab7d6a7
--- /dev/null
+++ b/records/sd1/sweep/7212c156ed55f97134e5/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "7212c156ed55f97134e5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:18:12Z",
+ "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": "input_blocks.2.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "medium",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_noise",
+   0.2,
+   "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": "4d3dcfa6718b9f082eef"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7212c156ed55f97134e5/workflow.json b/records/sd1/sweep/7212c156ed55f97134e5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..138e83a22b4d8f139ad7e65c4fa6f159313fe32b
--- /dev/null
+++ b/records/sd1/sweep/7212c156ed55f97134e5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7212c156ed55f97134e5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 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/721fad124e0a979805c6/measurements.json b/records/sd1/sweep/721fad124e0a979805c6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbb4c380d34c007628297473a899339437b0acb0
--- /dev/null
+++ b/records/sd1/sweep/721fad124e0a979805c6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "721fad124e0a979805c6",
+ "values": {
+  "std": {
+   "value": 61.889,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00461,
+   "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": 61.535,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.640564,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.732049,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.955384,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.447515,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.6573,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3334,
+   "method": "CLIP zero-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/721fad124e0a979805c6/record.json b/records/sd1/sweep/721fad124e0a979805c6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a56ce05dc91b657d7ba35b5df81549c3a41b3a41
--- /dev/null
+++ b/records/sd1/sweep/721fad124e0a979805c6/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "721fad124e0a979805c6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:33:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "rotate",
+   180.0,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 180.0,
+    "applied": 180.0,
+    "reason": "outside safe range [0.0, 5.0] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5b330ff2084d65ed06e8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/721fad124e0a979805c6/workflow.json b/records/sd1/sweep/721fad124e0a979805c6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbac969afc043bdcc83199a6ea05b948ceaa66a8
--- /dev/null
+++ b/records/sd1/sweep/721fad124e0a979805c6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 721fad124e0a979805c6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/7222d7b1744dded53944/measurements.json b/records/sd1/sweep/7222d7b1744dded53944/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bc3eea31c222aa14bd620ade36296aa9123c4ab
--- /dev/null
+++ b/records/sd1/sweep/7222d7b1744dded53944/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7222d7b1744dded53944",
+ "values": {
+  "std": {
+   "value": 55.794,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02063,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 8.127,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.008886,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.039582,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.023127,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.012358,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9652,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2717,
+   "method": "CLIP zero-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/7222d7b1744dded53944/record.json b/records/sd1/sweep/7222d7b1744dded53944/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b9e002ab6db1d2013c229e1af4f920734d62fc3
--- /dev/null
+++ b/records/sd1/sweep/7222d7b1744dded53944/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "7222d7b1744dded53944",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:22:31Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "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": 2.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "boost",
+   "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.hi",
+   "multiply",
+   2.0,
+   "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": "fef8f6966d8d3d392d01"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7222d7b1744dded53944/workflow.json b/records/sd1/sweep/7222d7b1744dded53944/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5df35d35f578e57c568fe6d720d4f6dcdf46afc4
--- /dev/null
+++ b/records/sd1/sweep/7222d7b1744dded53944/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7222d7b1744dded53944)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 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/7224e45fd82116f9ac78/interpretations.jsonl b/records/sd1/sweep/7224e45fd82116f9ac78/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..3e91df22557fd44669f49703c033493d948e094c
--- /dev/null
+++ b/records/sd1/sweep/7224e45fd82116f9ac78/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7224e45fd82116f9ac78"}, "kind": "caption", "value": "Crisper anime-style portrait with larger dark eyes, red lips and smoother hair", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cbf01b3d90d514e4"}
+{"target": {"record": "7224e45fd82116f9ac78"}, "kind": "change", "value": "The face becomes more polished and anime-like, with bigger darker eyes, brighter red lips and smoother, glossier hair against a cleaner grey ground.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d652906ae0062eaa"}
+{"target": {"record": "7224e45fd82116f9ac78"}, "kind": "keywords", "value": ["anime style", "big eyes", "glossy hair", "polished portrait", "red lips"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "eb174be68c3bf2f2"}
+{"target": {"record": "7224e45fd82116f9ac78"}, "kind": "effect_tags", "value": ["subject_kept", "style_change", "saturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:33:49Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1d30ef6410cdc534"}
diff --git a/records/sd1/sweep/7224e45fd82116f9ac78/measurements.json b/records/sd1/sweep/7224e45fd82116f9ac78/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1e6f738d4c464b6c49b6878261bfff1b5ae9172
--- /dev/null
+++ b/records/sd1/sweep/7224e45fd82116f9ac78/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7224e45fd82116f9ac78",
+ "values": {
+  "cosine_distance": {
+   "value": 0.0864,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.122633,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.04182,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023833,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.038,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00591,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.109,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9714,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2132,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7224e45fd82116f9ac78/record.json b/records/sd1/sweep/7224e45fd82116f9ac78/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..08512cd349bb13e7680bfb9a5819902bddddfbef
--- /dev/null
+++ b/records/sd1/sweep/7224e45fd82116f9ac78/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7224e45fd82116f9ac78",
+ "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.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/0c45bbd2672a901d628df87350f994038f91939d54c985d68f623775f7c0b72a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0c45bbd2672a901d628df87350f994038f91939d54c985d68f623775f7c0b72a",
+  "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": "cd16e6bef037a96fe498"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7224e45fd82116f9ac78/workflow.json b/records/sd1/sweep/7224e45fd82116f9ac78/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fddee90f3f86b6d0311d9b40cfd6b262ad41fec8
--- /dev/null
+++ b/records/sd1/sweep/7224e45fd82116f9ac78/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7224e45fd82116f9ac78)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7227c36d039101dd9afe/measurements.json b/records/sd1/sweep/7227c36d039101dd9afe/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..60740b7771aa410588b8a5225f70ce92c357b6e7
--- /dev/null
+++ b/records/sd1/sweep/7227c36d039101dd9afe/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "7227c36d039101dd9afe",
+ "values": {
+  "cosine_distance": {
+   "value": 0.052363,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.088044,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.049699,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.020232,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.714,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00524,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.841,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9481,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.168,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7227c36d039101dd9afe/record.json b/records/sd1/sweep/7227c36d039101dd9afe/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb73d61fdb988fbb994d9fee3c64d1cda79bedd5
--- /dev/null
+++ b/records/sd1/sweep/7227c36d039101dd9afe/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7227c36d039101dd9afe",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.proj_out",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/64c32240e10daaf24230b62526f2e9b2bfe98deee379060c0be0dafe5ca138cb.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "64c32240e10daaf24230b62526f2e9b2bfe98deee379060c0be0dafe5ca138cb",
+  "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": "ad7b9649fbe5b80a3e79"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7227c36d039101dd9afe/workflow.json b/records/sd1/sweep/7227c36d039101dd9afe/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1d70e55a5b43fa10d4e735f1149362be990ed972
--- /dev/null
+++ b/records/sd1/sweep/7227c36d039101dd9afe/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7227c36d039101dd9afe)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.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/722b91ac38a78296ccce/measurements.json b/records/sd1/sweep/722b91ac38a78296ccce/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1e27a8468c4bb0b84631896ae4b0b37409c3046e
--- /dev/null
+++ b/records/sd1/sweep/722b91ac38a78296ccce/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "722b91ac38a78296ccce",
+ "values": {
+  "cosine_distance": {
