diff --git a/records/sd1/sweep/70d55575f96bd82bc899/output.webp b/records/sd1/sweep/70d55575f96bd82bc899/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9f73b532b7be4fa36cf4cff7aab16f20de96a017
--- /dev/null
+++ b/records/sd1/sweep/70d55575f96bd82bc899/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e1a485a7a4389b81a3ad0744e5cce5895f36341723e73b433aa03ea81a2f2b0
+size 16468
diff --git a/records/sd1/sweep/70db5d0750f1bc2f197b/output.webp b/records/sd1/sweep/70db5d0750f1bc2f197b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4301a585c6d4d6135c5188f656ef3e571e9ce082
--- /dev/null
+++ b/records/sd1/sweep/70db5d0750f1bc2f197b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:824816b99bc404e5f837a9f99c77f5149086802845e0f4aed1d608f703580e91
+size 36498
diff --git a/records/sd1/sweep/70e59b7431636ae3034f/output.webp b/records/sd1/sweep/70e59b7431636ae3034f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5d28a37415c3e53e361cedd78d98fc5e9644168f
--- /dev/null
+++ b/records/sd1/sweep/70e59b7431636ae3034f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:864a75e1581e237d20764726491eb020c9cdd8308e6666321fbc924629acd196
+size 40252
diff --git a/records/sd1/sweep/70e7ed895e98a7d01d57/output.webp b/records/sd1/sweep/70e7ed895e98a7d01d57/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3c81d5fd8e2741a9dff8b279265344d69aac2d59
--- /dev/null
+++ b/records/sd1/sweep/70e7ed895e98a7d01d57/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88a262eb52227b83ceeb036db61051544aae178d8470999ed05c402c92cbe588
+size 17708
diff --git a/records/sd1/sweep/70ec05ccdbff0816e637/output.webp b/records/sd1/sweep/70ec05ccdbff0816e637/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..79bbc059af8c9c2730f92ee9158e80ce2fb219ab
--- /dev/null
+++ b/records/sd1/sweep/70ec05ccdbff0816e637/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5af40069a8f65b7d554337c85d5806e69e299ce4ed480c47df82725a582f1060
+size 51298
diff --git a/records/sd1/sweep/70f1e83ff915e2e3f5f0/output.webp b/records/sd1/sweep/70f1e83ff915e2e3f5f0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b717a681625953f3f515d7fda06b3d6f07b58691
--- /dev/null
+++ b/records/sd1/sweep/70f1e83ff915e2e3f5f0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a6492328ce023cc27b261841cf995dfc189590617b098438c0d2c05f0d203b5
+size 40542
diff --git a/records/sd1/sweep/71022e935c3bdfced4cb/output.webp b/records/sd1/sweep/71022e935c3bdfced4cb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b59b76423c901ea86955b9e65ce3f7d069afd243
--- /dev/null
+++ b/records/sd1/sweep/71022e935c3bdfced4cb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bef1ac11ab61f04220fc72dc717ea12cebea7e0f45d7402a022df825da88bbfa
+size 16620
diff --git a/records/sd1/sweep/71029697400db1e8284c/output.webp b/records/sd1/sweep/71029697400db1e8284c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6dfad63263879e6129f5be66df0f4d38d6111822
--- /dev/null
+++ b/records/sd1/sweep/71029697400db1e8284c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56d32b51cfaba799d630ceb6800001ae1b7c83b473b2d73b1269b395e438dfaa
+size 19932
diff --git a/records/sd1/sweep/71060966b1ffe379ceab/output.webp b/records/sd1/sweep/71060966b1ffe379ceab/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0c3ad7a8bd274f549fd85e2a6c912c4c387bc8f9
--- /dev/null
+++ b/records/sd1/sweep/71060966b1ffe379ceab/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1b32103aada8a9c29929ff7a3e5c1930d213b361f47a7108fb5a111267b87b9
+size 34876
diff --git a/records/sd1/sweep/71191a8f69507aa58d6f/output.webp b/records/sd1/sweep/71191a8f69507aa58d6f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..223af99bb7ab7463bbd6509cf9210cca8b1cbdf3
--- /dev/null
+++ b/records/sd1/sweep/71191a8f69507aa58d6f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eecd2e7be82b0077f64c9297d5706009deaec1464bb6e0bd768f1e5c8c17585c
+size 21678
diff --git a/records/sd1/sweep/711eb9da540a01c81534/output.webp b/records/sd1/sweep/711eb9da540a01c81534/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3fdb66579d95eea5041ae93512da457bf197be31
--- /dev/null
+++ b/records/sd1/sweep/711eb9da540a01c81534/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c461ec0441fa40d1ff05ae7e33e3800b188beeb0039298f903d5499db0555d6
+size 31672
diff --git a/records/sd1/sweep/7129ff9ae7aa4a636098/output.webp b/records/sd1/sweep/7129ff9ae7aa4a636098/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..35131288bd03b99d782fcfb3be509faae1e232ae
--- /dev/null
+++ b/records/sd1/sweep/7129ff9ae7aa4a636098/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a77b3b0bfe0f0b8a384ef2dcde55b2a95dc6d2f5bdc8254c76399bb6f6148c48
+size 24884
diff --git a/records/sd1/sweep/7133e19b0acc9d212ca7/output.webp b/records/sd1/sweep/7133e19b0acc9d212ca7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..082f0863dcf242376119c644b7e1ad9f12fb781e
--- /dev/null
+++ b/records/sd1/sweep/7133e19b0acc9d212ca7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e7f522d4f11932b6ed7fac7acbbdf6fea0429f0a199b8b61bae6af1e7a00556
+size 51066
diff --git a/records/sd1/sweep/713f3390b5cc3f45ba61/output.webp b/records/sd1/sweep/713f3390b5cc3f45ba61/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ad210c01934200929f6afec29e6ced6fac760d02
--- /dev/null
+++ b/records/sd1/sweep/713f3390b5cc3f45ba61/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2d3c07d7eff8b13b2657a8e5c5030000724356b55f4b647b61d2d9b95add13b
+size 31024
diff --git a/records/sd1/sweep/714169e100e849a237c8/output.webp b/records/sd1/sweep/714169e100e849a237c8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4b2ac6ab0b946a7254555228132d3aa8ad9a3284
--- /dev/null
+++ b/records/sd1/sweep/714169e100e849a237c8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:08e444fe6509e5265033d9ecfceea3f7be7480a6820bd50d506163fe91351dde
+size 16836
diff --git a/records/sd1/sweep/71430be7a1b7bedbc362/output.webp b/records/sd1/sweep/71430be7a1b7bedbc362/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..683561f455810aeee4f3d350c6231e03c5da49e0
--- /dev/null
+++ b/records/sd1/sweep/71430be7a1b7bedbc362/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f90ab37ec7c59ffe6a9a1ff675135c47f5d0ddb9852a854a5220713f5b815dc
+size 29214
diff --git a/records/sd1/sweep/71470dd252966636150e/output.webp b/records/sd1/sweep/71470dd252966636150e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..320df7f7a4c3faf39ff2f6905d1878d22a7e48ac
--- /dev/null
+++ b/records/sd1/sweep/71470dd252966636150e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f0e8bb7d7c4bff40ef719c49d8bbee5c71aa37f5b58b5d29c04f7a4b43767fa
+size 42274
diff --git a/records/sd1/sweep/7151d9433ef00c2b3cda/output.webp b/records/sd1/sweep/7151d9433ef00c2b3cda/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..88d75bbb8727fec7b69ee9e4e72f3790292566a9
--- /dev/null
+++ b/records/sd1/sweep/7151d9433ef00c2b3cda/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fad540229b5bf41451fe26dd3669bb265607c22a10e8c62d12ea05fe50be20e
+size 19932
diff --git a/records/sd1/sweep/71520b7f2b74367db816/output.webp b/records/sd1/sweep/71520b7f2b74367db816/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8e609fc6ba91c0be05a2dd6ddf2bdf56097d4fdb
--- /dev/null
+++ b/records/sd1/sweep/71520b7f2b74367db816/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9af03c0f6755f97ff502c5d69e11dfdffa0f5519d568247c31ff0e3fe3f7640b
+size 48842
diff --git a/records/sd1/sweep/7179aaa6d28804253b4e/output.webp b/records/sd1/sweep/7179aaa6d28804253b4e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..23035f29fb5b891462a4f53c120b34f4b483c01c
--- /dev/null
+++ b/records/sd1/sweep/7179aaa6d28804253b4e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ce6741be01c38aeb3760b245467822cff8926efe52eca52d3697fe51a75f94e
+size 46066
diff --git a/records/sd1/sweep/71933eb438f8500d105e/output.webp b/records/sd1/sweep/71933eb438f8500d105e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..332d12da11fa39f5043d3b5714364c01efc074f3
--- /dev/null
+++ b/records/sd1/sweep/71933eb438f8500d105e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:652146c01f6da639cd69ac1fb83f7fd6aa3128bc000646015bae96cfa2085765
+size 19936
diff --git a/records/sd1/sweep/7198dd9e359b42e8c5a8/output.webp b/records/sd1/sweep/7198dd9e359b42e8c5a8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1601a3ab0cfab6edb3b02073ae7ef4c1f24a6094
--- /dev/null
+++ b/records/sd1/sweep/7198dd9e359b42e8c5a8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5223e9ea42accc3843e1855c437f7c04c2cee4b97e8c9b4fe3874edf8093f454
+size 39958
diff --git a/records/sd1/sweep/719d47eee18d785c9537/output.webp b/records/sd1/sweep/719d47eee18d785c9537/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f3b97146ccbbdb6919f6c0d1bf1a575f76b8b9c0
--- /dev/null
+++ b/records/sd1/sweep/719d47eee18d785c9537/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:833ba4f2117e84605759ad9d161d3f32b92df66185310f2d5d38a882a6071954
+size 19930
diff --git a/records/sd1/sweep/71a0b17d70988c0b66e3/output.webp b/records/sd1/sweep/71a0b17d70988c0b66e3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..84e4d4c3f4f4f256f91f48cf349fd2fd15e08fc0
--- /dev/null
+++ b/records/sd1/sweep/71a0b17d70988c0b66e3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b10f031567e864f7f9f43943c852d7ada746d88f5c3dbeac2114c840f5ca8bc4
+size 29586
diff --git a/records/sd1/sweep/71a752dbeca05056caf0/output.webp b/records/sd1/sweep/71a752dbeca05056caf0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5b19cc607505bba08d3fd54c376c0e0be6d6b97c
--- /dev/null
+++ b/records/sd1/sweep/71a752dbeca05056caf0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d0766402bc58ccc132bb9aaab8239723dbaa07342946c349df61f237bb6a774
+size 27464
diff --git a/records/sd1/sweep/71a7a79b7a71898ec6fd/output.webp b/records/sd1/sweep/71a7a79b7a71898ec6fd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6d8a2fc265d56a17712e45c34b5eaaffac503628
--- /dev/null
+++ b/records/sd1/sweep/71a7a79b7a71898ec6fd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b19ec0b39efad75149db9b1684b7ba39e4883027d9b9f1ad8315de3aa310b5cb
+size 27982
diff --git a/records/sd1/sweep/71aaf97104bdc9e388a7/output.webp b/records/sd1/sweep/71aaf97104bdc9e388a7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5a67c308960af29815b98d8425e5a8800769f08e
--- /dev/null
+++ b/records/sd1/sweep/71aaf97104bdc9e388a7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59bc72713d792082fd88e5877d4ad11fa089f9cff65318853bf0e38483de3720
+size 14948
diff --git a/records/sd1/sweep/71b5042b73c4f47b3df3/output.webp b/records/sd1/sweep/71b5042b73c4f47b3df3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..650d751f9ef0d06f918a57f01c431451b5a27b6e
--- /dev/null
+++ b/records/sd1/sweep/71b5042b73c4f47b3df3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71e85771c5ac48f4fef3544851c747ca66f5826b149c1dd6f3ef342d1f4f1f9a
+size 40164
diff --git a/records/sd1/sweep/71b9322cff88615a71d1/output.webp b/records/sd1/sweep/71b9322cff88615a71d1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..17a22106c07dee0287303912eec9527160d565b1
--- /dev/null
+++ b/records/sd1/sweep/71b9322cff88615a71d1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8513ce4091a60b9ea9fc4f8bb56bc6b76ff6243b41e4bdae6b60255ab6f1adec
+size 30222
diff --git a/records/sd1/sweep/71ba725ec0e12ffddd0d/output.webp b/records/sd1/sweep/71ba725ec0e12ffddd0d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b8a83bef4c9c9983329c447ea77b2914bb86cf1d
--- /dev/null
+++ b/records/sd1/sweep/71ba725ec0e12ffddd0d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c140dac63aaf1723d67676b576df3db38767bfe007ed312dc4027510e2dade2
+size 33180
diff --git a/records/sd1/sweep/71bb5012164a18cdc389/output.webp b/records/sd1/sweep/71bb5012164a18cdc389/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4cf463fca2853cb778d12e2c85dce839b12664aa
--- /dev/null
+++ b/records/sd1/sweep/71bb5012164a18cdc389/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c52dbf89f72aed358c9d29dc27bfc7615adf70849b46d9dac8fb27f2370abbb
+size 18626
diff --git a/records/sd1/sweep/71bc1d3a1e929a71d165/output.webp b/records/sd1/sweep/71bc1d3a1e929a71d165/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2690b690b35ee4def99f7f129d5299ebe8e285c6
--- /dev/null
+++ b/records/sd1/sweep/71bc1d3a1e929a71d165/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb992e53e9c98de92d88d58576c5daec8d910c87edd1208f2472e9e5e1386963
+size 26678
diff --git a/records/sd1/sweep/71bfa7fc20d9d615bfb2/output.webp b/records/sd1/sweep/71bfa7fc20d9d615bfb2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7c7c5afdeeb23098c0bf84a9677d0757488e3b94
--- /dev/null
+++ b/records/sd1/sweep/71bfa7fc20d9d615bfb2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:59ede73c6c28f42700625c39d153b3cbdd94aa4032d63667aee445a2a2161ef1
+size 34384
diff --git a/records/sd1/sweep/71c034971a6ea01116d1/output.webp b/records/sd1/sweep/71c034971a6ea01116d1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cddc9fe171bf1a6352b1c9304069e18fdba86595
--- /dev/null
+++ b/records/sd1/sweep/71c034971a6ea01116d1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8e395d672b053c7b7d31c51ab45852ebf666dbbee73625ec8aca8436a0db4e6
+size 23764
diff --git a/records/sd1/sweep/71c6a089510ada38474e/output.webp b/records/sd1/sweep/71c6a089510ada38474e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f20e58b681a4a4bc5301bb0e2d0b396961c3fed3
--- /dev/null
+++ b/records/sd1/sweep/71c6a089510ada38474e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd33ff9abe1091514928ee2b6d83e2d87fdcd528fe4b3804e8c7f440001081d2
+size 15282
diff --git a/records/sd1/sweep/71c72442d1009e2dac97/output.webp b/records/sd1/sweep/71c72442d1009e2dac97/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9c4264f995f3c46579c193381672fa3d3ee132d1
--- /dev/null
+++ b/records/sd1/sweep/71c72442d1009e2dac97/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66e57659cf208b7a8de01b1dd10ee172b64059129e3baca118513197739e407a
+size 25030