+   "value": 0.004505,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.009271,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.003593,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0017,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.841,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00574,
+   "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": 3.013,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0218,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1462,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/722b91ac38a78296ccce/record.json b/records/sd1/sweep/722b91ac38a78296ccce/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fa28e7806c72b76b7c8d5951569c93a428f857d1
--- /dev/null
+++ b/records/sd1/sweep/722b91ac38a78296ccce/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "722b91ac38a78296ccce",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.1.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": "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/e16c00c395531c9728f2a88f2a2345b32b143d2508c5d429fe712c527ef85209.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e16c00c395531c9728f2a88f2a2345b32b143d2508c5d429fe712c527ef85209",
+  "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": "1360d6dc4db25525e5e3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/722b91ac38a78296ccce/workflow.json b/records/sd1/sweep/722b91ac38a78296ccce/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d94a81f80715be391b30097a4f55496d00579a7d
--- /dev/null
+++ b/records/sd1/sweep/722b91ac38a78296ccce/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 722b91ac38a78296ccce)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.transformer_blocks.0.attn2.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7234264bbaf2638f81b4/measurements.json b/records/sd1/sweep/7234264bbaf2638f81b4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..503e22fc4e5c6d46dbaa8086ebc9ca0ed54fe1ad
--- /dev/null
+++ b/records/sd1/sweep/7234264bbaf2638f81b4/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "7234264bbaf2638f81b4",
+ "values": {
+  "std": {
+   "value": 90.486,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0014,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": true,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": [
+    "clip_degenerate"
+   ]
+  },
+  "mae_vs_baseline": {
+   "value": 82.948,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.721504,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.746289,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 1.006238,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.421794,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.4541,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9064,
+   "method": "CLIP zero-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/7234264bbaf2638f81b4/record.json b/records/sd1/sweep/7234264bbaf2638f81b4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..16622c1e80a8130ccfd98dab632af231443e9f55
--- /dev/null
+++ b/records/sd1/sweep/7234264bbaf2638f81b4/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "7234264bbaf2638f81b4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:31:49Z",
+ "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": 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": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "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",
+   "scale",
+   0.7,
+   "all"
+  ],
+  "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": "a0d4eae54ec228b78f68"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7234264bbaf2638f81b4/workflow.json b/records/sd1/sweep/7234264bbaf2638f81b4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7094e11f3ca47238a7857b19f85444177d2fe506
--- /dev/null
+++ b/records/sd1/sweep/7234264bbaf2638f81b4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7234264bbaf2638f81b4)"
+  },
+  "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}",
+   "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/724be535c1a663c55995/measurements.json b/records/sd1/sweep/724be535c1a663c55995/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b790520471a9db143cea7e625e909357327f6fb
--- /dev/null
+++ b/records/sd1/sweep/724be535c1a663c55995/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "724be535c1a663c55995",
+ "values": {
+  "cosine_distance": {
+   "value": 0.538704,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.805378,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.874186,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.407909,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 27.455,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.06111,
+   "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": 43.423,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8899,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/724be535c1a663c55995/record.json b/records/sd1/sweep/724be535c1a663c55995/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..022eb257dc183adc67f8403246d70f528a41d732
--- /dev/null
+++ b/records/sd1/sweep/724be535c1a663c55995/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "724be535c1a663c55995",
+ "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.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "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/29eda81540c6ac3c7fb0e79bd5d9c9e22444df1ef9a89ee750e6055e4341295b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "29eda81540c6ac3c7fb0e79bd5d9c9e22444df1ef9a89ee750e6055e4341295b",
+  "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": "1fb7f3238f3109b220ab"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/724be535c1a663c55995/workflow.json b/records/sd1/sweep/724be535c1a663c55995/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..198b30c1d0922d532fe241067926d727fb9bece3
--- /dev/null
+++ b/records/sd1/sweep/724be535c1a663c55995/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 724be535c1a663c55995)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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/724bed0c272075060cf2/measurements.json b/records/sd1/sweep/724bed0c272075060cf2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b189231604c040b6e196366eab31232100ba2200
--- /dev/null
+++ b/records/sd1/sweep/724bed0c272075060cf2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "724bed0c272075060cf2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.300233,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.363018,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.263547,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.117142,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.894,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00255,
+   "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": 36.712,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0171,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.217,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/724bed0c272075060cf2/record.json b/records/sd1/sweep/724bed0c272075060cf2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2cdd2010c512883cd1905774858e274ce5683a6e
--- /dev/null
+++ b/records/sd1/sweep/724bed0c272075060cf2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "724bed0c272075060cf2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:19Z",
+ "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": "input_blocks.5.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/5078b045575ee684f44ede0df4dc0a052929e5559b51f639a1f6d68c6e1d4ce9.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5078b045575ee684f44ede0df4dc0a052929e5559b51f639a1f6d68c6e1d4ce9",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Dropout",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "aa16152deed950f49bb7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/724bed0c272075060cf2/workflow.json b/records/sd1/sweep/724bed0c272075060cf2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b56d0ceb00468d65baeb2a14d8af67a0d3aaf715
--- /dev/null
+++ b/records/sd1/sweep/724bed0c272075060cf2/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 724bed0c272075060cf2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/725c5417782018834dcb/measurements.json b/records/sd1/sweep/725c5417782018834dcb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..214872e2cadac103950d18db24098347bacda83c
--- /dev/null
+++ b/records/sd1/sweep/725c5417782018834dcb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "725c5417782018834dcb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.382023,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.431027,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.265168,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.145247,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.572,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00636,
+   "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": 43.338,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0112,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2271,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/725c5417782018834dcb/record.json b/records/sd1/sweep/725c5417782018834dcb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..29361918818e296f8fd9ff57e0aae9add4502f46
--- /dev/null
+++ b/records/sd1/sweep/725c5417782018834dcb/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "725c5417782018834dcb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:16Z",
+ "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": "input_blocks.4.0.out_layers.3",
+   "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": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/1b665bd219eee3e0d9365eaf292e52d602f3df59b13c8828558156fe6d6a802f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1b665bd219eee3e0d9365eaf292e52d602f3df59b13c8828558156fe6d6a802f",
+  "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": "ee6c7375dc8aa5107038"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/725c5417782018834dcb/workflow.json b/records/sd1/sweep/725c5417782018834dcb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..88388a3b9071da993ebbddada0a85a0d94e5f466
--- /dev/null
+++ b/records/sd1/sweep/725c5417782018834dcb/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 725c5417782018834dcb)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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/7262e0ce4373c27df4ee/measurements.json b/records/sd1/sweep/7262e0ce4373c27df4ee/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ff8d85434fc1f2cfb0969ea3537f6e18dff0e73
--- /dev/null
+++ b/records/sd1/sweep/7262e0ce4373c27df4ee/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7262e0ce4373c27df4ee",
+ "values": {
+  "std": {
+   "value": 37.677,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01474,