diff --git a/records/sd1/sweep/71cfc42591aab45cafcc/output.webp b/records/sd1/sweep/71cfc42591aab45cafcc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..baa80504d916d6d3402f3ec73a54fd41cc1e93e6
--- /dev/null
+++ b/records/sd1/sweep/71cfc42591aab45cafcc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8cde5071d618f643b3b97173fce4d2a3c67e2ef8c8e6b9447bcedd7bf9eabee3
+size 33926
diff --git a/records/sd1/sweep/71dc6ebf8593a2d4b663/output.webp b/records/sd1/sweep/71dc6ebf8593a2d4b663/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ed0c099e7ddcfb4ecc3696884b3ac5f9c1a67c5e
--- /dev/null
+++ b/records/sd1/sweep/71dc6ebf8593a2d4b663/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c28f832ed31d21911a4f2ae7cb038b2da6a5e0c0ad2cc5663aba0e6e1a9b69f
+size 21808
diff --git a/records/sd1/sweep/71de26a2cde7c146934c/output.webp b/records/sd1/sweep/71de26a2cde7c146934c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c80019bd275e5d1e54438d9381e415a6e1f24e24
--- /dev/null
+++ b/records/sd1/sweep/71de26a2cde7c146934c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:270b4999010e08f07af2033c78e2425b76e49a7592e89e925a4e19674f2e231f
+size 19696
diff --git a/records/sd1/sweep/71e43f983c41558872aa/output.webp b/records/sd1/sweep/71e43f983c41558872aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..120bb992307c0a64756b02d47161514bd4287e81
--- /dev/null
+++ b/records/sd1/sweep/71e43f983c41558872aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:efca80433f853a2fc314ad94f08722b1cb7307f41880399a4ae71d26c5b685d4
+size 15518
diff --git a/records/sd1/sweep/71e9b69f821ef3764a3c/output.webp b/records/sd1/sweep/71e9b69f821ef3764a3c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b30fdcb9d69901e86997ae13d423f4e2d5b0c5b5
--- /dev/null
+++ b/records/sd1/sweep/71e9b69f821ef3764a3c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8d64d80c1902280bfc6fb7a0a88ed943fa0a59b00c668cc97c431297da348f0
+size 40228
diff --git a/records/sd1/sweep/71f34cda7a2725fbcce2/output.webp b/records/sd1/sweep/71f34cda7a2725fbcce2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a703f7a99219ebec740fd775d1bec46c6335c6ce
--- /dev/null
+++ b/records/sd1/sweep/71f34cda7a2725fbcce2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d45d33be2914af3045d4cd2e27ac4c3302b80062f16a92515c0afdfd6e70a95e
+size 52372
diff --git a/records/sd1/sweep/71f683eb119fdf0c2b9e/output.webp b/records/sd1/sweep/71f683eb119fdf0c2b9e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..64dcb44e1c6855df49561d32034f16baefa63fdf
--- /dev/null
+++ b/records/sd1/sweep/71f683eb119fdf0c2b9e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e50769a2653750f98f390b4634202a4bcbb86d136e4ed977e475283ac3435004
+size 22288
diff --git a/records/sd1/sweep/71f70031e8998fa61a65/output.webp b/records/sd1/sweep/71f70031e8998fa61a65/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..863f82ac435f01c8dc3129aa473b41e74b87b8d1
--- /dev/null
+++ b/records/sd1/sweep/71f70031e8998fa61a65/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d8ea6184fcea6b9c4b176c0528f044dc697a5d04ef5372edf624a8d4398ec284
+size 19510
diff --git a/records/sd1/sweep/7200c124fec9e0d5c97d/output.webp b/records/sd1/sweep/7200c124fec9e0d5c97d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9b0d5a010e2c839f33b3c24f8bb04b04591437f0
--- /dev/null
+++ b/records/sd1/sweep/7200c124fec9e0d5c97d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c76cc1a9df8ce60d570ee34ac0b1305a47fa4fb7c701ce919cc25fd09b17d1f2
+size 29270
diff --git a/records/sd1/sweep/720350c0a5677b8d7543/output.webp b/records/sd1/sweep/720350c0a5677b8d7543/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2325d6e6c4fb29bc6955c9e7a01094fc096082c7
--- /dev/null
+++ b/records/sd1/sweep/720350c0a5677b8d7543/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d9225a23286413b9aae141b0927284204ed43818dd34a9081a034200c22652b
+size 20186
diff --git a/records/sd1/sweep/720359581e80966a0cfa/output.webp b/records/sd1/sweep/720359581e80966a0cfa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..608f6608d48ce424f26ff880d11b8ced57239e1c
--- /dev/null
+++ b/records/sd1/sweep/720359581e80966a0cfa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74f342a60e4e7bf00517d80080d5b9d68e7e8ea0011a83e72264f1e74b51e8e0
+size 44880
diff --git a/records/sd1/sweep/72052c7a937402cb6cce/output.webp b/records/sd1/sweep/72052c7a937402cb6cce/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3b893d7b69efff4afd02b403e9df42f698897d28
--- /dev/null
+++ b/records/sd1/sweep/72052c7a937402cb6cce/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:88f8d7b4f5cd9885d22ab4c4e944d37f3dbd4ff4274f70f2cda3a105754ed1b6
+size 14746
diff --git a/records/sd1/sweep/7205941fbacacc13fcc9/output.webp b/records/sd1/sweep/7205941fbacacc13fcc9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2879f52193e70d0974b1797e70fe08f152fc0a72
--- /dev/null
+++ b/records/sd1/sweep/7205941fbacacc13fcc9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06fe2ac1bfdc0927b0d699066c7c061b1f4b4756b91d0419c9c0ebd58a8f0b71
+size 38480
diff --git a/records/sd1/sweep/7205a5b4096fc8aa770f/output.webp b/records/sd1/sweep/7205a5b4096fc8aa770f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ded1130d2c6464613b0411c8d36b02bcbf6cc1e3
--- /dev/null
+++ b/records/sd1/sweep/7205a5b4096fc8aa770f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:71b914f29afe35f9ec3f340229cbe41ce521ea2ef134a40f5aa3c9571081b25a
+size 22796
diff --git a/records/sd1/sweep/7212c156ed55f97134e5/output.webp b/records/sd1/sweep/7212c156ed55f97134e5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c796ead54f07b579f56f5486afb7b7cf2bf99cff
--- /dev/null
+++ b/records/sd1/sweep/7212c156ed55f97134e5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b84e12ff4fc568cdf3e8fad329da2e2625b52f933e6747440c1946e58c1edfa6
+size 57824
diff --git a/records/sd1/sweep/721fad124e0a979805c6/output.webp b/records/sd1/sweep/721fad124e0a979805c6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3889d3f3f0c9bd18ff8451fdfcd95ce1f0bae5e8
--- /dev/null
+++ b/records/sd1/sweep/721fad124e0a979805c6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14aefc5a5aaaa3f9a4c09f19da764318ef258ce131dcba2ee8a6f92dc35a21ee
+size 48412
diff --git a/records/sd1/sweep/7222d7b1744dded53944/output.webp b/records/sd1/sweep/7222d7b1744dded53944/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4b47409c4863e7f2e044d32481a64a50409d0dca
--- /dev/null
+++ b/records/sd1/sweep/7222d7b1744dded53944/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f1da9e9e75860b7c0277d81f524b4f163b0963acaec87935a41b91f58ac7c50
+size 36618
diff --git a/records/sd1/sweep/7224e45fd82116f9ac78/output.webp b/records/sd1/sweep/7224e45fd82116f9ac78/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..898d41c3330c605169a0d32e4f76c5b5ca9f3986
--- /dev/null
+++ b/records/sd1/sweep/7224e45fd82116f9ac78/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:121a04385b93cef1199d8b1936d806aa179d349627a45b28e55dbf5980a8010b
+size 30650
diff --git a/records/sd1/sweep/7227c36d039101dd9afe/output.webp b/records/sd1/sweep/7227c36d039101dd9afe/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a985730f943ace47e032f80d33cce23924d3ee92
--- /dev/null
+++ b/records/sd1/sweep/7227c36d039101dd9afe/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e47559f48e893c72e125364335ceff6ddc3206a94d80fa7b4a3fe9e0c2c2ff13
+size 35412
diff --git a/records/sd1/sweep/722b91ac38a78296ccce/output.webp b/records/sd1/sweep/722b91ac38a78296ccce/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2fa9685575552c94d81c2df98985f8786756380b
--- /dev/null
+++ b/records/sd1/sweep/722b91ac38a78296ccce/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ca64acd1437f64c48e7dfc453d3ba72a65e49ca2e59dab06ad634df43124017
+size 38096
diff --git a/records/sd1/sweep/7234264bbaf2638f81b4/output.webp b/records/sd1/sweep/7234264bbaf2638f81b4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d93f3a7ee7c05a968de88de24ab4169ad5985e1a
--- /dev/null
+++ b/records/sd1/sweep/7234264bbaf2638f81b4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:40753c06ca46bb4f8a800ae7f60961c3f2ffec6511498d8136579d60dcc5141f
+size 19046
diff --git a/records/sd1/sweep/724be535c1a663c55995/output.webp b/records/sd1/sweep/724be535c1a663c55995/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7624453af76189208bad090891414b14fa9cdfd0
--- /dev/null
+++ b/records/sd1/sweep/724be535c1a663c55995/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05cad061384d79b5dbdefe7df43e7cc8b632a63d1a10f4769ee33c9e7f5c0070
+size 38810
diff --git a/records/sd1/sweep/724bed0c272075060cf2/output.webp b/records/sd1/sweep/724bed0c272075060cf2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4e7cc48059cd7e147ebb32ec75cecdf080ee33a3
--- /dev/null
+++ b/records/sd1/sweep/724bed0c272075060cf2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2cc73f434c9826aa26443ce80cf2b33683ce33f95ac0eeca24910e12c3992419
+size 16404
diff --git a/records/sd1/sweep/725c5417782018834dcb/output.webp b/records/sd1/sweep/725c5417782018834dcb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e5a2c64fb774b9a08f017b8f25d243829c9112c9
--- /dev/null
+++ b/records/sd1/sweep/725c5417782018834dcb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94b26a3cd6164f79e27b7734eb8d8d2e62f10c81767a4c820e4e7117820a5453
+size 30938
diff --git a/records/sd1/sweep/7262e0ce4373c27df4ee/output.webp b/records/sd1/sweep/7262e0ce4373c27df4ee/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..edb6bdff41813e370bf2b0210c912222f2e2c239
--- /dev/null
+++ b/records/sd1/sweep/7262e0ce4373c27df4ee/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb52bdd0e2852ade548c3e53a4c922e4ba9c3e30b327a6e6e4102e428cd4eae6
+size 39764
diff --git a/records/sd1/sweep/7265998521c57e0c18b8/output.webp b/records/sd1/sweep/7265998521c57e0c18b8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..49bfda27e27825b549960c77b8d919486848159f
--- /dev/null
+++ b/records/sd1/sweep/7265998521c57e0c18b8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:806ce9c1e6825c810948a8fea958a4fc7d87675661c93e7e0db93027e3c9d2ac
+size 33724
diff --git a/records/sd1/sweep/72698a0f4b6f9a14f5e3/output.webp b/records/sd1/sweep/72698a0f4b6f9a14f5e3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..da0aeead893ccf3ba197cf8a26c2bc11dd40eeb6
--- /dev/null
+++ b/records/sd1/sweep/72698a0f4b6f9a14f5e3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ed26024c3dd7627f967f2f7390abea88aedfc5af0ba40f2951ee79df9e5830e
+size 25624
diff --git a/records/sd1/sweep/72810a4465bc5e7e0c3b/output.webp b/records/sd1/sweep/72810a4465bc5e7e0c3b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..221fc0dc4a8542918e3931c8d402f6b5d0190fd6
--- /dev/null
+++ b/records/sd1/sweep/72810a4465bc5e7e0c3b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f67b1f8523809d31d418a3615a54ee3819a301f724958a8ae419f059f8f649e0
+size 27628
diff --git a/records/sd1/sweep/728aa7860c96aa024786/output.webp b/records/sd1/sweep/728aa7860c96aa024786/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c419108fa6e640000a1b2291047ac567fa1bc45e
--- /dev/null
+++ b/records/sd1/sweep/728aa7860c96aa024786/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ef199fd2d1f13e61a626ac4811a30e953c5fa342d266b7183a941ebe01fb6e4
+size 21610
diff --git a/records/sd1/sweep/728beea8ead44b3f411a/output.webp b/records/sd1/sweep/728beea8ead44b3f411a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1926a831987197652d673a982040dde54ec10e13
--- /dev/null
+++ b/records/sd1/sweep/728beea8ead44b3f411a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45f73c4c968d5a55fad99df0b124b3e5976f0a54db5fcd5b53081ade4a2276ae
+size 39114
diff --git a/records/sd1/sweep/728eb170171a61d87cc3/output.webp b/records/sd1/sweep/728eb170171a61d87cc3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6982d178d15ff0cd6d239d051daad3849c7efde5
--- /dev/null
+++ b/records/sd1/sweep/728eb170171a61d87cc3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:de244976fbfc455ecec57584b6ac45c354400a93cb53b17f175b0d6f57ef89e9
+size 33388
diff --git a/records/sd1/sweep/728faa0d7b35958a4a45/output.webp b/records/sd1/sweep/728faa0d7b35958a4a45/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b1bb7e7a031f46a28e885825671affdd9640c8bb
--- /dev/null
+++ b/records/sd1/sweep/728faa0d7b35958a4a45/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f568a1bc1ea8beb4bdfd5dd05f75411c69806c353e1ddb5ff0694d828256005c
+size 18408
diff --git a/records/sd1/sweep/7297142e874b96245a0e/output.webp b/records/sd1/sweep/7297142e874b96245a0e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..df52fddaa4c3e55d5860291fef3ffdd74b001f03
--- /dev/null
+++ b/records/sd1/sweep/7297142e874b96245a0e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:399d1d88f784d1741be33091d0114ef19e8326ff0a1bb682d2ba529e32ddd23c
+size 31740
diff --git a/records/sd1/sweep/729ca7d47084c875cc74/output.webp b/records/sd1/sweep/729ca7d47084c875cc74/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..67c515ebadbfb2ea8e1ac4c6d4d2a33cffc80ed6
--- /dev/null
+++ b/records/sd1/sweep/729ca7d47084c875cc74/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e78faedbb808ee6de7ec0d31ce8efb5096bcaf58dd1e8a30fceda4e4ba3655e
+size 51428
diff --git a/records/sd1/sweep/72a3f829435102271dc3/output.webp b/records/sd1/sweep/72a3f829435102271dc3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4e9d554ed545aaa6db8996acda3c5fe34ba65deb
--- /dev/null
+++ b/records/sd1/sweep/72a3f829435102271dc3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d39b24f01da320a3a2c13e7e8cff5a8ef3ffd9768b702a7f48cb80dc0c31cfce
+size 22134
diff --git a/records/sd1/sweep/72a6e954b32daef7bd3b/output.webp b/records/sd1/sweep/72a6e954b32daef7bd3b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..66510ce923c43e11069dfd2b40b64ee489fd4360
--- /dev/null
+++ b/records/sd1/sweep/72a6e954b32daef7bd3b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:140632ec96f10a26b297bb8a423537da83ecb3fd8aaa6dfd9c118a264ac2c396
+size 19932