+   "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": 6.584,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.0271,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.07327,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.01726,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.005416,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0071,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0125,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7262e0ce4373c27df4ee/record.json b/records/sd1/sweep/7262e0ce4373c27df4ee/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..443c77426402ee0ddd0afe5785d8622c21b1bb5c
--- /dev/null
+++ b/records/sd1/sweep/7262e0ce4373c27df4ee/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "7262e0ce4373c27df4ee",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:01:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "medium"
+  }
+ ],
+ "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",
+   "add_noise",
+   0.2,
+   "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": "4e35a87e1461bcf31067"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7262e0ce4373c27df4ee/workflow.json b/records/sd1/sweep/7262e0ce4373c27df4ee/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..47e1c42791f2dd6672de2f18deb9930cafbd363b
--- /dev/null
+++ b/records/sd1/sweep/7262e0ce4373c27df4ee/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7262e0ce4373c27df4ee)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 1234\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7265998521c57e0c18b8/interpretations.jsonl b/records/sd1/sweep/7265998521c57e0c18b8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..4ba26c73c310dba447d0d298dbd249cca5aceff5
--- /dev/null
+++ b/records/sd1/sweep/7265998521c57e0c18b8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "7265998521c57e0c18b8"}, "kind": "caption", "value": "Portrait of woman with wavy dark hair and turtleneck, slightly more muted tones", "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": "f9374af497455acd"}
+{"target": {"record": "7265998521c57e0c18b8"}, "kind": "change", "value": "Almost no visible change; the face is a touch softer and the colours slightly more muted.", "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": "2e5c9c03bca73385"}
+{"target": {"record": "7265998521c57e0c18b8"}, "kind": "keywords", "value": ["nearly identical", "muted", "minimal change", "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": "4c47c9fa0845a700"}
+{"target": {"record": "7265998521c57e0c18b8"}, "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": "c71b93f72fb84a5b"}
diff --git a/records/sd1/sweep/7265998521c57e0c18b8/measurements.json b/records/sd1/sweep/7265998521c57e0c18b8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b1a8893f8490335dc84121fceab44d067f26e42
--- /dev/null
+++ b/records/sd1/sweep/7265998521c57e0c18b8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7265998521c57e0c18b8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.086035,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.119999,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.063043,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027598,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.341,
+   "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": 13.009,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9568,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1073,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7265998521c57e0c18b8/record.json b/records/sd1/sweep/7265998521c57e0c18b8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc6267ff550a711740a37cfb8e8ce2684709fb1a
--- /dev/null
+++ b/records/sd1/sweep/7265998521c57e0c18b8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7265998521c57e0c18b8",
+ "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": 1.5
+   },
+   "group": "mid",
+   "kind": "ff",
+   "module_type": "Linear",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/8d490a8b6bbf5bc6ba83516f3d42d6bcf597bb083c9911523362009d94c921ce.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8d490a8b6bbf5bc6ba83516f3d42d6bcf597bb083c9911523362009d94c921ce",
+  "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": "e315f9f45354422c34b5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7265998521c57e0c18b8/workflow.json b/records/sd1/sweep/7265998521c57e0c18b8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..28a584955f905fae4e6db048b9de8a439cd1e14a
--- /dev/null
+++ b/records/sd1/sweep/7265998521c57e0c18b8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7265998521c57e0c18b8)"
+  },
+  "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\": 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/72698a0f4b6f9a14f5e3/interpretations.jsonl b/records/sd1/sweep/72698a0f4b6f9a14f5e3/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ce667d77536b53abb283b0b54f264364915ad44a
--- /dev/null
+++ b/records/sd1/sweep/72698a0f4b6f9a14f5e3/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72698a0f4b6f9a14f5e3"}, "kind": "caption", "value": "Woman with long brown hair resting her chin on her hand, framed print on teal", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e08b2837cd9c8a8e"}
+{"target": {"record": "72698a0f4b6f9a14f5e3"}, "kind": "change", "value": "A new pose: a long-haired woman tilts her head and leans on her hand in a grey knit sweater, with a teal print frame replacing the grey one.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "018380665bda4f56"}
+{"target": {"record": "72698a0f4b6f9a14f5e3"}, "kind": "keywords", "value": ["hand on chin", "long hair", "teal frame", "pose change", "film portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "46245a47e0a3307c"}
+{"target": {"record": "72698a0f4b6f9a14f5e3"}, "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": "2026-10-05T08:08:15Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f19b925be1e693d0"}
diff --git a/records/sd1/sweep/72698a0f4b6f9a14f5e3/measurements.json b/records/sd1/sweep/72698a0f4b6f9a14f5e3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd516fa03c22c1f31d03d77a3b01f94ef8986ba8
--- /dev/null
+++ b/records/sd1/sweep/72698a0f4b6f9a14f5e3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72698a0f4b6f9a14f5e3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.344539,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.39671,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.253027,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.18666,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.981,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00526,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.663,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9964,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0912,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72698a0f4b6f9a14f5e3/record.json b/records/sd1/sweep/72698a0f4b6f9a14f5e3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9ed07081a68d8f7384449bf1e5efccc5eb5007b1
--- /dev/null
+++ b/records/sd1/sweep/72698a0f4b6f9a14f5e3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72698a0f4b6f9a14f5e3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:30Z",
+ "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": "input_blocks.9.0.op",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.lo",
+   "kind": "downsample",
+   "module_type": "Conv2d",
+   "container": "Downsample",
+   "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/300db4e843bc9e32d0b846d32b2e1f00b19ac596e12e9ef39b1f55fe8a730c65.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "300db4e843bc9e32d0b846d32b2e1f00b19ac596e12e9ef39b1f55fe8a730c65",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.Downsample.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "11addcb1f3eaa6414254"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72698a0f4b6f9a14f5e3/workflow.json b/records/sd1/sweep/72698a0f4b6f9a14f5e3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6fddbc4d1d5a99399683e5648563e844b3abe833
--- /dev/null
+++ b/records/sd1/sweep/72698a0f4b6f9a14f5e3/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 72698a0f4b6f9a14f5e3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.9.0.op\",\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/72810a4465bc5e7e0c3b/measurements.json b/records/sd1/sweep/72810a4465bc5e7e0c3b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d2a722c9c8c4f1422a053dba5fb8e0c103ed983
--- /dev/null
+++ b/records/sd1/sweep/72810a4465bc5e7e0c3b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72810a4465bc5e7e0c3b",
+ "values": {
+  "std": {
+   "value": 29.601,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01689,
+   "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": 29.11,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.263871,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.378409,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.328961,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.136006,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0379,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0419,
+   "method": "CLIP zero-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/72810a4465bc5e7e0c3b/record.json b/records/sd1/sweep/72810a4465bc5e7e0c3b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0420b9321322fea6b531133e9024d18ecf322dea
--- /dev/null
+++ b/records/sd1/sweep/72810a4465bc5e7e0c3b/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "72810a4465bc5e7e0c3b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:17:44Z",
+ "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 still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   0.5,
+   "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": "c2d3216647cd67dee097"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72810a4465bc5e7e0c3b/workflow.json b/records/sd1/sweep/72810a4465bc5e7e0c3b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b16035c866cc6a8aae81c00917ac476422cb311e
--- /dev/null
+++ b/records/sd1/sweep/72810a4465bc5e7e0c3b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72810a4465bc5e7e0c3b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "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/728aa7860c96aa024786/measurements.json b/records/sd1/sweep/728aa7860c96aa024786/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b214767c614c739e7641bb7a20ce26363897f8d9
--- /dev/null
+++ b/records/sd1/sweep/728aa7860c96aa024786/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "728aa7860c96aa024786",
+ "values": {
+  "cosine_distance": {
+   "value": 0.236142,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.483133,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.303206,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.155033,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 53.345,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00222,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.948,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0563,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4691,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728aa7860c96aa024786/record.json b/records/sd1/sweep/728aa7860c96aa024786/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab71c8fc3cab4a4e2a127c64899ba70575964072