diff --git a/records/sd1/sweep/72a7f9f209f54568febc/output.webp b/records/sd1/sweep/72a7f9f209f54568febc/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..01893fa9deab81739e077efbd8e8e320aa416748
--- /dev/null
+++ b/records/sd1/sweep/72a7f9f209f54568febc/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c2472c834b5cf2c554cd64236e670bc6d5dc857976a7a68cac606bf9859ee3a
+size 36058
diff --git a/records/sd1/sweep/72b406513e57170e69aa/output.webp b/records/sd1/sweep/72b406513e57170e69aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..370f151c2bf5a8cbdd5c7cd3fbc6f02cf60a2456
--- /dev/null
+++ b/records/sd1/sweep/72b406513e57170e69aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b7c900463aa81344268cf6d0363f24da6d59ed2451697f1b13ef98693029e1f
+size 16264
diff --git a/records/sd1/sweep/72ba0e5fecc77ad396e9/output.webp b/records/sd1/sweep/72ba0e5fecc77ad396e9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0c2c27ad0f3d3eba1b8436e1c9158a475f1bfc3a
--- /dev/null
+++ b/records/sd1/sweep/72ba0e5fecc77ad396e9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd0e164bc445606a04ffd3bced8b2381ac07eec39e8dabd1d78752a444e94d17
+size 62040
diff --git a/records/sd1/sweep/72bded680bb6c36aace2/output.webp b/records/sd1/sweep/72bded680bb6c36aace2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e71f33f473dce0b4f78a7b18635cb9a35b341fe6
--- /dev/null
+++ b/records/sd1/sweep/72bded680bb6c36aace2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d55d39164d1794df1641e82dab701d4432ed5df32bb811facf3542175dba4a86
+size 12912
diff --git a/records/sd1/sweep/72bfa37bd610aa6d1c17/output.webp b/records/sd1/sweep/72bfa37bd610aa6d1c17/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..248b33c18ee639e3ae6d4b9459013c92a67f0c78
--- /dev/null
+++ b/records/sd1/sweep/72bfa37bd610aa6d1c17/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:02d5b2ed4b2871de3f783b1acb71cd16c582caff786771cace4e01846ddceb9a
+size 40234
diff --git a/records/sd1/sweep/72c1ca8384a0712bffa6/output.webp b/records/sd1/sweep/72c1ca8384a0712bffa6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..41adb665211357e6d3fe64253c3dbf44c8b9374d
--- /dev/null
+++ b/records/sd1/sweep/72c1ca8384a0712bffa6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee69a01018f4c2aa2e7a62bfee0ad60b62f2e323f2ef70327259c59afba8dce8
+size 51744
diff --git a/records/sd1/sweep/72c4d9471ef372664a58/output.webp b/records/sd1/sweep/72c4d9471ef372664a58/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..47f30eea7f53064f7e94f70afd13c74f5631ecfd
--- /dev/null
+++ b/records/sd1/sweep/72c4d9471ef372664a58/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:853414032f12dd44a1334a1a1bfb234cbcc222c58c269ed02c344cf790bac1ac
+size 32888
diff --git a/records/sd1/sweep/72c727454b0eda006785/output.webp b/records/sd1/sweep/72c727454b0eda006785/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e4387f102f9fde73417b550c8f8fd3adc5c1102a
--- /dev/null
+++ b/records/sd1/sweep/72c727454b0eda006785/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c605613158c9646ab7e0ea544b7ce3582e05e594332d503fae8a9d3740e063b4
+size 3736
diff --git a/records/sd1/sweep/72c8eca2005882a6f656/output.webp b/records/sd1/sweep/72c8eca2005882a6f656/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..42ca3dbc29a4e5b6b64c18a09e1066dbd7017674
--- /dev/null
+++ b/records/sd1/sweep/72c8eca2005882a6f656/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:172934b9172240acb6aba9c868bd8f004c3271e5926537174a5edb8fae27245b
+size 58100
diff --git a/records/sd1/sweep/72c9daa51b779cd41394/output.webp b/records/sd1/sweep/72c9daa51b779cd41394/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cb477efdf5a977ccf3823f274e4a4e071f4eaffb
--- /dev/null
+++ b/records/sd1/sweep/72c9daa51b779cd41394/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10add345fc2b1de3a2656ed2210765b939a4b4375efd211e9d101ddc5ff84ec6
+size 71804
diff --git a/records/sd1/sweep/72d385765ac8eccfa890/output.webp b/records/sd1/sweep/72d385765ac8eccfa890/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..293c707d2031fbd04d3dfafeccf2001b4c2f0cce
--- /dev/null
+++ b/records/sd1/sweep/72d385765ac8eccfa890/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2f14208fd320fc8844e39c6d7bb37ffcc2f9484f0686d98b405b9d4fd5f547c
+size 51788
diff --git a/records/sd1/sweep/72de2027c08353263ed7/output.webp b/records/sd1/sweep/72de2027c08353263ed7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8f9614cbc82a1f5c0a772d8d18eb232fea046ae9
--- /dev/null
+++ b/records/sd1/sweep/72de2027c08353263ed7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd466e9da2a44cc268ee589bee762e9f654386421063f8e56162a41461a80c8b
+size 50864
diff --git a/records/sd1/sweep/72deabe7f513eb417241/output.webp b/records/sd1/sweep/72deabe7f513eb417241/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..dbf26848a9426a6a8dc88b561ccee377ad59e241
--- /dev/null
+++ b/records/sd1/sweep/72deabe7f513eb417241/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfef258f46d44442abebaf04272e9800e1748839cb7400609a80f826168c57ba
+size 43294
diff --git a/records/sd1/sweep/72e144124be9e7a4d687/output.webp b/records/sd1/sweep/72e144124be9e7a4d687/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a8281fb139512935eee75c89bccfcc50a590782c
--- /dev/null
+++ b/records/sd1/sweep/72e144124be9e7a4d687/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6b8c7a076e5ed350ef33ca06cf6f2393b51dffecee286ddfab3d6867188ea15d
+size 33516
diff --git a/records/sd1/sweep/72f10af829385f698900/output.webp b/records/sd1/sweep/72f10af829385f698900/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5d8228ad819b3e35a473764703eccd72bc3123af
--- /dev/null
+++ b/records/sd1/sweep/72f10af829385f698900/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61a018f3ed8fef67c669b4b18baedca651f8db96ae2993ab91058c09704fba35
+size 33998
diff --git a/records/sd1/sweep/72f3d8ac2f466462da67/output.webp b/records/sd1/sweep/72f3d8ac2f466462da67/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4186170a34e6429b1a173f7c6b9e86c537e56bb6
--- /dev/null
+++ b/records/sd1/sweep/72f3d8ac2f466462da67/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccea3d20238b275ef68138bf00c30b23f96455faf3d55ef61b0c1c60878b98e1
+size 17590
diff --git a/records/sd1/sweep/72f887128d77cc17d072/output.webp b/records/sd1/sweep/72f887128d77cc17d072/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..78a3e8198578b4543830ab4137cd3a93cd72f436
--- /dev/null
+++ b/records/sd1/sweep/72f887128d77cc17d072/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:863c9080bffb41cae8dadaf81878d39175fe5071d1ea397a534a4988f46fef1d
+size 40864
diff --git a/records/sd1/sweep/72fbce8849f449a75909/output.webp b/records/sd1/sweep/72fbce8849f449a75909/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..acda803147788fe4a1195eb5178a13d2388cc538
--- /dev/null
+++ b/records/sd1/sweep/72fbce8849f449a75909/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3d7354a71fbb0650f2503e7004d4c9aeb6aac201ff748f7e830a8a60884637c
+size 33862
diff --git a/records/sd1/sweep/730c5c69715b0987256f/output.webp b/records/sd1/sweep/730c5c69715b0987256f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8cad2ec8fc848a9fa14b794e822defe2b60e7f0f
--- /dev/null
+++ b/records/sd1/sweep/730c5c69715b0987256f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd4fb92668aa0f82f4fd25aeff90319da8ebbf5e3eea382161ba5892323d0a9e
+size 30936
diff --git a/records/sd1/sweep/730f92eec2f6f3125e0a/output.webp b/records/sd1/sweep/730f92eec2f6f3125e0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e88f6650185f27704495f9ad8afc9c817fa5679a
--- /dev/null
+++ b/records/sd1/sweep/730f92eec2f6f3125e0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa07b0206e5547bdef2d7e1e77764a833a59b6b30c6e30ad4891e51fe8e2711e
+size 30420
diff --git a/records/sd1/sweep/7316eb0b2b21066ebd43/output.webp b/records/sd1/sweep/7316eb0b2b21066ebd43/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..25b89b0b9108ab5e76abe108754e8d512454a807
--- /dev/null
+++ b/records/sd1/sweep/7316eb0b2b21066ebd43/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:61079b5c898c83c4110ca0cbb05cf7c7105166969ff709b0aef264b057bc8817
+size 68848
diff --git a/records/sd1/sweep/73219689c6028d6fc083/output.webp b/records/sd1/sweep/73219689c6028d6fc083/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0f507bee56a1a88a465d5346f3e8339fcb3f8a2d
--- /dev/null
+++ b/records/sd1/sweep/73219689c6028d6fc083/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:98260134447986791dd9e540e4dc25ff2cff301f4d3d31d35c2945e51410fd96
+size 134002
diff --git a/records/sd1/sweep/73226bc2d0661341572c/output.webp b/records/sd1/sweep/73226bc2d0661341572c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e61cd68b9c261376847722b204d47ae9ddada62e
--- /dev/null
+++ b/records/sd1/sweep/73226bc2d0661341572c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fbdd5ef3a177b04b6c7eb11f7fed570bd407852787b949ce099a4df6df890c7e
+size 33764
diff --git a/records/sd1/sweep/732b393b4b7ccb9f3acf/output.webp b/records/sd1/sweep/732b393b4b7ccb9f3acf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..753b6a80f6366fc9d4558f2c442f4d4a14bc4325
--- /dev/null
+++ b/records/sd1/sweep/732b393b4b7ccb9f3acf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c261ebff55f06239dd90316f652a66a823ac0c0b984064b87c06191349c51a1a
+size 46640
diff --git a/records/sd1/sweep/732d8899e03ebf59f70f/output.webp b/records/sd1/sweep/732d8899e03ebf59f70f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..75c71ffe31b46da1a3cd401a9c12dd8b1eedf0f6
--- /dev/null
+++ b/records/sd1/sweep/732d8899e03ebf59f70f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:575267e5ad8d686f5d052412a1d78554a587b6a1a3ea540f1a8e3334e1cb3c7c
+size 31746
diff --git a/records/sd1/sweep/732f94fc011903bcaaf6/output.webp b/records/sd1/sweep/732f94fc011903bcaaf6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a14ac8218a64ac0ef599a111a1a824260aee5e38
--- /dev/null
+++ b/records/sd1/sweep/732f94fc011903bcaaf6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6404ab022b3cd0a47d1668d92172cd8d6ef37ae2691555559e3c7c4275bb0838
+size 22862
diff --git a/records/sd1/sweep/7336f9026ff06236cd6b/output.webp b/records/sd1/sweep/7336f9026ff06236cd6b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ea6f735d63d03d54f9c1c6b244331425c1caf516
--- /dev/null
+++ b/records/sd1/sweep/7336f9026ff06236cd6b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f456a7f20d19b29d410c3978d6e5d2d9f69686c20b56b8a92223fe0a3e38844a
+size 29328
diff --git a/records/sd1/sweep/73376160781fd728bc55/output.webp b/records/sd1/sweep/73376160781fd728bc55/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..166efbbb69ef2a6a21973591734c5256556d0f10
--- /dev/null
+++ b/records/sd1/sweep/73376160781fd728bc55/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:13786c15094a06da587bdfdd154fb18d6eb6f9c5111fc2c3f87e6e57e8834f34
+size 22396
diff --git a/records/sd1/sweep/733a3f07eeef31d34c69/output.webp b/records/sd1/sweep/733a3f07eeef31d34c69/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..542f2bc3bd1456aed14259c0005c21282ab49620
--- /dev/null
+++ b/records/sd1/sweep/733a3f07eeef31d34c69/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d743d5b41bea9e706db91d84be39a334f13a75e56ee849a0c34f78576931ae55
+size 18742
diff --git a/records/sd1/sweep/733f6770a9639e7ab512/output.webp b/records/sd1/sweep/733f6770a9639e7ab512/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9c51ab402d4d04da141ba5104a1deea42f8333ef
--- /dev/null
+++ b/records/sd1/sweep/733f6770a9639e7ab512/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73b9e57591f0840795e5f3dee6fdfb84c748a574da0e28d687896ecea1c8c69d
+size 26988
diff --git a/records/sd1/sweep/734afedf39a825340b59/output.webp b/records/sd1/sweep/734afedf39a825340b59/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4ea1641f0d7c6a5082221b6fdb02fe200753f6d5
--- /dev/null
+++ b/records/sd1/sweep/734afedf39a825340b59/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44a4cab2d3e9ad626687b0cb4891617be2106e154d020446a8c63be03a79c895
+size 34428
diff --git a/records/sd1/sweep/734b81edd74addba3f1d/output.webp b/records/sd1/sweep/734b81edd74addba3f1d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..eba5bc23d6bb9928e76473ce40419bf32e562a5f
--- /dev/null
+++ b/records/sd1/sweep/734b81edd74addba3f1d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a14a8816b567c914ad827bf6e6bc779042cb9154192e96266582c7e54a3e1b3
+size 35260
diff --git a/records/sd1/sweep/73559c1b91a5fc2af581/output.webp b/records/sd1/sweep/73559c1b91a5fc2af581/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e6d9cbe867263c6a77ff44b75d306a81b928f066
--- /dev/null
+++ b/records/sd1/sweep/73559c1b91a5fc2af581/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:76831b9c76f61ee8ae63755d741202c93f2002b2057709d016409ef9b09235c0
+size 29528
diff --git a/records/sd1/sweep/736285ae44fdda792df5/output.webp b/records/sd1/sweep/736285ae44fdda792df5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5ea1ad7426d88be5e55e4a612e237edab683edc1
--- /dev/null
+++ b/records/sd1/sweep/736285ae44fdda792df5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8f31962f0455cbe4450a2ca80442ff7780ed486766c2a7f13f7e06b3611e1d9b
+size 31946
diff --git a/records/sd1/sweep/739183a62a8374194c5f/output.webp b/records/sd1/sweep/739183a62a8374194c5f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..573556f6fbd4c1c71a74a103820a3483d74f4278
--- /dev/null
+++ b/records/sd1/sweep/739183a62a8374194c5f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:560f79a81e3d4d79f75148d3b74d3ef8262a5992233d7b4199f8ceff9196c129
+size 49210
diff --git a/records/sd1/sweep/73a766727a1845225b2c/output.webp b/records/sd1/sweep/73a766727a1845225b2c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..eee344826357401d866372df29f89aa214278dd7
--- /dev/null
+++ b/records/sd1/sweep/73a766727a1845225b2c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d76343e86067844e9bcfac1c6f67cd2effff0be01f29ea9d52dca8def1f4e5d
+size 17598
diff --git a/records/sd1/sweep/73ae6eed500115b0f86a/output.webp b/records/sd1/sweep/73ae6eed500115b0f86a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4f158618b321e28a6cb43375e4ceda5b5915c423
--- /dev/null
+++ b/records/sd1/sweep/73ae6eed500115b0f86a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fe48a1d233586d3af72dbc6d83f683a0a9937de20a02b937da5722a2c4e570a
+size 20354
diff --git a/records/sd1/sweep/73b675c11fcf0dd9d856/output.webp b/records/sd1/sweep/73b675c11fcf0dd9d856/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..80a9c3dca9d88416cfb1f66ca65e640625c38e3b
--- /dev/null
+++ b/records/sd1/sweep/73b675c11fcf0dd9d856/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe141a787599bb30e72395bd264cd960a7a909575300ae6d4eb64f3014dd6687
+size 21520
diff --git a/records/sd1/sweep/73bb2db8e85e07039d09/output.webp b/records/sd1/sweep/73bb2db8e85e07039d09/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b09ad48d618d47e0e2d513bff6e0fafebeadbbff
--- /dev/null
+++ b/records/sd1/sweep/73bb2db8e85e07039d09/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d858cc87e3b5abb460619fa3f5522c7f2cc5c3f375a59153c46748f272e8e261
+size 40730
diff --git a/records/sd1/sweep/73bcfe7e51da7c7d9096/output.webp b/records/sd1/sweep/73bcfe7e51da7c7d9096/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6390d4d11f4562e1bffbc3df945b88d95618f190
--- /dev/null
+++ b/records/sd1/sweep/73bcfe7e51da7c7d9096/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9f583db4908accd96ec52f577fbf17c5e17077b1e5ca4d013eee05791981e03e
+size 21128
diff --git a/records/sd1/sweep/73be2c1d5c1e2b169da6/output.webp b/records/sd1/sweep/73be2c1d5c1e2b169da6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0e9000bbb4014319140eef667a28daa927239bf6
--- /dev/null
+++ b/records/sd1/sweep/73be2c1d5c1e2b169da6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0c8162afac8a1adcc7d59f0859c0aedcd33372e7b82838d3f7a5a8111a880554
+size 52924
diff --git a/records/sd1/sweep/73c85a7d3b4381917ba8/output.webp b/records/sd1/sweep/73c85a7d3b4381917ba8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fb6dcad107f301447a5e62400eef592b9787e6f2
--- /dev/null
+++ b/records/sd1/sweep/73c85a7d3b4381917ba8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bf71b5607803db59b851bc44d8c5a834422061e38fc4b3310343a6b6d4bf2f3
+size 53558
diff --git a/records/sd1/sweep/73e10a6dd3a667d8ba1d/output.webp b/records/sd1/sweep/73e10a6dd3a667d8ba1d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5b849026df7664a61db658d52cf680b0206eb5c4
--- /dev/null
+++ b/records/sd1/sweep/73e10a6dd3a667d8ba1d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e722c20ca3d2b0f22c8aed14d0f846ef16774a3ee708c5659d76ebd239f1b476
+size 41314
diff --git a/records/sd1/sweep/73e338bcf73603b1089a/output.webp b/records/sd1/sweep/73e338bcf73603b1089a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6b9de160df64d309766424febccc1698a1a917b0
--- /dev/null
+++ b/records/sd1/sweep/73e338bcf73603b1089a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b2f43153adc4ddfa0574f5855d3b42eb98ac300256267bd4887b38cd080cb86e
+size 34668
diff --git a/records/sd1/sweep/73ec2da2367997be24b8/output.webp b/records/sd1/sweep/73ec2da2367997be24b8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..6c48ae86e24d0b4b361e77eb9d9ca023132778b6
--- /dev/null
+++ b/records/sd1/sweep/73ec2da2367997be24b8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bbcecb23bedaca5468f9d912ab742437a34c4708bf6a4b887165f2455c3251c1
+size 42154
diff --git a/records/sd1/sweep/73f220e9adfcaa1cf470/output.webp b/records/sd1/sweep/73f220e9adfcaa1cf470/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a98937505982cc1b605fd66087d0ae798d53aea9
--- /dev/null
+++ b/records/sd1/sweep/73f220e9adfcaa1cf470/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e9a4e7890434bf3cd53228ab75df46c8af9ce2405d10ba626bbd55176d6f3a8c
+size 23364
diff --git a/records/sd1/sweep/73f665316e96d5258c89/output.webp b/records/sd1/sweep/73f665316e96d5258c89/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..489004abb7d2f9d880bef31f4e2a0c4b7be3178f
--- /dev/null
+++ b/records/sd1/sweep/73f665316e96d5258c89/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f910f7f9ce4390d9fc48c642f337ef570736bbc3b9bc66d84538a2b8bf614c9
+size 15802
diff --git a/records/sd1/sweep/73fa5ade84571fbaf1fe/output.webp b/records/sd1/sweep/73fa5ade84571fbaf1fe/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2a8e2f4b69b251595b9fed4dd4cafdc7b38aafdf
--- /dev/null
+++ b/records/sd1/sweep/73fa5ade84571fbaf1fe/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ffdc3bea3f3fa1f298333ad72432221af38eb6ddaf08be437e4cda2701e00d71
+size 35862
diff --git a/records/sd1/sweep/7404cf5a30eedab1f384/output.webp b/records/sd1/sweep/7404cf5a30eedab1f384/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..17254d93c807a4cb1f9eff72bdfbf190d262c850
--- /dev/null
+++ b/records/sd1/sweep/7404cf5a30eedab1f384/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:470b2ead387ffd4ecdf33cdcb7177d1821999243495821b72da46a9760f527c1
+size 67582
diff --git a/records/sd1/sweep/74057ca6a5c0195915a7/output.webp b/records/sd1/sweep/74057ca6a5c0195915a7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..73a92c5ee537c1f458746fde98e458fff3f09b1d
--- /dev/null
+++ b/records/sd1/sweep/74057ca6a5c0195915a7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9205dd59fb1f29b0a368aac02e1349191538b5791a2d699e80285a2b0dd455e8
+size 40142
diff --git a/records/sd1/sweep/7405efd905d50c99d67e/output.webp b/records/sd1/sweep/7405efd905d50c99d67e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bf880f7e8b8fdc5ed46e13484d1b43666f473937
--- /dev/null
+++ b/records/sd1/sweep/7405efd905d50c99d67e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96fb4a068b6524517f3411786f05c1956db62a0852365e760e9ad26dfb671d31
+size 15236
diff --git a/records/sd1/sweep/740a6f29135d8bb09549/output.webp b/records/sd1/sweep/740a6f29135d8bb09549/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5d436d95cc0aa74dec80717c31d810dd30a63e8f
--- /dev/null
+++ b/records/sd1/sweep/740a6f29135d8bb09549/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cd41576bbc644a89253920c82d88d235d2e99c0c9c60670f03c35120fec55fa
+size 58516
diff --git a/records/sd1/sweep/741bd67fb33512f04634/output.webp b/records/sd1/sweep/741bd67fb33512f04634/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f475d3b56a5166ef1e29ce1eaed4182db83b2a47
--- /dev/null
+++ b/records/sd1/sweep/741bd67fb33512f04634/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5ae910c683488464dffaac5ad0731c20de8e1fbcbb92dbe05eb4a0ab20fcbba4
+size 30620
diff --git a/records/sd1/sweep/742256ab1f88cf4609b5/output.webp b/records/sd1/sweep/742256ab1f88cf4609b5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e7bd8b892371d4b1c6dda78e36c33060ac3875ca
--- /dev/null
+++ b/records/sd1/sweep/742256ab1f88cf4609b5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2299cf0bf6729e73fc4bf93de470159bb21820074204371bd606e4679314d7a
+size 51498
diff --git a/records/sd1/sweep/7424872fb64798275128/output.webp b/records/sd1/sweep/7424872fb64798275128/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f66599abb4cee1800c7e18d80c3f5a9f4f31223d
--- /dev/null
+++ b/records/sd1/sweep/7424872fb64798275128/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ef211134737f9770d8182efd3e3beab34d9b6e3d6e1db1aa3c49182dfc035ca
+size 19928
diff --git a/records/sd1/sweep/742d79019e57b2ffc91e/output.webp b/records/sd1/sweep/742d79019e57b2ffc91e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..19bb20e201f8405f500dd8740d3f8c79b97226c9
--- /dev/null
+++ b/records/sd1/sweep/742d79019e57b2ffc91e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:891989e302558b5b7015d7d7b23d677d96609c47d8e143c2df1b06c091fa5d95
+size 37938
diff --git a/records/sd1/sweep/74307a4d6b60bb5daff9/output.webp b/records/sd1/sweep/74307a4d6b60bb5daff9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bd37143ce0a20e6b9d6f4d4deb573ff4bf18a207
--- /dev/null
+++ b/records/sd1/sweep/74307a4d6b60bb5daff9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f9cde1cabf69302ec156cee88d2b7e7a9e7b0e06882800f11ca54e11ee7fb25b
+size 84054
diff --git a/records/sd1/sweep/7439bde027b9df6fa082/output.webp b/records/sd1/sweep/7439bde027b9df6fa082/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cb713c97da4e792cbfb72d92be57433e6a39a9a1
--- /dev/null
+++ b/records/sd1/sweep/7439bde027b9df6fa082/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a3d4d629a7e0d7ab44153789aade64aba8a507bb222ceb399661d8dcc56ed2c
+size 34818
diff --git a/records/sd1/sweep/743a0f1b7e59b738e41a/output.webp b/records/sd1/sweep/743a0f1b7e59b738e41a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..476f242ab544a6a16805fa400b0dd02818c58b5c
--- /dev/null
+++ b/records/sd1/sweep/743a0f1b7e59b738e41a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:016e92c492d8e53c55bd8778c13e7adfe96f0532fad51ef95730c99272f77b02
+size 34574
diff --git a/records/sd1/sweep/743ab557aabd5946d746/output.webp b/records/sd1/sweep/743ab557aabd5946d746/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a417f896bc2c7d02e203c7b2364e01cb4c5e5a5c
--- /dev/null
+++ b/records/sd1/sweep/743ab557aabd5946d746/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eaca2ee1b8d23e58722960a7abe54e15dd336f95c71ae2c2c0dcb48f99014c55
+size 25238
diff --git a/records/sd1/sweep/743ed4d7632da9d3caa8/output.webp b/records/sd1/sweep/743ed4d7632da9d3caa8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7db71c6f71c78f36d16dd454433c20c160ba85ae
--- /dev/null
+++ b/records/sd1/sweep/743ed4d7632da9d3caa8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e14471516b5259fd4efee058a71e671fe417ed2c1c3a4e9fdc120077ce9622dd
+size 18342
diff --git a/records/sd1/sweep/745eac3fb155dab9cb8d/output.webp b/records/sd1/sweep/745eac3fb155dab9cb8d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..35a8fe28ec587f3488b4a869fb0a72ece7587811
--- /dev/null
+++ b/records/sd1/sweep/745eac3fb155dab9cb8d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f1c14b90b36c13d6af3ce094f87a6f5daad4584203cfc13604aecc79ba99cca
+size 14024
diff --git a/records/sd1/sweep/746314b8a88d993ee25c/output.webp b/records/sd1/sweep/746314b8a88d993ee25c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..adf39777bbc7e1a27fa1ce5e6843448b56499176
--- /dev/null
+++ b/records/sd1/sweep/746314b8a88d993ee25c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f2b82211a325c659ac08a3b4aa0c4b61a2aa4ce5f092443839b45b036e5ec233
+size 32658
diff --git a/records/sd1/sweep/746392216076de39263b/output.webp b/records/sd1/sweep/746392216076de39263b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e372171ebdfaaa40de16b9665577ab268c1450c2
--- /dev/null
+++ b/records/sd1/sweep/746392216076de39263b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a3e45d8131ec43bef0255550efb7e39af0a8f0eb9c62852656b6ebd0a15aa2c
+size 88092
diff --git a/records/sd1/sweep/7469a389c9690bb236c8/output.webp b/records/sd1/sweep/7469a389c9690bb236c8/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a7fd825ddb2887573508115774876ff399b477b6
--- /dev/null
+++ b/records/sd1/sweep/7469a389c9690bb236c8/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa7b57e85adee529a8ca9115a0437257a854b5e4fc359b33d721eece65ce7f2b
+size 51648
diff --git a/records/sd1/sweep/7475786922888486cb27/output.webp b/records/sd1/sweep/7475786922888486cb27/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3fba4712c96295e064ac44207d2bf34706273d96
--- /dev/null
+++ b/records/sd1/sweep/7475786922888486cb27/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8e0b11a8b210430f2ed86454121fc078671bb6d62e4f9a716600b3a63d4cd0f7
+size 36364
diff --git a/records/sd1/sweep/7478ba810aab5674788b/output.webp b/records/sd1/sweep/7478ba810aab5674788b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9041d7af46bc578b40161ddaf9cc777075f937e2
--- /dev/null
+++ b/records/sd1/sweep/7478ba810aab5674788b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:225091390903922aee448ccc5189e72d5ad7ccd98ff184581c41ffda8ac20207
+size 39344
diff --git a/records/sd1/sweep/747ee8001296819322d9/output.webp b/records/sd1/sweep/747ee8001296819322d9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0c9475a44c82f8a52667bd4d8337750dc2d59d66
--- /dev/null
+++ b/records/sd1/sweep/747ee8001296819322d9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5e84966f87a9dd14336f56d60ab26d851d3804bf69fd028b19177eaffa7ddbe
+size 20298
diff --git a/records/sd1/sweep/74827528ffb45c07215f/output.webp b/records/sd1/sweep/74827528ffb45c07215f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4c5e0011142980c1b20ddbeb36863be217947833
--- /dev/null
+++ b/records/sd1/sweep/74827528ffb45c07215f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1a641822764bffa8d3fb89e77fce8df5e8c5aacac543caa7e754d122270b00c
+size 22284
diff --git a/records/sd1/sweep/74859eeb59b96029b94b/output.webp b/records/sd1/sweep/74859eeb59b96029b94b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..36a040662173e8ecb2c79403c059fa0b362e291d
--- /dev/null
+++ b/records/sd1/sweep/74859eeb59b96029b94b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6730ef02e8a1ef270e7d3c31c759f15464f0caff1c39784724c1139be7af9f06
+size 34184
diff --git a/records/sd1/sweep/7499dae1a6d73fcb31bb/output.webp b/records/sd1/sweep/7499dae1a6d73fcb31bb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f765698f7c47b7fe6c146ecc5c88ced6e01b737f
--- /dev/null
+++ b/records/sd1/sweep/7499dae1a6d73fcb31bb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e40df5e43577fe5339da92c67dd61b946ec814a94138a8eb82c52b5b1ca6fb90
+size 26400
diff --git a/records/sd1/sweep/74a089484cb3e4b8fd7b/output.webp b/records/sd1/sweep/74a089484cb3e4b8fd7b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..587447d0c2010efb856be9f032a481a593dc9638
--- /dev/null
+++ b/records/sd1/sweep/74a089484cb3e4b8fd7b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:74565ddc57ac1c17d007bd2cf9cfe533c3e7409a842a2a670faaff69888d9e67
+size 17032