--- /dev/null
+++ b/records/sd1/sweep/728aa7860c96aa024786/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "728aa7860c96aa024786",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/33188e4a56aa8c9f63bf556db6b8f0cceabed09d49b4626592dca4d810814192.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "33188e4a56aa8c9f63bf556db6b8f0cceabed09d49b4626592dca4d810814192",
+  "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": "322e69e6c4b6242e205f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728aa7860c96aa024786/workflow.json b/records/sd1/sweep/728aa7860c96aa024786/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bf3b80d5bbe598b6bf4f7c902e8e44c9efce5f8f
--- /dev/null
+++ b/records/sd1/sweep/728aa7860c96aa024786/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 728aa7860c96aa024786)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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": [
+    "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/728beea8ead44b3f411a/measurements.json b/records/sd1/sweep/728beea8ead44b3f411a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a6c021365362d000adf986382f9edf9d55278a9c
--- /dev/null
+++ b/records/sd1/sweep/728beea8ead44b3f411a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "728beea8ead44b3f411a",
+ "values": {
+  "std": {
+   "value": 42.192,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00508,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.92,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.169675,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.580749,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.670845,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.322499,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8994,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0169,
+   "method": "CLIP zero-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/728beea8ead44b3f411a/record.json b/records/sd1/sweep/728beea8ead44b3f411a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3cffac6d98d76aaddf5ef78933d008482a615d18
--- /dev/null
+++ b/records/sd1/sweep/728beea8ead44b3f411a/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "728beea8ead44b3f411a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:16:55Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "zero",
+   "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.hi",
+   "multiply",
+   0.0,
+   "late"
+  ],
+  "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": "4c100db2dc1c2fa25de8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728beea8ead44b3f411a/workflow.json b/records/sd1/sweep/728beea8ead44b3f411a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f0921110a422ebe35fce9661ca7bbe0bfb908d6
--- /dev/null
+++ b/records/sd1/sweep/728beea8ead44b3f411a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 728beea8ead44b3f411a)"
+  },
+  "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\": 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/728eb170171a61d87cc3/measurements.json b/records/sd1/sweep/728eb170171a61d87cc3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8c8b719e4e934ab8008ad11d10bc52684b907e9
--- /dev/null
+++ b/records/sd1/sweep/728eb170171a61d87cc3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "728eb170171a61d87cc3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.048539,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.062967,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.02703,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.012039,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.046,
+   "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": 14.223,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0211,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1651,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728eb170171a61d87cc3/record.json b/records/sd1/sweep/728eb170171a61d87cc3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..604b78e764a005f74d137ed5693bdaeb8ed97160
--- /dev/null
+++ b/records/sd1/sweep/728eb170171a61d87cc3/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "728eb170171a61d87cc3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:29Z",
+ "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": "output_blocks.11.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.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/9f22392b7e9dcdcfea8a216c3cc14223557563b9adff6e8b2d78f7841498830e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9f22392b7e9dcdcfea8a216c3cc14223557563b9adff6e8b2d78f7841498830e",
+  "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": "f0bf729ad47817bb0231"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728eb170171a61d87cc3/workflow.json b/records/sd1/sweep/728eb170171a61d87cc3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b2989cc36238a9a0ae22269d45ba1ebbcdfe7b6
--- /dev/null
+++ b/records/sd1/sweep/728eb170171a61d87cc3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 728eb170171a61d87cc3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.attn1.to_out.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728faa0d7b35958a4a45/measurements.json b/records/sd1/sweep/728faa0d7b35958a4a45/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..44f4e2be74116f3efc051900ebd37b724e7135fd
--- /dev/null
+++ b/records/sd1/sweep/728faa0d7b35958a4a45/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "728faa0d7b35958a4a45",
+ "values": {
+  "cosine_distance": {
+   "value": 0.121057,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.174221,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.085215,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.151377,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.992,
+   "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": 15.873,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9444,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5931,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728faa0d7b35958a4a45/record.json b/records/sd1/sweep/728faa0d7b35958a4a45/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..20607f00a329febaf4fc34e87224bca3d4dd50ba
--- /dev/null
+++ b/records/sd1/sweep/728faa0d7b35958a4a45/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "728faa0d7b35958a4a45",
+ "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.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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/a39a1de926175cb85d3a508bdaf969b0878f513eb6f81b1fa8bbb855b704b9b0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a39a1de926175cb85d3a508bdaf969b0878f513eb6f81b1fa8bbb855b704b9b0",
+  "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": "8a87c90bf219b7c356bf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/728faa0d7b35958a4a45/workflow.json b/records/sd1/sweep/728faa0d7b35958a4a45/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae03d13010aa30359b8512d3831c02a972a22509
--- /dev/null
+++ b/records/sd1/sweep/728faa0d7b35958a4a45/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 728faa0d7b35958a4a45)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.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/7297142e874b96245a0e/measurements.json b/records/sd1/sweep/7297142e874b96245a0e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..251611a18e348a5dd60eec6b5ff34f3907a337b7
--- /dev/null
+++ b/records/sd1/sweep/7297142e874b96245a0e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "7297142e874b96245a0e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.207011,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.319935,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.114707,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.06267,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.834,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00649,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.308,
+   "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": 0.1072,
+   "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/7297142e874b96245a0e/record.json b/records/sd1/sweep/7297142e874b96245a0e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..81becb81badf9ce01a423c7dfaffa49d3ad71cf6
--- /dev/null
+++ b/records/sd1/sweep/7297142e874b96245a0e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "7297142e874b96245a0e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:54Z",
+ "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.5.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "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/947e6f0d04ef394b135e3637208d58396afcd6ea4fbe4d27b18d733e864110ae.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "947e6f0d04ef394b135e3637208d58396afcd6ea4fbe4d27b18d733e864110ae",
+  "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": "076ff5722cc65dff4499"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/7297142e874b96245a0e/workflow.json b/records/sd1/sweep/7297142e874b96245a0e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..72df5f9a98f8510704f5216956e4277e5a1d938f
--- /dev/null
+++ b/records/sd1/sweep/7297142e874b96245a0e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 7297142e874b96245a0e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.in_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/729ca7d47084c875cc74/measurements.json b/records/sd1/sweep/729ca7d47084c875cc74/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..73370bc071b5049a6466bf8971711b5129842488
--- /dev/null
+++ b/records/sd1/sweep/729ca7d47084c875cc74/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "729ca7d47084c875cc74",
+ "values": {
+  "std": {
+   "value": 57.133,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00773,
+   "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": 18.164,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.112542,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.367302,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.486002,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.184716,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9245,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6207,
+   "method": "CLIP zero-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/729ca7d47084c875cc74/record.json b/records/sd1/sweep/729ca7d47084c875cc74/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..83e0826e8890888932f6eba4c8f573b74f5664ef
--- /dev/null
+++ b/records/sd1/sweep/729ca7d47084c875cc74/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "729ca7d47084c875cc74",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:50:11Z",