diff --git a/records/sd1/sweep/74a8e04cfaa2f2e91e0a/output.webp b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..af07b0904680363e37bd9030e1396b59b9f41934
--- /dev/null
+++ b/records/sd1/sweep/74a8e04cfaa2f2e91e0a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6770ad30b7b4c13de8c93c4a5eb54fa399717e57aef0dfe2628bb56672834491
+size 31904
diff --git a/records/sd1/sweep/74aa245972c75dabe395/output.webp b/records/sd1/sweep/74aa245972c75dabe395/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..82ca95909a5bd23800ab94b86ae35dd16cedc72c
--- /dev/null
+++ b/records/sd1/sweep/74aa245972c75dabe395/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28120275f3add15a54dcb9c6139e4e5ac3b1e47c0e609aa68141c09453010b02
+size 36998
diff --git a/records/sd1/sweep/74ba99159e0b8f7dd0c7/output.webp b/records/sd1/sweep/74ba99159e0b8f7dd0c7/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..26fa30c53ebcb597a75f0dce97ac5c2976c83ae1
--- /dev/null
+++ b/records/sd1/sweep/74ba99159e0b8f7dd0c7/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ec02f94da9a9be8514412760d1c34fa112594d017af9662677313220abdd2ef
+size 20870
diff --git a/records/sd1/sweep/74bcad6e1bf961dbd754/output.webp b/records/sd1/sweep/74bcad6e1bf961dbd754/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..be8e5a92c8b195c7d242e9bf462aeeb1fe294c40
--- /dev/null
+++ b/records/sd1/sweep/74bcad6e1bf961dbd754/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:81797753844fd71a79d9a27d11f26d5955fbec03d766c8d7bbf488249bfc74f4
+size 52504
diff --git a/records/sd1/sweep/74bd81d571aa43d22fb6/output.webp b/records/sd1/sweep/74bd81d571aa43d22fb6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a5eef21876a8fc226a26c2140957590c61c6d518
--- /dev/null
+++ b/records/sd1/sweep/74bd81d571aa43d22fb6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9601269e5e08fc51cbe82ffd00bd89b40b038c82c9b9c7512f004bf2ac675520
+size 44572
diff --git a/records/sd1/sweep/74c41ebb67d808d6ff33/output.webp b/records/sd1/sweep/74c41ebb67d808d6ff33/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cc21a038422aa9be78064511cd95634ad4960624
--- /dev/null
+++ b/records/sd1/sweep/74c41ebb67d808d6ff33/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:34ad7a99ff53a24ce68e7d01fe9a377bbddca960a273f11d778a113b363f2850
+size 21172
diff --git a/records/sd1/sweep/74c51e1674389d1fac6c/output.webp b/records/sd1/sweep/74c51e1674389d1fac6c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..62b86ff722f4d684e78d9cfc8f4d23dc81d75b41
--- /dev/null
+++ b/records/sd1/sweep/74c51e1674389d1fac6c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d4ed75a7a685b76a75af92bc111244249762f2e5a1caa6c38fa170f4d466ebc
+size 18002
diff --git a/records/sd1/sweep/74ccc9325dbe264a870a/output.webp b/records/sd1/sweep/74ccc9325dbe264a870a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4f5321e7d2d241b2ad1c7edc0a79581908b4b6b6
--- /dev/null
+++ b/records/sd1/sweep/74ccc9325dbe264a870a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5e3b6a6fee9fc7bb03ee7bab74cc9ea95436ae73c04883c5a266dfbc2afb9ca0
+size 54804
diff --git a/records/sd1/sweep/74d233af02933d3913eb/output.webp b/records/sd1/sweep/74d233af02933d3913eb/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..112ef0e3c446f78679b08e159eca208f6dd051a6
--- /dev/null
+++ b/records/sd1/sweep/74d233af02933d3913eb/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:632e5652b682ee10014384d7408d2ce41d56d6b4a692289b4eda88fe25dd1c20
+size 94564
diff --git a/records/sd1/sweep/74d27d2121e2c497c48f/output.webp b/records/sd1/sweep/74d27d2121e2c497c48f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e08ab6cfdbff69c82ce27e2503b31622c8b372db
--- /dev/null
+++ b/records/sd1/sweep/74d27d2121e2c497c48f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:64c46901fa6726f4d4715e8025d59350dd213e7e467f45e8426cd2e31818a49c
+size 21140
diff --git a/records/sd1/sweep/74d34817a345e33aba0f/output.webp b/records/sd1/sweep/74d34817a345e33aba0f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0932d3f8535b2dd2bc8aaa7118eeb63922e61be5
--- /dev/null
+++ b/records/sd1/sweep/74d34817a345e33aba0f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85c64b1a6fe3698b45a25621b4db418686a3e9b8b5af9ce3f0a609345a6961ff
+size 25556
diff --git a/records/sd1/sweep/74db7eaba0d5521af51a/output.webp b/records/sd1/sweep/74db7eaba0d5521af51a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..977ddc480e40ba86f52cb50bae3c41a205db6a80
--- /dev/null
+++ b/records/sd1/sweep/74db7eaba0d5521af51a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4bbefcdb11782234d62812db3ed14ac1a4566eb3fe20592d8ef39bd4f008743d
+size 31898
diff --git a/records/sd1/sweep/74e4f048bba1ce2340b4/output.webp b/records/sd1/sweep/74e4f048bba1ce2340b4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..badf2cc7456f7e49a81d18f5b356f83a46bd4e80
--- /dev/null
+++ b/records/sd1/sweep/74e4f048bba1ce2340b4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c05291ef1b0dcb7dfb8ed08ba6fd6cedab9501882c17348f0bb8feaf7717af46
+size 47252
diff --git a/records/sd1/sweep/74eb40765b650595c75d/output.webp b/records/sd1/sweep/74eb40765b650595c75d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1072d8e2478158f5b3d4bdecd27bcf302e2e6633
--- /dev/null
+++ b/records/sd1/sweep/74eb40765b650595c75d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:032e8e76c4dc5e5a29d0e5587f7ce07ba5b58d40acc2b1aabebc0ed0e2d05e5e
+size 28306
diff --git a/records/sd1/sweep/74f0326d60879943e257/output.webp b/records/sd1/sweep/74f0326d60879943e257/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..caffe00b7a033d238492a4133b0cc143eb412a0c
--- /dev/null
+++ b/records/sd1/sweep/74f0326d60879943e257/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7d23408d72b516998b0b41f18094114000419f9582ce55de8bcbb9383797e136
+size 26176
diff --git a/records/sd1/sweep/74f0940e4ab8127aef65/output.webp b/records/sd1/sweep/74f0940e4ab8127aef65/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b99d814feec60f30356b5fc2a6036c21d600c61c
--- /dev/null
+++ b/records/sd1/sweep/74f0940e4ab8127aef65/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:178265730328886589d80802ccc9fb649b800a50d7ec998a8fc389b8214651a5
+size 48150
diff --git a/records/sd1/sweep/74f1982961914c180316/output.webp b/records/sd1/sweep/74f1982961914c180316/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..b5247ae7025a1f72803cd76fb189b7abf70c888a
--- /dev/null
+++ b/records/sd1/sweep/74f1982961914c180316/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a625dd726e461123a2c6db51d34d96976285d0815744d0867259cfc1fb633f0e
+size 29444
diff --git a/records/sd1/sweep/751110323e7b730885e4/output.webp b/records/sd1/sweep/751110323e7b730885e4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d92045ca60b70a92ca310121c4cf2f3ed343e7cb
--- /dev/null
+++ b/records/sd1/sweep/751110323e7b730885e4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af2ee2894076d60e60657388312adf5b8da9492b7013753b0423a42930fc00b7
+size 50308
diff --git a/records/sd1/sweep/751556cab678a4cb7aee/output.webp b/records/sd1/sweep/751556cab678a4cb7aee/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fd683acd8143eaed7b0e28ba6b4fc70b48e27d3c
--- /dev/null
+++ b/records/sd1/sweep/751556cab678a4cb7aee/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bc9e3b51a31261de67e4bf9f0df5f7e7d6b34c657af494e1c4ed62af23501f4
+size 57018
diff --git a/records/sd1/sweep/751f8261227aa1d68c26/output.webp b/records/sd1/sweep/751f8261227aa1d68c26/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..644ba4226cf7d7d6e749eb2ee74b62b9a7e9435b
--- /dev/null
+++ b/records/sd1/sweep/751f8261227aa1d68c26/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:237a988dd69eb048db4c5b6a7b9475fae561cbffb741e2baac66bc3d5038a12b
+size 32606
diff --git a/records/sd1/sweep/752486356d81dc018f27/output.webp b/records/sd1/sweep/752486356d81dc018f27/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..54956aa2028dcf72fb5eb58c89e61c743cbb62ed
--- /dev/null
+++ b/records/sd1/sweep/752486356d81dc018f27/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f69d7bc5501cdea18e376718dca0e1c072d690a25a0a1632383d7f66a48a63c4
+size 36496
diff --git a/records/sd1/sweep/752a3257da42b124b57a/output.webp b/records/sd1/sweep/752a3257da42b124b57a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..463b424287c8e42a1a1754feb57e788fc0c375aa
--- /dev/null
+++ b/records/sd1/sweep/752a3257da42b124b57a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:261c8c67d8129abeb0f1507bdfbdf48db36d439b11f83a4b0f215b097baf3929
+size 38570
diff --git a/records/sd1/sweep/7535b8f2f0c2a997740c/output.webp b/records/sd1/sweep/7535b8f2f0c2a997740c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fa0d6742fcb29b25f555d742b74f2d2b16f3f0ad
--- /dev/null
+++ b/records/sd1/sweep/7535b8f2f0c2a997740c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cebda9e471c94b142558211e88c22449015cb146037ad7d84506b49e9cf6e927
+size 50088
diff --git a/records/sd1/sweep/7542fd0536db25472e2a/output.webp b/records/sd1/sweep/7542fd0536db25472e2a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3ead2423072a91ec3d2eb8cbbb4b28e87546c07d
--- /dev/null
+++ b/records/sd1/sweep/7542fd0536db25472e2a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf0e7e0d4e32e682063de2cfd7402a2e1218e4e1f3b431421a27a38267d87afe
+size 31412
diff --git a/records/sd1/sweep/755245009e763508bb4a/output.webp b/records/sd1/sweep/755245009e763508bb4a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..71e8a3a4aacf7ac09b24c3228a6eb66e850e1728
--- /dev/null
+++ b/records/sd1/sweep/755245009e763508bb4a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a5e56427a5f320b0fdd2c84a0d9b8971606d25f0fb2b320887bd0d84a0d83a81
+size 32886
diff --git a/records/sd1/sweep/756274ffca019a89953c/output.webp b/records/sd1/sweep/756274ffca019a89953c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..17a8d50a83ee010b6aa5adb91802d4ea22483515
--- /dev/null
+++ b/records/sd1/sweep/756274ffca019a89953c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cd4820d6bb06761c12b8eb658e692b9abebd0c5cbb466c61aafe9ccd05f7f7a
+size 19746
diff --git a/records/sd1/sweep/756ea17f64ab2d98df8e/output.webp b/records/sd1/sweep/756ea17f64ab2d98df8e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..15407afa8b9a676b5800f02244ef8db20ac68f66
--- /dev/null
+++ b/records/sd1/sweep/756ea17f64ab2d98df8e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:06963f2f60c0bc313e0bc57e1cc8e803797d411d39b7db36b8bc78fcfa2ded36
+size 17874
diff --git a/records/sd1/sweep/7570a32d193c99a53ccd/output.webp b/records/sd1/sweep/7570a32d193c99a53ccd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..16a06b03d79c3bb43f8abb86ae4963c0dc8b7a27
--- /dev/null
+++ b/records/sd1/sweep/7570a32d193c99a53ccd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82427e583a7758acad0b42464fb8d94274529c4edb236d5d11e4a68eee21beeb
+size 45376
diff --git a/records/sd1/sweep/7574bbafa84ae25245dd/output.webp b/records/sd1/sweep/7574bbafa84ae25245dd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..60e1aa29ee2a0d28458c56f5db490e674bd41ddc
--- /dev/null
+++ b/records/sd1/sweep/7574bbafa84ae25245dd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ad3128dee1614701b2cd1dbec58e3e923be5c91c906bd6aeb7ed4ca75d771b
+size 26690
diff --git a/records/sd1/sweep/757a0e6a310aa974f7e2/output.webp b/records/sd1/sweep/757a0e6a310aa974f7e2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1fbdaf78100f86628780dcdca0ab22463daf6855
--- /dev/null
+++ b/records/sd1/sweep/757a0e6a310aa974f7e2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e5c3341d6f14e4d472b939ddc25eca8272561814d6e9e585c680d2edd27660e
+size 23508
diff --git a/records/sd1/sweep/7580d74d132c48a96d9d/output.webp b/records/sd1/sweep/7580d74d132c48a96d9d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..41148f0ba53330ede15bd25fee514d0b61dba18e
--- /dev/null
+++ b/records/sd1/sweep/7580d74d132c48a96d9d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbeb489ba4346b27d8b6dbd7f8abb6832a16457a6bdc4628a8d9ceafdfdcc450
+size 30672
diff --git a/records/sd1/sweep/75879c0efb7c5f841a18/output.webp b/records/sd1/sweep/75879c0efb7c5f841a18/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ce1d43191f6a49c258d4de080099ddac6769251b
--- /dev/null
+++ b/records/sd1/sweep/75879c0efb7c5f841a18/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46e89091197548c94403954f021f23690ebc6a3be560cc9be1c8a04b7fe1b0b3
+size 43272
diff --git a/records/sd1/sweep/7587eef302e82ae6125b/output.webp b/records/sd1/sweep/7587eef302e82ae6125b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..159d6ef9beda0f4a5fa96f01fc5e59087467f44d
--- /dev/null
+++ b/records/sd1/sweep/7587eef302e82ae6125b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5240940508846a3bee7e4436856b5cb15ca13975bba34858502cf486ff52cc66
+size 47482
diff --git a/records/sd1/sweep/75a986b7999dcb2c492e/output.webp b/records/sd1/sweep/75a986b7999dcb2c492e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..728c850e9f19b2976d0c3c927f225f6c3374ef1f
--- /dev/null
+++ b/records/sd1/sweep/75a986b7999dcb2c492e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb45f288f8e80d6ad4567da5a48ba68fd55623e23c092e8b0286e59c01f14a32
+size 23252
diff --git a/records/sd1/sweep/75b4ab9b7a463a9f0df3/output.webp b/records/sd1/sweep/75b4ab9b7a463a9f0df3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fd097fee36600155934f66febd8f633222ee83be
--- /dev/null
+++ b/records/sd1/sweep/75b4ab9b7a463a9f0df3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:55cc02397ef23bffad7c61df089fa74fcf741968c9b7824d96c37d364de87460
+size 38980
diff --git a/records/sd1/sweep/75c059b0fa1465204dcd/output.webp b/records/sd1/sweep/75c059b0fa1465204dcd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..296cd11003359aa0698f2bf9f6c835e026e92c93
--- /dev/null
+++ b/records/sd1/sweep/75c059b0fa1465204dcd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3e55a3a89c0d42f2e949636c52002e9a4a3fe39b8cf2ad14b05d3c93c70542e4
+size 34820
diff --git a/records/sd1/sweep/75c1ec8c5737fd06c8ff/output.webp b/records/sd1/sweep/75c1ec8c5737fd06c8ff/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..db4905f72d84ea45e3013b38cee7b15cf96bc0f8