+ "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": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "zero",
+   "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": [
+   "out.mid",
+   "multiply",
+   0.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] 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": "c2975cb8e7522ff72140"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/729ca7d47084c875cc74/workflow.json b/records/sd1/sweep/729ca7d47084c875cc74/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b61afe5d7f44b9fc8fb4b43a5e42af143f4ef95
--- /dev/null
+++ b/records/sd1/sweep/729ca7d47084c875cc74/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 729ca7d47084c875cc74)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\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/72a3f829435102271dc3/measurements.json b/records/sd1/sweep/72a3f829435102271dc3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd7d9aabf6031e861a897515d65f5f5cbdaabfc0
--- /dev/null
+++ b/records/sd1/sweep/72a3f829435102271dc3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72a3f829435102271dc3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.397442,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.473137,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.265245,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.094119,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 78.198,
+   "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": 51.9,
+   "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": 0.251,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72a3f829435102271dc3/record.json b/records/sd1/sweep/72a3f829435102271dc3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..837436a6fae3fdb02c1449317b6b85091e3a301e
--- /dev/null
+++ b/records/sd1/sweep/72a3f829435102271dc3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72a3f829435102271dc3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:07Z",
+ "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": "input_blocks.1.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.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/92242c97804e446b0d53f90781961edd3b3a8103dc61fd58ac32c8f0b6835420.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "92242c97804e446b0d53f90781961edd3b3a8103dc61fd58ac32c8f0b6835420",
+  "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": "f4b6e4810e9d7d991dfa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72a3f829435102271dc3/workflow.json b/records/sd1/sweep/72a3f829435102271dc3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7fbb7860e0fac060a7cb6bffc9d48734ce476125
--- /dev/null
+++ b/records/sd1/sweep/72a3f829435102271dc3/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 72a3f829435102271dc3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/72a6e954b32daef7bd3b/measurements.json b/records/sd1/sweep/72a6e954b32daef7bd3b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c4a72e539ea6949faca24052b70ed706a77a4c5c
--- /dev/null
+++ b/records/sd1/sweep/72a6e954b32daef7bd3b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72a6e954b32daef7bd3b",
+ "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/72a6e954b32daef7bd3b/record.json b/records/sd1/sweep/72a6e954b32daef7bd3b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a8f395aee629f12a3a4e966a1051539905f0389
--- /dev/null
+++ b/records/sd1/sweep/72a6e954b32daef7bd3b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72a6e954b32daef7bd3b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/ca5403acfc98b96391ebc5a233120c67a219868af66796cda48aa58b456d5955.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "ca5403acfc98b96391ebc5a233120c67a219868af66796cda48aa58b456d5955",
+  "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": "b1679fcf5d0fc249b0ad"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72a6e954b32daef7bd3b/workflow.json b/records/sd1/sweep/72a6e954b32daef7bd3b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bee652593ae286a5e4749154fceb2df261cea878
--- /dev/null
+++ b/records/sd1/sweep/72a6e954b32daef7bd3b/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 72a6e954b32daef7bd3b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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": [
+    "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/72a7f9f209f54568febc/interpretations.jsonl b/records/sd1/sweep/72a7f9f209f54568febc/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..b71e05bba5ed65cff29171fd7f6aec3ff605bcbc
--- /dev/null
+++ b/records/sd1/sweep/72a7f9f209f54568febc/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72a7f9f209f54568febc"}, "kind": "caption", "value": "Painted portrait of a woman with dark brown bob and collared black top on grey", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6732fade43b12999"}
+{"target": {"record": "72a7f9f209f54568febc"}, "kind": "change", "value": "The hair becomes darker and shorter-looking, the background flattens to plain grey without the border, and the face is rendered more painterly with slightly altered features.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c48ae45d2602e495"}
+{"target": {"record": "72a7f9f209f54568febc"}, "kind": "keywords", "value": ["dark bob", "grey background", "painted portrait", "collared top"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ad83906b9d954daa"}
+{"target": {"record": "72a7f9f209f54568febc"}, "kind": "effect_tags", "value": ["subject_kept", "painterly", "subtle", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:20:58Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7b9b91d509b0c701"}
diff --git a/records/sd1/sweep/72a7f9f209f54568febc/measurements.json b/records/sd1/sweep/72a7f9f209f54568febc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb74075c4d81c33678744f990c6dd489ae1e6567
--- /dev/null
+++ b/records/sd1/sweep/72a7f9f209f54568febc/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72a7f9f209f54568febc",
+ "values": {
+  "cosine_distance": {
+   "value": 0.283598,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.335528,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.188005,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.05606,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.883,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00488,
+   "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": 34.234,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9662,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0743,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72a7f9f209f54568febc/record.json b/records/sd1/sweep/72a7f9f209f54568febc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..19a745a2744086417b98dc3ec46bed62ca724162
--- /dev/null
+++ b/records/sd1/sweep/72a7f9f209f54568febc/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "72a7f9f209f54568febc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:08Z",
+ "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": "input_blocks.4.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7cd8d2935b521ddbbdd163a246b5dd9201d0b49379566d1a46380e1b63fc6eed.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7cd8d2935b521ddbbdd163a246b5dd9201d0b49379566d1a46380e1b63fc6eed",
+  "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": "48a0bcc87bd850509acd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72a7f9f209f54568febc/workflow.json b/records/sd1/sweep/72a7f9f209f54568febc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2e34129686231aee3da5cbc14d5ef0a8d82e77d
--- /dev/null
+++ b/records/sd1/sweep/72a7f9f209f54568febc/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72a7f9f209f54568febc)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72b406513e57170e69aa/interpretations.jsonl b/records/sd1/sweep/72b406513e57170e69aa/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..614a6ca5a01c57fcdf1135eea16f1922a692597a
--- /dev/null
+++ b/records/sd1/sweep/72b406513e57170e69aa/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72b406513e57170e69aa"}, "kind": "caption", "value": "Frontal portrait of a person with slicked-back hair and calm gaze on grey", "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": "736a6390fc26ea6b"}
+{"target": {"record": "72b406513e57170e69aa"}, "kind": "change", "value": "The subject turns frontal and more androgynous with slicked-back hair, keeping the muted palette and frame.", "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": "a945ec57b7b95578"}
+{"target": {"record": "72b406513e57170e69aa"}, "kind": "keywords", "value": ["frontal portrait", "slicked hair", "androgynous", "grey backdrop"], "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": "0835c8195a71b96c"}
+{"target": {"record": "72b406513e57170e69aa"}, "kind": "effect_tags", "value": ["subject_kept", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e238fcec57c761ff"}
diff --git a/records/sd1/sweep/72b406513e57170e69aa/measurements.json b/records/sd1/sweep/72b406513e57170e69aa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..83bf5d295b964efe380342619a92b1930f3f6a15
--- /dev/null
+++ b/records/sd1/sweep/72b406513e57170e69aa/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72b406513e57170e69aa",
+ "values": {
+  "cosine_distance": {
+   "value": 0.153237,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.182488,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.145807,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.167774,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.521,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00394,
+   "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": 17.488,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.931,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.603,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72b406513e57170e69aa/record.json b/records/sd1/sweep/72b406513e57170e69aa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e323c4facb19cc1771bd9a9c95505d2fa8fbbc1
--- /dev/null
+++ b/records/sd1/sweep/72b406513e57170e69aa/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72b406513e57170e69aa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.in_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.mid",
+   "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/826c891b4c8d5935eec2d5bdccf9bef1188bd7e08378e2391ed082dc9ba36e66.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "826c891b4c8d5935eec2d5bdccf9bef1188bd7e08378e2391ed082dc9ba36e66",
+  "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": "ed3e1426ed6f455b2040"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72b406513e57170e69aa/workflow.json b/records/sd1/sweep/72b406513e57170e69aa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1c72acf33a6ec2170bb4f8b3441ce1d4b688f7b