--- /dev/null
+++ b/records/sd1/sweep/75c1ec8c5737fd06c8ff/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f121dfbcfcae12e2f7aad7853b7bc2c82754f2ed9addb14b91e1d2c4c141a8f
+size 86336
diff --git a/records/sd1/sweep/75ce512ff8ac7e0c8b24/output.webp b/records/sd1/sweep/75ce512ff8ac7e0c8b24/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4f0f09c7df88432590c5f0949c6083d6fcf4357c
--- /dev/null
+++ b/records/sd1/sweep/75ce512ff8ac7e0c8b24/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccf19aafb0cf7666272d6c4cef28e688c11e748b842edc56c7bcdccebc63b2b9
+size 27774
diff --git a/records/sd1/sweep/75d3cc2243862c0bcf0e/output.webp b/records/sd1/sweep/75d3cc2243862c0bcf0e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..50813e9dd55d46408c363342572fc3df0860c2b7
--- /dev/null
+++ b/records/sd1/sweep/75d3cc2243862c0bcf0e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:865a9f88963f7346647933dda55f8803eb3c01f31b7d2fdc29765e604ab07613
+size 28958
diff --git a/records/sd1/sweep/75d479a1f49f4c2ba3de/output.webp b/records/sd1/sweep/75d479a1f49f4c2ba3de/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8d76e2f3c3278a0db5deeb0892bfdbe865fe8662
--- /dev/null
+++ b/records/sd1/sweep/75d479a1f49f4c2ba3de/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7dc58a8e8db1a1334dc2d4428ae0843c05223d8b5c640bcca0402507ab520d3
+size 33894
diff --git a/records/sd1/sweep/75e3bd4bc0791981f505/output.webp b/records/sd1/sweep/75e3bd4bc0791981f505/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..028f553de758e01bfb20b4f36b7ee0302a510608
--- /dev/null
+++ b/records/sd1/sweep/75e3bd4bc0791981f505/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b9e922d666ff09236e2b8eaa1c495dff08bc97a2c60b2dbea291b8375fa9ee5
+size 18088
diff --git a/records/sd1/sweep/75e7474411f18b773382/output.webp b/records/sd1/sweep/75e7474411f18b773382/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e0211619f52e2f00ce72f5acbbd16b96f19f9520
--- /dev/null
+++ b/records/sd1/sweep/75e7474411f18b773382/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b87c2f02880f4793451645eff9be016b11219f78db8dc3e92eb4050e539cad
+size 15782
diff --git a/records/sd1/sweep/75e7cb4119b49780d8d2/output.webp b/records/sd1/sweep/75e7cb4119b49780d8d2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..40d675e1a470efeb4620f96f3c94511084aace0e
--- /dev/null
+++ b/records/sd1/sweep/75e7cb4119b49780d8d2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e854a3ea0c25b0e3f08ea4347b4ab24b6c5df793886c775d194f261bbeb2e1b
+size 37686
diff --git a/records/sd1/sweep/75f38de5297e1989463e/output.webp b/records/sd1/sweep/75f38de5297e1989463e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8a00cc1be9989d00c83ce69fe4fb1895db3e1bd8
--- /dev/null
+++ b/records/sd1/sweep/75f38de5297e1989463e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0cbc8150eb602a5ec71b4ac552b5f0b29b706900162ec06318bf3bfe65dda1e1
+size 34532
diff --git a/records/sd1/sweep/75f5ff5179d4387dd92d/output.webp b/records/sd1/sweep/75f5ff5179d4387dd92d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..03a8de37364e1ef0387400e4f178d203c87744ea
--- /dev/null
+++ b/records/sd1/sweep/75f5ff5179d4387dd92d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3ac85563f2ca2cc6bfacfa812d38c93cdc7f619a155f4bd62a2132ea6cc5918
+size 23460
diff --git a/records/sd1/sweep/75fe25b64b48b44eb4aa/output.webp b/records/sd1/sweep/75fe25b64b48b44eb4aa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..54974b69ee80681d92b4e1edef4cb22750531391
--- /dev/null
+++ b/records/sd1/sweep/75fe25b64b48b44eb4aa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:848755cb05bf3d0a22947e07ce014389775859ad720cb7ecb787c190983daa80
+size 51430
diff --git a/records/sd1/sweep/7607e37d955e05ab9a7f/output.webp b/records/sd1/sweep/7607e37d955e05ab9a7f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f9a8a11ed42906c7f6e06600af143f4d206b1d41
--- /dev/null
+++ b/records/sd1/sweep/7607e37d955e05ab9a7f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25a6ff66e19e2ae61ec96125a07f7f89a5bbd920955a422dfeac31fbf2322382
+size 30326
diff --git a/records/sd1/sweep/760bda95ecd9ca137c50/output.webp b/records/sd1/sweep/760bda95ecd9ca137c50/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..42202056362468be97c252d482f0edcff8c75a91
--- /dev/null
+++ b/records/sd1/sweep/760bda95ecd9ca137c50/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56e74510f0f329d8faf86aa09832215f900bda0b7c1615a3d905587fb9e43bb9
+size 19934
diff --git a/records/sd1/sweep/760d7c0efc493f771daa/output.webp b/records/sd1/sweep/760d7c0efc493f771daa/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ff156ffe377243be4214cdfd798ced92f71c10cf
--- /dev/null
+++ b/records/sd1/sweep/760d7c0efc493f771daa/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcdb5cffafc52c96a4486b67bc9230369515bbd00682342b17cb6eea3ab54383
+size 50034
diff --git a/records/sd1/sweep/760f3d1c9d55514d8160/output.webp b/records/sd1/sweep/760f3d1c9d55514d8160/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9e1e8ac62140777f530ff916a02a8f4545196643
--- /dev/null
+++ b/records/sd1/sweep/760f3d1c9d55514d8160/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:726e585c6675ef6cb8f96d81370f3e49bfc17939c82ba6d5762d64ae402def10
+size 75116
diff --git a/records/sd1/sweep/760f67a43f188d356a1e/output.webp b/records/sd1/sweep/760f67a43f188d356a1e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f2eed10cdfec8acbeb2ca16ec7aa00a0601a84e6
--- /dev/null
+++ b/records/sd1/sweep/760f67a43f188d356a1e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bec1baad28d96025b6f78bef01d7281e0f502adbfcf94606f27f9256821b3d28
+size 34828
diff --git a/records/sd1/sweep/761022169d673379fdbd/output.webp b/records/sd1/sweep/761022169d673379fdbd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bc7b584c51e9ec62166c2175df8e13b02f651e16
--- /dev/null
+++ b/records/sd1/sweep/761022169d673379fdbd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a744468664713e68249319b6e7a7ebf7e671ed349db7465875ddd8605e9ce493
+size 32012
diff --git a/records/sd1/sweep/7614eeafca06dd585cc5/output.webp b/records/sd1/sweep/7614eeafca06dd585cc5/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fe9b27a7e09a0a6385c91b48fea09a6456bccde4
--- /dev/null
+++ b/records/sd1/sweep/7614eeafca06dd585cc5/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa33ab7b6c5d683136eb41cb779e2433d9f5a5874f36847f58404aa3526e12a8
+size 31394
diff --git a/records/sd1/sweep/7615a556e5c9ac0aee04/output.webp b/records/sd1/sweep/7615a556e5c9ac0aee04/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fa9560838487cd7bd9a6e3cb7fe78dd49a8a14b0
--- /dev/null
+++ b/records/sd1/sweep/7615a556e5c9ac0aee04/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e7370752e6ab5f73606103ddc763e1197875966236fe3a33f84cbd21ce59174d
+size 16036
diff --git a/records/sd1/sweep/76263c0ebfa26080e9bd/output.webp b/records/sd1/sweep/76263c0ebfa26080e9bd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..142856d5aca1471e0171b546562f2aca67a9fe5e
--- /dev/null
+++ b/records/sd1/sweep/76263c0ebfa26080e9bd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2351cc03faa75e27889f7d85545cceadb8042c47ec271833bd2faa18094ffb1
+size 19934
diff --git a/records/sd1/sweep/763681832d5629b5b429/output.webp b/records/sd1/sweep/763681832d5629b5b429/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2baffeba8cb23f823dc4918cdbd079c1b58bd5f4
--- /dev/null
+++ b/records/sd1/sweep/763681832d5629b5b429/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2760c9cc36183be6ec70731f8eb282f317d8bffdc4fd2373de48831b6b96ab0
+size 47012
diff --git a/records/sd1/sweep/76421b672b13f1812ea6/output.webp b/records/sd1/sweep/76421b672b13f1812ea6/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ae98f9d782607e8cdbd1b5784674660810a2edcd
--- /dev/null
+++ b/records/sd1/sweep/76421b672b13f1812ea6/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2449d0ff5b8634b2f2227a267de877d408fc0fe92c39f72843afcaafdd42e165
+size 46804
diff --git a/records/sd1/sweep/76465717f0d7d62cfa8f/output.webp b/records/sd1/sweep/76465717f0d7d62cfa8f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..90309f7301317c6ecca09de48960b3be24046a29
--- /dev/null
+++ b/records/sd1/sweep/76465717f0d7d62cfa8f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65f3454f0814381c75ead9f69c3e72a03d01b3f252da5978532ded7af527a61e
+size 33552
diff --git a/records/sd1/sweep/7649464d5a645064e65e/output.webp b/records/sd1/sweep/7649464d5a645064e65e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9d44cc39eeb2d0ce6c5df2d4bbf06223b292f236
--- /dev/null
+++ b/records/sd1/sweep/7649464d5a645064e65e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:418a2d2e437dae5aab655aea9a9e346516f200e0cea63badadf688c1c0e7d0bf
+size 32624
diff --git a/records/sd1/sweep/764f9a0e4cadbd6cf26d/output.webp b/records/sd1/sweep/764f9a0e4cadbd6cf26d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9b2dd8aa282f265fda07c570cd53228a69b18bcd
--- /dev/null
+++ b/records/sd1/sweep/764f9a0e4cadbd6cf26d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5483d25dd36ab9f79ba6556196ee3671e216641bb39cffc473ef3a811e759f9a
+size 48690
diff --git a/records/sd1/sweep/7650c0a8d541b0010070/output.webp b/records/sd1/sweep/7650c0a8d541b0010070/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..66ea348dfbb16cc0bdf17382109c0f3e62fd3d59
--- /dev/null
+++ b/records/sd1/sweep/7650c0a8d541b0010070/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:15869994e3137f3ce5a631f30f5adaf53d93e09e10b29c5221e97496d468c9ea
+size 21158
diff --git a/records/sd1/sweep/766a2f066556da8c0b94/output.webp b/records/sd1/sweep/766a2f066556da8c0b94/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..92feca28bbe7745a91c6432125db4ff1fb4ee1b6
--- /dev/null
+++ b/records/sd1/sweep/766a2f066556da8c0b94/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7899b3d0556c785497cc8b095310e17313f612b8529ef30419ee38c1f7003f3
+size 36910
diff --git a/records/sd1/sweep/766c5d31a465fa3f5263/output.webp b/records/sd1/sweep/766c5d31a465fa3f5263/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ab1b4f210ec47bc132c4c75e9e6874c981bc7a1d
--- /dev/null
+++ b/records/sd1/sweep/766c5d31a465fa3f5263/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc22bc7c0db6943ef294e2f98b117b4c0dd4520d516f8e4ab3b6c7867bc39637
+size 47034
diff --git a/records/sd1/sweep/767437988df570d061d0/output.webp b/records/sd1/sweep/767437988df570d061d0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e5ae4c0c1d23d19e63f492649beb56039e6df48a
--- /dev/null
+++ b/records/sd1/sweep/767437988df570d061d0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcc502b01978cfb692f71b5b86158786dcc77cf1425cb14942fcd82257bec111
+size 41398
diff --git a/records/sd1/sweep/7686a1e354e8d12ee8e1/output.webp b/records/sd1/sweep/7686a1e354e8d12ee8e1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..754f6db74138ff7c54d30d716c9a278e13fdc405
--- /dev/null
+++ b/records/sd1/sweep/7686a1e354e8d12ee8e1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46b1546372d62c3e1fe6bae73185ea4d4287de52cc3ab6661b3c378bbbc4c668
+size 42716
diff --git a/records/sd1/sweep/768f9c8a0aaf6fb6e690/output.webp b/records/sd1/sweep/768f9c8a0aaf6fb6e690/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..94d8da8153505f0c3ec335e3feeff924ae285046
--- /dev/null
+++ b/records/sd1/sweep/768f9c8a0aaf6fb6e690/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90997f50ddc38a939765dbdf89d505b1fd2a7374f35b5dffe8a546cc809fb5c4
+size 19936
diff --git a/records/sd1/sweep/769fa3c9bc07097531e9/output.webp b/records/sd1/sweep/769fa3c9bc07097531e9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9a5a7d15d7fdd9c1c1f33eb288fb65f1286eb187
--- /dev/null
+++ b/records/sd1/sweep/769fa3c9bc07097531e9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f8bf78bef184d0ef30d2636dc92116ba602370afb4979af61d81ea1d096c1913
+size 50922
diff --git a/records/sd1/sweep/76a5ac11f615a02d85a1/output.webp b/records/sd1/sweep/76a5ac11f615a02d85a1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..7133f9f58c207607671195d9bfb44057c2b380d6
--- /dev/null
+++ b/records/sd1/sweep/76a5ac11f615a02d85a1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff6e34f9d04a63f8d1452a5219c0ec266df34e8bad5411da9ab661f5c66ac5e7
+size 37268
diff --git a/records/sd1/sweep/76c31564fe769336dc7f/output.webp b/records/sd1/sweep/76c31564fe769336dc7f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0e02c9044d904a3db7c5372ae2c9968506d802dd
--- /dev/null
+++ b/records/sd1/sweep/76c31564fe769336dc7f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:04888bbfa11656eae5737c4a27ebcfe3dbda4008f4ca42d1d70b944e753604a4
+size 42650
diff --git a/records/sd1/sweep/76c6f8813df950fc44a9/output.webp b/records/sd1/sweep/76c6f8813df950fc44a9/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..811d7a99bdecbbfdc4cfb7728f45fc4d4a974aff
--- /dev/null
+++ b/records/sd1/sweep/76c6f8813df950fc44a9/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fba5cf31a9600ff378b30a02b539ce793c84b9f9ca487448f42087d5283046df
+size 33696
diff --git a/records/sd1/sweep/76c96b2115238f16e764/output.webp b/records/sd1/sweep/76c96b2115238f16e764/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5d6cca4eae3a80efa22d1c973baee5c81d8f4ba6
--- /dev/null
+++ b/records/sd1/sweep/76c96b2115238f16e764/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0ee0bdcfe1d9e841cba4486a76530833a30d172f298c56d7baba6f6a87cd5983
+size 37074
diff --git a/records/sd1/sweep/76ccabcc92e22173932a/output.webp b/records/sd1/sweep/76ccabcc92e22173932a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..80e6997a892cfe1945a5f9df2abd023687625d44
--- /dev/null
+++ b/records/sd1/sweep/76ccabcc92e22173932a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af3ce74831a582359fabba7c30f99d34d85ac942c1b9d6f8fb7eed4d2763605b
+size 30456
diff --git a/records/sd1/sweep/76ce621ac63011ab87a0/output.webp b/records/sd1/sweep/76ce621ac63011ab87a0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f2b4247e35a196788e9269b4d58c71e04f8c636b
--- /dev/null
+++ b/records/sd1/sweep/76ce621ac63011ab87a0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:663c105b1af5c95176d8da32c7208540bbbfb9ef4993f82860cbf776aa7cc785
+size 35432
diff --git a/records/sd1/sweep/76e5311849a8ead02a49/output.webp b/records/sd1/sweep/76e5311849a8ead02a49/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..bf3772c7f412acb8bc0f9655e6b5a181d3f3981f
--- /dev/null
+++ b/records/sd1/sweep/76e5311849a8ead02a49/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:05aea4076575743ffaed8d5fd564c336f18b250701061fdb576a019c843cba54
+size 25138
diff --git a/records/sd1/sweep/76e9a7eee645551f0df1/output.webp b/records/sd1/sweep/76e9a7eee645551f0df1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..25bc49ad5af0dac0aafe7f5a79cb4c197636b7a5
--- /dev/null
+++ b/records/sd1/sweep/76e9a7eee645551f0df1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4be67756948ec5be1e7ebf84ec72a516bcc7bb1b6f714894c80ba36965c56e0
+size 19214
diff --git a/records/sd1/sweep/76ed281908a353f7860b/output.webp b/records/sd1/sweep/76ed281908a353f7860b/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..85b723c803d5167926f5d2a0bb54099c9979fb60
--- /dev/null
+++ b/records/sd1/sweep/76ed281908a353f7860b/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70cf29b350976e06141b486c207453adf9d5ce64a6990ca68e6c4b7fb62f9c21
+size 24382
diff --git a/records/sd1/sweep/76f06d3d3896548e1c36/output.webp b/records/sd1/sweep/76f06d3d3896548e1c36/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..621abb26aeef5d679dccc9c51736c7d466c775cc
--- /dev/null
+++ b/records/sd1/sweep/76f06d3d3896548e1c36/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfae25d0f2792c079d633b81e9f1eaebbc2c3740d6e99e96cabf87bed6cc3a25
+size 51786
diff --git a/records/sd1/sweep/76f09b12c08703304709/output.webp b/records/sd1/sweep/76f09b12c08703304709/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cbf34f2b89419d313a6cd66f8ef467ee992a147a
--- /dev/null
+++ b/records/sd1/sweep/76f09b12c08703304709/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ba6a4bb2cfad5d1112ccadaac6fa4e37ce57d0433ee3152f6fd233069edd8ba
+size 24978
diff --git a/records/sd1/sweep/7704ed862c810e1b9f5d/output.webp b/records/sd1/sweep/7704ed862c810e1b9f5d/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2589dd869d345cd2335cf1faf7965776f0b5786b
--- /dev/null
+++ b/records/sd1/sweep/7704ed862c810e1b9f5d/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd39a66284df98f222df55d2f9490aa12b1ae0849f20967209455d1ce8823051
+size 33734
diff --git a/records/sd1/sweep/7711d93b53cd2103d978/output.webp b/records/sd1/sweep/7711d93b53cd2103d978/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..f1ae864691ddb45bd02f937575f2b8e5ad52540d
--- /dev/null
+++ b/records/sd1/sweep/7711d93b53cd2103d978/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1ba4b91fcce2798afef82aa02540dd34b20dac755b7cc6cb9a9fe4ba7c26bcb
+size 37588
diff --git a/records/sd1/sweep/771ab44400857ecaa31c/output.webp b/records/sd1/sweep/771ab44400857ecaa31c/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..660cdb2301d9a5d380ef270021d2439b8a5ed388
--- /dev/null
+++ b/records/sd1/sweep/771ab44400857ecaa31c/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d232c26c3579adb702390d683c199145611ddae074c0234ba609ac2850d3cd0
+size 19932
diff --git a/records/sd1/sweep/7721fd671efa916092cd/output.webp b/records/sd1/sweep/7721fd671efa916092cd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8da347ac84de7cf11240a184a6583f9c886c7f6a
--- /dev/null
+++ b/records/sd1/sweep/7721fd671efa916092cd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19d2f562ebb47fedde3e2dff194d00f1790de851462819d63588fd6b34779cf0
+size 36504
diff --git a/records/sd1/sweep/7723a7a55c2419799386/output.webp b/records/sd1/sweep/7723a7a55c2419799386/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e0674a57fd7c346e0157952de6f8f8f60cd959ec
--- /dev/null
+++ b/records/sd1/sweep/7723a7a55c2419799386/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0bba7e1be46d037d79a5259fd2f1e0ed0b499c2acfbe9b46dd8530fe03ec15b
+size 19936
diff --git a/records/sd1/sweep/7727f8c6478ab69fed00/output.webp b/records/sd1/sweep/7727f8c6478ab69fed00/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..4b39147f8d3e129e90ae184659d568c98da07e10
--- /dev/null
+++ b/records/sd1/sweep/7727f8c6478ab69fed00/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:359ed6f8425411fc107b8f418e13fe0ab9b98eab3060971162998a350d4f5bcb
+size 27002
diff --git a/records/sd1/sweep/7729864c3cc7ced84036/output.webp b/records/sd1/sweep/7729864c3cc7ced84036/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..8702f2391f7ee4cad0ce355689e1f90bee9ebca1
--- /dev/null
+++ b/records/sd1/sweep/7729864c3cc7ced84036/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:853047f97e0c627cb75aeebd5c9b269e123c0743f39b9b1c5db5a3ee0ba9eda8
+size 37292
diff --git a/records/sd1/sweep/77309306157d0e3ba826/output.webp b/records/sd1/sweep/77309306157d0e3ba826/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c833c2f6bfc8ffb7446f70a0343376835caed68d
--- /dev/null
+++ b/records/sd1/sweep/77309306157d0e3ba826/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af23734c29dcf871f231001665f6d37712b6c201025b466c125807eba6cb734a
+size 37768
diff --git a/records/sd1/sweep/77360af8089199cf2a10/output.webp b/records/sd1/sweep/77360af8089199cf2a10/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..229fbca3b75756f162f6f62b0feaa6c2060ebedc
--- /dev/null
+++ b/records/sd1/sweep/77360af8089199cf2a10/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef5a12312a5e874b652eee8a3593d7dd7663a7f90d9c568a6724768e06a0639
+size 19664
diff --git a/records/sd1/sweep/773a8c6bf9fd24479fe1/output.webp b/records/sd1/sweep/773a8c6bf9fd24479fe1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d8fcf94ef3d77e8f3a23f5c1ddfc645adc3c70db
--- /dev/null
+++ b/records/sd1/sweep/773a8c6bf9fd24479fe1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1968cd0fa3d21278fbe7bc3b266e4433de4f5bceceb20e603b0d5a6ad41a3197
+size 62606
diff --git a/records/sd1/sweep/773c6662fee5b68b9579/output.webp b/records/sd1/sweep/773c6662fee5b68b9579/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e87bda501cac06cd126fef0aa9d92f6b4e16037e
--- /dev/null
+++ b/records/sd1/sweep/773c6662fee5b68b9579/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fd3572383b6a4858b4c02424dfd28d5a12271800edf0e16721becf26dd37572f
+size 30166
diff --git a/records/sd1/sweep/773df460d23c71c022f4/output.webp b/records/sd1/sweep/773df460d23c71c022f4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0df98aafc8e9de205b347e6d539a195dab22d6e6
--- /dev/null
+++ b/records/sd1/sweep/773df460d23c71c022f4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dd829095ffbb3c937c73aa62db75ac02104916a9bd4d04a16c88c27f1e3067c
+size 31038
diff --git a/records/sd1/sweep/773e185bdcc32839431e/output.webp b/records/sd1/sweep/773e185bdcc32839431e/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0f36bd1f45c6ff8cf14beff34bb26878dc3f99a2
--- /dev/null
+++ b/records/sd1/sweep/773e185bdcc32839431e/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:916e090e48e2516d1740bac18b2c4e45bde9ba541e22ad2aed839f241fff5b1a
+size 35972
diff --git a/records/sd1/sweep/7745a95002895fbd48f0/output.webp b/records/sd1/sweep/7745a95002895fbd48f0/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..48956505e555af6fdd715c2cf027458514ea179b
--- /dev/null
+++ b/records/sd1/sweep/7745a95002895fbd48f0/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b456ba3963b4939b71b5d58660703d7b08688e963748b07576c65ca1c04161af
+size 44148
diff --git a/records/sd1/sweep/774d92fb612576c939cf/output.webp b/records/sd1/sweep/774d92fb612576c939cf/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..50c1ed4f874d2255f5d15615b42e52024e0eb111
--- /dev/null
+++ b/records/sd1/sweep/774d92fb612576c939cf/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9ff367bb003422d3ab9d96ba750bd90fd5c720aa93455b07d50dfe065aeb9a3
+size 34234
diff --git a/records/sd1/sweep/774f902671c21085e6a4/output.webp b/records/sd1/sweep/774f902671c21085e6a4/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3109d3bae079fb195b6f35178fb97109be73a054
--- /dev/null
+++ b/records/sd1/sweep/774f902671c21085e6a4/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57d6220c76ebc9135c47551a63561f4ebde87aa2bab75d3f00e0f6a3269ac5fa
+size 29354
diff --git a/records/sd1/sweep/776165928da293234a58/output.webp b/records/sd1/sweep/776165928da293234a58/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e2d667a818b18afe0e4cedd4acc8b552cc539a51
--- /dev/null
+++ b/records/sd1/sweep/776165928da293234a58/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:70c0b2df9cd62cf93eeb636a7f4c1ff3162e9b108415ecb9f21a087e7637020e
+size 25622
diff --git a/records/sd1/sweep/7766ec77d9b1c0c14020/output.webp b/records/sd1/sweep/7766ec77d9b1c0c14020/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..fe8bd3694c609a600a2e2f82ed9cde910a128cc8
--- /dev/null
+++ b/records/sd1/sweep/7766ec77d9b1c0c14020/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1937aa7cfc0806f028e047e4fa58a115a50629fbf91f7a9c4ab3e18ac22c0d93
+size 23744
diff --git a/records/sd1/sweep/77684dfd84da99e04c03/output.webp b/records/sd1/sweep/77684dfd84da99e04c03/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..aa4422426f86db3a0d3d57782afdc99e70e58eb6
--- /dev/null
+++ b/records/sd1/sweep/77684dfd84da99e04c03/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b7d1f0ed18d3ec69fd54fd407c8ccfca0617e97d183484813c1f7b7f070d5def
+size 54330
diff --git a/records/sd1/sweep/7769ff941bb4634787e2/output.webp b/records/sd1/sweep/7769ff941bb4634787e2/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..ac2530574361466c8b2e3267c49198437c420e65
--- /dev/null
+++ b/records/sd1/sweep/7769ff941bb4634787e2/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0c32fcc984f1727aca1598a0012a77e4fb32655157d7c173be32b2d9507c783
+size 27804
diff --git a/records/sd1/sweep/776a1ff454943ee0151f/output.webp b/records/sd1/sweep/776a1ff454943ee0151f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..db24aa222ea96e71dbff1e0533c3e454f129197d
--- /dev/null
+++ b/records/sd1/sweep/776a1ff454943ee0151f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:244a3cd9cb5febc2ad5e7231a59ad7ce9f607d04a2ffe6e9f1a0021da56e1d94
+size 34690
diff --git a/records/sd1/sweep/776f1756e2360bff46d3/output.webp b/records/sd1/sweep/776f1756e2360bff46d3/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..e525f37d4427d4272d40706c7bae802544cb0887
--- /dev/null
+++ b/records/sd1/sweep/776f1756e2360bff46d3/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c79c7706636a8544f6539ca4e4168043a8c6594cfd6f770c7684320dd5cf07dd
+size 56638
diff --git a/records/sd1/sweep/777eb2bbeb0450a00126/output.webp b/records/sd1/sweep/777eb2bbeb0450a00126/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..71ea26df13c02a0933d5f3d67d632d6ce49b4a7c
--- /dev/null
+++ b/records/sd1/sweep/777eb2bbeb0450a00126/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7013d9ceb7591148c8c3c378e84c86bd6a6ce77c2a42b5b7c201d36cd9b59699
+size 29528
diff --git a/records/sd1/sweep/77bb6e1ea3a67fd45780/output.webp b/records/sd1/sweep/77bb6e1ea3a67fd45780/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..5659f76adeb52adaf555b122c836a5065c5b3949
--- /dev/null
+++ b/records/sd1/sweep/77bb6e1ea3a67fd45780/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0f7783cb2752f7d802fbb27897e1f7bdd2d4733e8296a6fcedd9b7e81cb5a410
+size 37734
diff --git a/records/sd1/sweep/77c0e9e77219ac5eeb94/output.webp b/records/sd1/sweep/77c0e9e77219ac5eeb94/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..86bcdc8881f4887cf1ea27a3e75296fdd3e3c8ff
--- /dev/null
+++ b/records/sd1/sweep/77c0e9e77219ac5eeb94/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1c5894f82d749a3de2c765ec2419746efff48fb25bee436c9afec04bf09791f1
+size 22520
diff --git a/records/sd1/sweep/77c850588b59d524ea36/output.webp b/records/sd1/sweep/77c850588b59d524ea36/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..a0b4ff32b1f15906ddc30f67b5491ff1853ed66e
--- /dev/null
+++ b/records/sd1/sweep/77c850588b59d524ea36/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a91abcbe3a76f6e6b6fd0135584d07410791a2540e5fc2714992d1673a03edbf
+size 38516
diff --git a/records/sd1/sweep/77ced9b2372f268fdc8f/output.webp b/records/sd1/sweep/77ced9b2372f268fdc8f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..3973eb8cf18833ee740a01445efdb2afae5099ce
--- /dev/null
+++ b/records/sd1/sweep/77ced9b2372f268fdc8f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f65ea14352734c4137feb0b63d00527fe390f3a708ef4a40f730fefe14dea072
+size 31154
diff --git a/records/sd1/sweep/77d7e83376bc888e94ff/output.webp b/records/sd1/sweep/77d7e83376bc888e94ff/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..cd87cfb3e4e2fd7aeffbc6b3742a67f1b4591d4c
--- /dev/null
+++ b/records/sd1/sweep/77d7e83376bc888e94ff/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:425880e37fb53b67c4de426a74781d60f62cb4b6e3ab9371779ba0fe893e6e03
+size 26244
diff --git a/records/sd1/sweep/77e054db3c219b1f0838/output.webp b/records/sd1/sweep/77e054db3c219b1f0838/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..d97f94cd7733d49f4f847faa30af886a5c511de8
--- /dev/null
+++ b/records/sd1/sweep/77e054db3c219b1f0838/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9d31568f7563a07ed61301e4cd4af51cea94436904b175de09bcfa039df41fea
+size 36504
diff --git a/records/sd1/sweep/77ecf594e6e71c88f48a/output.webp b/records/sd1/sweep/77ecf594e6e71c88f48a/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..0653205855b9ff680aad07b126c827da57156808
--- /dev/null
+++ b/records/sd1/sweep/77ecf594e6e71c88f48a/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dbd9f7376fd3ccc79d78f4c23466e19b1385d9e42fbc1a18ce4174bc7f0e0601
+size 17722