--- /dev/null
+++ b/records/sd1/sweep/72b406513e57170e69aa/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 72b406513e57170e69aa)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.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/72ba0e5fecc77ad396e9/measurements.json b/records/sd1/sweep/72ba0e5fecc77ad396e9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4007c607c12b4772800f2b7176571c04bdfdd379
--- /dev/null
+++ b/records/sd1/sweep/72ba0e5fecc77ad396e9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72ba0e5fecc77ad396e9",
+ "values": {
+  "std": {
+   "value": 51.005,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01314,
+   "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": 20.78,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.183747,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.226951,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.064891,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.037442,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0245,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1208,
+   "method": "CLIP zero-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/72ba0e5fecc77ad396e9/record.json b/records/sd1/sweep/72ba0e5fecc77ad396e9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e7b68aa10777e2a66fcfc15e0311e4fd31b3aea
--- /dev/null
+++ b/records/sd1/sweep/72ba0e5fecc77ad396e9/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "72ba0e5fecc77ad396e9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:20:22Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "heavy"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "add_noise",
+   0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2c79403fa824b8b9291a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72ba0e5fecc77ad396e9/workflow.json b/records/sd1/sweep/72ba0e5fecc77ad396e9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a041e8d37b157b1cbe5232b04f6079f1851263f
--- /dev/null
+++ b/records/sd1/sweep/72ba0e5fecc77ad396e9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72ba0e5fecc77ad396e9)"
+  },
+  "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}",
+   "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/72bded680bb6c36aace2/measurements.json b/records/sd1/sweep/72bded680bb6c36aace2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d45c05a237afd725ad71ab212896adc46411e20
--- /dev/null
+++ b/records/sd1/sweep/72bded680bb6c36aace2/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72bded680bb6c36aace2",
+ "values": {
+  "cosine_distance": {
+   "value": 0.571613,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.564243,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.72394,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.344408,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.263,
+   "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": 54.408,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8824,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6964,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72bded680bb6c36aace2/record.json b/records/sd1/sweep/72bded680bb6c36aace2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e447368829993d6af30262a7e3a23eba9c0d9d4
--- /dev/null
+++ b/records/sd1/sweep/72bded680bb6c36aace2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72bded680bb6c36aace2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:13Z",
+ "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.2.0.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": "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/f6585aeec2517d26198f5c4b6b17fb363cb6ff2959a765d6e7a07ca68839b8c7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f6585aeec2517d26198f5c4b6b17fb363cb6ff2959a765d6e7a07ca68839b8c7",
+  "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": "ef52c8b7434caece5ae0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72bded680bb6c36aace2/workflow.json b/records/sd1/sweep/72bded680bb6c36aace2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..22e95452affd577a7906fa66456c6b1a3547c7e7
--- /dev/null
+++ b/records/sd1/sweep/72bded680bb6c36aace2/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 72bded680bb6c36aace2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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/72bfa37bd610aa6d1c17/interpretations.jsonl b/records/sd1/sweep/72bfa37bd610aa6d1c17/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5e67de5d0aba7eacd7d9863e4d660be226f7ca8f
--- /dev/null
+++ b/records/sd1/sweep/72bfa37bd610aa6d1c17/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72bfa37bd610aa6d1c17"}, "kind": "caption", "value": "Dark hazy neon street with glowing red and cyan signs blurred in mist", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0b2405c896e6fde6"}
+{"target": {"record": "72bfa37bd610aa6d1c17"}, "kind": "change", "value": "The neon street becomes dark, misty and smudged; signs turn into soft glowing patches and pedestrians disappear into murky teal shadows.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6ccf07f8a43845de"}
+{"target": {"record": "72bfa37bd610aa6d1c17"}, "kind": "keywords", "value": ["hazy neon", "foggy city night", "smudged signs", "dark street", "glowing blobs"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "315d8e6a68189011"}
+{"target": {"record": "72bfa37bd610aa6d1c17"}, "kind": "effect_tags", "value": ["blur", "dark", "painterly", "subject_kept", "texture"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:42Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2292014251f0b6ed"}
diff --git a/records/sd1/sweep/72bfa37bd610aa6d1c17/measurements.json b/records/sd1/sweep/72bfa37bd610aa6d1c17/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3466e7200e32f5d1eb5c117530f0c73f9e0d1eff
--- /dev/null
+++ b/records/sd1/sweep/72bfa37bd610aa6d1c17/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72bfa37bd610aa6d1c17",
+ "values": {
+  "std": {
+   "value": 26.741,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02012,
+   "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": 15.898,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.085469,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.281404,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.4761,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.116247,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9377,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2166,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/72bfa37bd610aa6d1c17/record.json b/records/sd1/sweep/72bfa37bd610aa6d1c17/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d84da490d6da1d08ac584ce1cca081be0d7b6d67
--- /dev/null
+++ b/records/sd1/sweep/72bfa37bd610aa6d1c17/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "72bfa37bd610aa6d1c17",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:28:24Z",
+ "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 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": 90.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "quarter",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "623f5902a3cda0d19664"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72bfa37bd610aa6d1c17/workflow.json b/records/sd1/sweep/72bfa37bd610aa6d1c17/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e4deea91702a0cb38039b787ef12ef58039f077
--- /dev/null
+++ b/records/sd1/sweep/72bfa37bd610aa6d1c17/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72bfa37bd610aa6d1c17)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 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/72c1ca8384a0712bffa6/interpretations.jsonl b/records/sd1/sweep/72c1ca8384a0712bffa6/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..02f55b6e0b733f5d1c25d1069401310fed1d5318
--- /dev/null
+++ b/records/sd1/sweep/72c1ca8384a0712bffa6/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72c1ca8384a0712bffa6"}, "kind": "caption", "value": "Old man with white beard and wool hat in front of a grey sea, nearly unchanged.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b5cba978153e5dbb"}
+{"target": {"record": "72c1ca8384a0712bffa6"}, "kind": "change", "value": "Almost no visible change; tiny differences in facial expression and beard.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c3f4711cba4720cc"}
+{"target": {"record": "72c1ca8384a0712bffa6"}, "kind": "keywords", "value": ["old man portrait", "wool hat", "seaside", "minimal change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:45:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "320f5ab9a47aa645"}
+{"target": {"record": "72c1ca8384a0712bffa6"}, "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": "2026-10-05T04:45:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "54f992964296500d"}
diff --git a/records/sd1/sweep/72c1ca8384a0712bffa6/measurements.json b/records/sd1/sweep/72c1ca8384a0712bffa6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ca10263e5e2485716cbc220ea1afa9c5adde013
--- /dev/null
+++ b/records/sd1/sweep/72c1ca8384a0712bffa6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72c1ca8384a0712bffa6",
+ "values": {
+  "std": {
+   "value": 58.76,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0103,
+   "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": 5.563,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.007483,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.023136,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.006663,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002166,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0047,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3673,
+   "method": "CLIP zero-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/72c1ca8384a0712bffa6/record.json b/records/sd1/sweep/72c1ca8384a0712bffa6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e4cb4f84a2f0744c78482b7177bab3fa6f86ecc2
--- /dev/null
+++ b/records/sd1/sweep/72c1ca8384a0712bffa6/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "72c1ca8384a0712bffa6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:37:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "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": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "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.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": "207fa4efdf7bcacaa912"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c1ca8384a0712bffa6/workflow.json b/records/sd1/sweep/72c1ca8384a0712bffa6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0427653218905a96d1b6eb67eb138a36b5e170a7
--- /dev/null
+++ b/records/sd1/sweep/72c1ca8384a0712bffa6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72c1ca8384a0712bffa6)"
+  },
+  "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": "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/72c4d9471ef372664a58/interpretations.jsonl b/records/sd1/sweep/72c4d9471ef372664a58/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..c3846f8de746239d65b56d0abd75a40b7369a7be
--- /dev/null
+++ b/records/sd1/sweep/72c4d9471ef372664a58/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72c4d9471ef372664a58"}, "kind": "caption", "value": "Nearly identical frontal portrait of a woman with wavy hair and black collared top", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a947d676c0a0be6d"}
+{"target": {"record": "72c4d9471ef372664a58"}, "kind": "change", "value": "Almost no visible change apart from slightly lighter hair, softer features and a collared shirt instead of a turtleneck.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "316d1ccd02afbecc"}
+{"target": {"record": "72c4d9471ef372664a58"}, "kind": "keywords", "value": ["near identical", "minor variation", "portrait", "collared shirt"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0c73769fcb2f2fbb"}
+{"target": {"record": "72c4d9471ef372664a58"}, "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": "2026-10-05T21:30:06Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e410c86d6b74739f"}
diff --git a/records/sd1/sweep/72c4d9471ef372664a58/measurements.json b/records/sd1/sweep/72c4d9471ef372664a58/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fccf41dedf3af0fcb00535b99cf3805e89dd41c8
--- /dev/null
+++ b/records/sd1/sweep/72c4d9471ef372664a58/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72c4d9471ef372664a58",
+ "values": {
+  "cosine_distance": {
+   "value": 0.095685,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.124001,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.12353,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.031419,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.956,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00508,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.661,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0438,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1441,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c4d9471ef372664a58/record.json b/records/sd1/sweep/72c4d9471ef372664a58/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f4e0772e0278320654c1bdcad0d7869f9c702df
--- /dev/null
+++ b/records/sd1/sweep/72c4d9471ef372664a58/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "72c4d9471ef372664a58",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:54Z",
+ "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.1.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1aa35423b37723efefae7f413c7ffe4d5e4a2bfec64bd51e8fdc49132fe7295e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1aa35423b37723efefae7f413c7ffe4d5e4a2bfec64bd51e8fdc49132fe7295e",
+  "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": "c4de57e368d166b9a304"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c4d9471ef372664a58/workflow.json b/records/sd1/sweep/72c4d9471ef372664a58/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..01d72c465391359b2fa2d25ca661999740bff4a4
--- /dev/null
+++ b/records/sd1/sweep/72c4d9471ef372664a58/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72c4d9471ef372664a58)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c727454b0eda006785/measurements.json b/records/sd1/sweep/72c727454b0eda006785/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5598d30b5dc7d86dbda5024f597e13cc54458ef1
--- /dev/null
+++ b/records/sd1/sweep/72c727454b0eda006785/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72c727454b0eda006785",
+ "values": {
+  "cosine_distance": {
+   "value": 0.48324,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.685501,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.802688,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.417065,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 18.062,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00942,
+   "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": 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": 44.527,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8295,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9704,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c727454b0eda006785/record.json b/records/sd1/sweep/72c727454b0eda006785/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..24480215ed66b4c0754991288ee336c8bd9e8c87
--- /dev/null
+++ b/records/sd1/sweep/72c727454b0eda006785/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72c727454b0eda006785",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/c8ecbd51a1db8dde05320c0f7c934668dc722ef85d6a401641bc18045c1a399d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c8ecbd51a1db8dde05320c0f7c934668dc722ef85d6a401641bc18045c1a399d",
+  "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": "f91fc6b77781f3f2f444"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c727454b0eda006785/workflow.json b/records/sd1/sweep/72c727454b0eda006785/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdf265e6b040822f402b1495d791c143bc76901f
--- /dev/null
+++ b/records/sd1/sweep/72c727454b0eda006785/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 72c727454b0eda006785)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.transformer_blocks.0.norm1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c8eca2005882a6f656/measurements.json b/records/sd1/sweep/72c8eca2005882a6f656/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..57011ba091e1fcaf6a742c205f44921e28cbe7e5
--- /dev/null
+++ b/records/sd1/sweep/72c8eca2005882a6f656/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72c8eca2005882a6f656",
+ "values": {
+  "std": {
+   "value": 47.198,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01384,
+   "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": 0.000616,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.002568,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000764,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000397,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0114,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2821,
+   "method": "CLIP zero-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/72c8eca2005882a6f656/record.json b/records/sd1/sweep/72c8eca2005882a6f656/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d91ed3bec0d4452a80eb7b985276aedf0a0a800
--- /dev/null
+++ b/records/sd1/sweep/72c8eca2005882a6f656/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "72c8eca2005882a6f656",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:29:59Z",
+ "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.6.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "mid",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "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,
+   "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": "73ad87ed6980d7060fc4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c8eca2005882a6f656/workflow.json b/records/sd1/sweep/72c8eca2005882a6f656/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b105de6a2baf65205c0193a72c9b0afc96c7ee5f
--- /dev/null
+++ b/records/sd1/sweep/72c8eca2005882a6f656/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72c8eca2005882a6f656)"
+  },
+  "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\": 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/72c9daa51b779cd41394/measurements.json b/records/sd1/sweep/72c9daa51b779cd41394/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9543795ad767dbdff006b1a207ec7b76a4b0206
--- /dev/null
+++ b/records/sd1/sweep/72c9daa51b779cd41394/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72c9daa51b779cd41394",
+ "values": {
+  "cosine_distance": {
+   "value": 0.434237,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.524176,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.299493,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.267482,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 41.897,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.09598,
+   "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": 42.389,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.2466,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c9daa51b779cd41394/record.json b/records/sd1/sweep/72c9daa51b779cd41394/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7972d54584e44a968992e9225d7e3b630efd14a3
--- /dev/null
+++ b/records/sd1/sweep/72c9daa51b779cd41394/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "72c9daa51b779cd41394",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:01Z",
+ "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.11.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.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/6d55322735250c3ec9e478eaefa13b6075507c579f5cf83a97112bea665cc2e7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6d55322735250c3ec9e478eaefa13b6075507c579f5cf83a97112bea665cc2e7",
+  "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": "73c3f5f8ea8f806fa1d6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72c9daa51b779cd41394/workflow.json b/records/sd1/sweep/72c9daa51b779cd41394/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c18e1b7c1741049e4ae8f8384646033360db788
--- /dev/null
+++ b/records/sd1/sweep/72c9daa51b779cd41394/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 72c9daa51b779cd41394)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.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/72d385765ac8eccfa890/interpretations.jsonl b/records/sd1/sweep/72d385765ac8eccfa890/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..10a2206d5f46bac473bfe510502b201c7f3a1baf
--- /dev/null
+++ b/records/sd1/sweep/72d385765ac8eccfa890/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72d385765ac8eccfa890"}, "kind": "caption", "value": "Old man in a tweed cap with a reddened face, sea background", "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": "afcd6afacaa61219"}
+{"target": {"record": "72d385765ac8eccfa890"}, "kind": "change", "value": "Almost no visible change apart from slightly redder, warmer skin and minor shifts in his features.", "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": "9f6e482e79617b2d"}
+{"target": {"record": "72d385765ac8eccfa890"}, "kind": "keywords", "value": ["old man portrait", "warmer skin", "minor variation"], "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": "1e9666c0f9d6fdb8"}
+{"target": {"record": "72d385765ac8eccfa890"}, "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": "0c77f4f72c0175fa"}
diff --git a/records/sd1/sweep/72d385765ac8eccfa890/measurements.json b/records/sd1/sweep/72d385765ac8eccfa890/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..71cc90bbe7fe5437228d860260cb9cd4ef711ed4
--- /dev/null
+++ b/records/sd1/sweep/72d385765ac8eccfa890/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72d385765ac8eccfa890",
+ "values": {
+  "std": {
+   "value": 60.051,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00995,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 8.401,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.020392,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.044866,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.009864,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.007658,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0114,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3514,
+   "method": "CLIP zero-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/72d385765ac8eccfa890/record.json b/records/sd1/sweep/72d385765ac8eccfa890/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d473b9ba768d65fd99c40c0336977d7077c9eaeb
--- /dev/null
+++ b/records/sd1/sweep/72d385765ac8eccfa890/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "72d385765ac8eccfa890",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:50:31Z",
+ "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": 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": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.mid",
+   "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": [
+   "out.mid",
+   "multiply",
+   2.0,
+   "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": "c3fc4b53bee0ff2c350f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72d385765ac8eccfa890/workflow.json b/records/sd1/sweep/72d385765ac8eccfa890/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..eba0e25081f46820b573f7106460007d4a5f263b
--- /dev/null
+++ b/records/sd1/sweep/72d385765ac8eccfa890/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72d385765ac8eccfa890)"
+  },
+  "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 \"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/72de2027c08353263ed7/interpretations.jsonl b/records/sd1/sweep/72de2027c08353263ed7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2c7081c29f0aba0520a1aa70fa8c894713a5c3c4
--- /dev/null
+++ b/records/sd1/sweep/72de2027c08353263ed7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "72de2027c08353263ed7"}, "kind": "caption", "value": "Rainy neon street at dusk with pink and blue signs and pedestrians on wet pavement", "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": "05f51ac8b3830fd3"}
+{"target": {"record": "72de2027c08353263ed7"}, "kind": "change", "value": "Still a neon street, but redrawn with a new sign layout, more pink and magenta neon, a lighter dusk sky, and a more centred perspective.", "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": "43858784d18cd5d0"}
+{"target": {"record": "72de2027c08353263ed7"}, "kind": "keywords", "value": ["neon street", "dusk city", "pink neon", "wet pavement", "pedestrians"], "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": "2cd7f0e1214ae86a"}
+{"target": {"record": "72de2027c08353263ed7"}, "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": "2026-10-05T04:43:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "058d4941a201bffb"}
diff --git a/records/sd1/sweep/72de2027c08353263ed7/measurements.json b/records/sd1/sweep/72de2027c08353263ed7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ffe17a0849c936fd3e366455c7a6a95d3680389c
--- /dev/null
+++ b/records/sd1/sweep/72de2027c08353263ed7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72de2027c08353263ed7",
+ "values": {
+  "std": {
+   "value": 32.936,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02787,
+   "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": 33.752,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.324109,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.432648,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.148483,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.051351,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9894,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4326,
+   "method": "CLIP zero-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/72de2027c08353263ed7/record.json b/records/sd1/sweep/72de2027c08353263ed7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d61acea47abf4662de8b232685e39d75365c0b87
--- /dev/null
+++ b/records/sd1/sweep/72de2027c08353263ed7/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "72de2027c08353263ed7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:15:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.hi",
+   "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": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_scalar",
+   -0.8,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": -0.8,
+    "applied": -0.8,
+    "reason": "outside safe range [-0.5, 1.5] 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": "a746c308b3871987b74a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72de2027c08353263ed7/workflow.json b/records/sd1/sweep/72de2027c08353263ed7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc7cbddbbd73be882c972ecf140b9f6d64952ab3
--- /dev/null
+++ b/records/sd1/sweep/72de2027c08353263ed7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72de2027c08353263ed7)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 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/72deabe7f513eb417241/measurements.json b/records/sd1/sweep/72deabe7f513eb417241/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a6edd8f3bc3c4b0458a8efdcbcb5748989d51120
--- /dev/null
+++ b/records/sd1/sweep/72deabe7f513eb417241/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72deabe7f513eb417241",
+ "values": {
+  "cosine_distance": {
+   "value": 0.294504,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.368244,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.18172,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.062721,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.373,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00484,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.28,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9783,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.057,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72deabe7f513eb417241/record.json b/records/sd1/sweep/72deabe7f513eb417241/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..daf5e713ce7b63f1d1dde36bbc6b2672de07b1f2
--- /dev/null
+++ b/records/sd1/sweep/72deabe7f513eb417241/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "72deabe7f513eb417241",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e66bd807750caf222e1421a09bd49cb86417dcdcdb49731d72973c5b171335d0.png",
+  "batch_id": "kb_lcm_h2h_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",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0166a70419e120102e0c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72deabe7f513eb417241/workflow.json b/records/sd1/sweep/72deabe7f513eb417241/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc5b19075f175954c9ad46602ae83dae07341401
--- /dev/null
+++ b/records/sd1/sweep/72deabe7f513eb417241/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72deabe7f513eb417241)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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": [
+    "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/72e144124be9e7a4d687/measurements.json b/records/sd1/sweep/72e144124be9e7a4d687/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e6aabf8bdefd23a0e39e1cdd6cd1ba181fb96753
--- /dev/null
+++ b/records/sd1/sweep/72e144124be9e7a4d687/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "72e144124be9e7a4d687",
+ "values": {
+  "cosine_distance": {
+   "value": 0.091208,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.136003,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.106997,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040173,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 83.239,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00615,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 28.386,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.118,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1441,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72e144124be9e7a4d687/record.json b/records/sd1/sweep/72e144124be9e7a4d687/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d98a2664c2b06a43002b47d6c2b611740a5998e
--- /dev/null
+++ b/records/sd1/sweep/72e144124be9e7a4d687/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "72e144124be9e7a4d687",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:30Z",
+ "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": "output_blocks.11.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "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/f4431285b161c49d6f4145f24c06d0b4e0fc6443aea05fd1fefbb79558291be4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "f4431285b161c49d6f4145f24c06d0b4e0fc6443aea05fd1fefbb79558291be4",
+  "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": "6622dca89d55d88dd342"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72e144124be9e7a4d687/workflow.json b/records/sd1/sweep/72e144124be9e7a4d687/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..da1ae5e581bac422d3cd6ef3e4652e8582321a93
--- /dev/null
+++ b/records/sd1/sweep/72e144124be9e7a4d687/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 72e144124be9e7a4d687)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.transformer_blocks.0.attn1.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f10af829385f698900/measurements.json b/records/sd1/sweep/72f10af829385f698900/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..90d7025c3d94901caf8dfb5ef719d9302673fd0e
--- /dev/null
+++ b/records/sd1/sweep/72f10af829385f698900/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "72f10af829385f698900",
+ "values": {
+  "cosine_distance": {
+   "value": 0.532786,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.694257,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.795275,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.402738,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.634,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00275,
+   "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": 56.827,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9594,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3413,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/72f10af829385f698900/workflow.json b/records/sd1/sweep/72f10af829385f698900/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..09ac0ccf4f5ac0cdf6ca8479d89e770c6c81c52a
--- /dev/null
+++ b/records/sd1/sweep/72f10af829385f698900/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 72f10af829385f698900)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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": [
+    "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