diff --git a/records/sd1/sweep/77f38075fa85ac261fcd/output.webp b/records/sd1/sweep/77f38075fa85ac261fcd/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..c9bdf15987d0d1a8be81be70ca0b03be94e79420
--- /dev/null
+++ b/records/sd1/sweep/77f38075fa85ac261fcd/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9244f9ab02b66882872a5aed4c8c12937ce4b6c1f0fb61b20df9ea0f4f484cd4
+size 42780
diff --git a/records/sd1/sweep/780491e63d6aa6950a0f/output.webp b/records/sd1/sweep/780491e63d6aa6950a0f/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..1aab6d908d7a06b2746acfeec41c5981d1d60569
--- /dev/null
+++ b/records/sd1/sweep/780491e63d6aa6950a0f/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb2dd232ded4536b2729f5dbc36a855b6546ddb34536392c41b3768d442069cf
+size 43998
diff --git a/records/sd1/sweep/780ca188aa4fc6b233f1/output.webp b/records/sd1/sweep/780ca188aa4fc6b233f1/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..2dcd133543fb2692d539c9cc90661c8e09397d74
--- /dev/null
+++ b/records/sd1/sweep/780ca188aa4fc6b233f1/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eaa41344b766bf30c2ef4f8e6fa4dea441344ad055c55ca3d75da888675c9fb6
+size 47902
diff --git a/records/sd1/sweep/780fe68ea2b0185dd348/output.webp b/records/sd1/sweep/780fe68ea2b0185dd348/output.webp
new file mode 100644
index 0000000000000000000000000000000000000000..9d7e61659dc4232d870799f0801508bf9ff304ea
--- /dev/null
+++ b/records/sd1/sweep/780fe68ea2b0185dd348/output.webp
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03274c2d9ce27d1bfebac49d53d4d1db169a25759449de4a369a9543b12dcfb5
+size 28448
diff --git a/records/sd1/sweep/817dba243af5c13fbf1a/measurements.json b/records/sd1/sweep/817dba243af5c13fbf1a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6911d64e161921d096970dc08604f94beca9de3b
--- /dev/null
+++ b/records/sd1/sweep/817dba243af5c13fbf1a/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "817dba243af5c13fbf1a",
+ "values": {
+  "std": {
+   "value": 75.289,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00279,
+   "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": 69.726,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.639239,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.738802,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 1.030736,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.400332,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.5323,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9607,
+   "method": "CLIP zero-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/817dba243af5c13fbf1a/workflow.json b/records/sd1/sweep/817dba243af5c13fbf1a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b04a1adfe263fb48cc42230d108319dc5bf73777
--- /dev/null
+++ b/records/sd1/sweep/817dba243af5c13fbf1a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 817dba243af5c13fbf1a)"
+  },
+  "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 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/817fd5f9d0b9234d5050/measurements.json b/records/sd1/sweep/817fd5f9d0b9234d5050/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f98c6c7a2df80b8f666f8f3764545ce110a7efe
--- /dev/null
+++ b/records/sd1/sweep/817fd5f9d0b9234d5050/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "817fd5f9d0b9234d5050",
+ "values": {
+  "cosine_distance": {
+   "value": 0.313868,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.415122,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.287404,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.119734,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.706,
+   "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": 36.868,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9768,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0445,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/817fd5f9d0b9234d5050/record.json b/records/sd1/sweep/817fd5f9d0b9234d5050/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5fe7e0747ad1bab20abfe4a9963052d6403779f
--- /dev/null
+++ b/records/sd1/sweep/817fd5f9d0b9234d5050/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "817fd5f9d0b9234d5050",
+ "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
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/2470226a3ec89435a1724950c4c71623fdf6b4c96199b8e241ddf0850b12b268.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2470226a3ec89435a1724950c4c71623fdf6b4c96199b8e241ddf0850b12b268",
+  "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": "051cfa0d3612be65aa5a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/817fd5f9d0b9234d5050/workflow.json b/records/sd1/sweep/817fd5f9d0b9234d5050/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ed36bbe9b198b8aab09e6623bb4148058d65f27b
--- /dev/null
+++ b/records/sd1/sweep/817fd5f9d0b9234d5050/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 817fd5f9d0b9234d5050)"
+  },
+  "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\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/818516d094ac240b8f22/measurements.json b/records/sd1/sweep/818516d094ac240b8f22/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ad84858ffeadee49f191cd77d6f180ad9cae224
--- /dev/null
+++ b/records/sd1/sweep/818516d094ac240b8f22/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "818516d094ac240b8f22",
+ "values": {
+  "std": {
+   "value": 51.574,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00844,
+   "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": 32.218,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.324699,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.484151,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.459215,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.063361,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9573,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.016,
+   "method": "CLIP zero-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/818516d094ac240b8f22/record.json b/records/sd1/sweep/818516d094ac240b8f22/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca0be8e7793148fefa098915e68803d146eb8cb2
--- /dev/null
+++ b/records/sd1/sweep/818516d094ac240b8f22/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "818516d094ac240b8f22",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:36:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "zero",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "multiply",
+   0.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for in.lo (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "a4de854741286c63c9fb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/818516d094ac240b8f22/workflow.json b/records/sd1/sweep/818516d094ac240b8f22/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..58ff68e5b5a769e3d7e58ead32d655b0dfda1202
--- /dev/null
+++ b/records/sd1/sweep/818516d094ac240b8f22/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 818516d094ac240b8f22)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"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/8187b872456ef3858359/measurements.json b/records/sd1/sweep/8187b872456ef3858359/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c820d7c2663dca6906ac4c7d78d44e344be70a1
--- /dev/null
+++ b/records/sd1/sweep/8187b872456ef3858359/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8187b872456ef3858359",
+ "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/8187b872456ef3858359/record.json b/records/sd1/sweep/8187b872456ef3858359/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc1bc31c2c6738c326ec24f6b7e1db130d0eb97a
--- /dev/null
+++ b/records/sd1/sweep/8187b872456ef3858359/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8187b872456ef3858359",
+ "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": 0.5
+   },
+   "group": "out.lo",
+   "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/397d02bcd1d0b52c600436a9b30bdc2c66382a0ecd015923161569e553a83189.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "397d02bcd1d0b52c600436a9b30bdc2c66382a0ecd015923161569e553a83189",
+  "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": "caba9584db8668cd39f7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8187b872456ef3858359/workflow.json b/records/sd1/sweep/8187b872456ef3858359/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..69950731a37b8c98ec985e54b26d75693e305e22
--- /dev/null
+++ b/records/sd1/sweep/8187b872456ef3858359/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 8187b872456ef3858359)"
+  },
+  "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\": 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/818c75c6c4c1a1d8f47f/measurements.json b/records/sd1/sweep/818c75c6c4c1a1d8f47f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..636d03eeced4b99b733ebe9e29e62881adc78d6e
--- /dev/null
+++ b/records/sd1/sweep/818c75c6c4c1a1d8f47f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "818c75c6c4c1a1d8f47f",
+ "values": {
+  "std": {
+   "value": 92.193,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00276,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 82.908,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.779074,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.823212,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.489013,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.329501,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9244,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2672,
+   "method": "CLIP zero-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/818c75c6c4c1a1d8f47f/record.json b/records/sd1/sweep/818c75c6c4c1a1d8f47f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..175f77354b707b04f356e830cf96733dcfac1523
--- /dev/null
+++ b/records/sd1/sweep/818c75c6c4c1a1d8f47f/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "818c75c6c4c1a1d8f47f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:15:22Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "slight",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   30.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 30.0,
+    "applied": 30.0,
+    "reason": "outside safe range [0.0, 5.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "93e18f53bca74ec4a31b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/818c75c6c4c1a1d8f47f/workflow.json b/records/sd1/sweep/818c75c6c4c1a1d8f47f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..47b67d6dbaa648a2901b67680aed7a178a8b68e9
--- /dev/null
+++ b/records/sd1/sweep/818c75c6c4c1a1d8f47f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 818c75c6c4c1a1d8f47f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/818ce53dbab9e275b513/measurements.json b/records/sd1/sweep/818ce53dbab9e275b513/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..613b69dad7f6ead40d6caf4279f1070407ec1485
--- /dev/null
+++ b/records/sd1/sweep/818ce53dbab9e275b513/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "818ce53dbab9e275b513",
+ "values": {
+  "std": {
+   "value": 57.731,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01423,
+   "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": 32.85,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.268352,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.377687,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.366541,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.042898,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9832,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.287,
+   "method": "CLIP zero-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/818ce53dbab9e275b513/record.json b/records/sd1/sweep/818ce53dbab9e275b513/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..aae31cbc76dd1247298e0b567ba95b9d3c5d265e
--- /dev/null
+++ b/records/sd1/sweep/818ce53dbab9e275b513/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "818ce53dbab9e275b513",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:08:56Z",
+ "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": 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": "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": "13216623a742ccb32c5e",
+  "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": "ca7891cf1aa933659464"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/818ce53dbab9e275b513/workflow.json b/records/sd1/sweep/818ce53dbab9e275b513/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecab9cecf155ce26087abedec874d3665edc665e
--- /dev/null
+++ b/records/sd1/sweep/818ce53dbab9e275b513/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 818ce53dbab9e275b513)"
+  },
+  "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 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/818cff8380b61382211e/measurements.json b/records/sd1/sweep/818cff8380b61382211e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9045cac4191aae6f212fec7435c1ce71efa355b0
--- /dev/null
+++ b/records/sd1/sweep/818cff8380b61382211e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "818cff8380b61382211e",
+ "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/818cff8380b61382211e/record.json b/records/sd1/sweep/818cff8380b61382211e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..62dda4eb59bfc7f163a3c0938b8c4d473a8ba8a1
--- /dev/null
+++ b/records/sd1/sweep/818cff8380b61382211e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "818cff8380b61382211e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:42Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "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/67ff263668fd51a10ccbc8e764151b6dfa173e7356fb22a21d31d65b7c5141d7.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "67ff263668fd51a10ccbc8e764151b6dfa173e7356fb22a21d31d65b7c5141d7",
+  "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": "2c786a2130d5c7494119"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/818cff8380b61382211e/workflow.json b/records/sd1/sweep/818cff8380b61382211e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5904995c112fb68f470ee02567dd7e7624454415
--- /dev/null
+++ b/records/sd1/sweep/818cff8380b61382211e/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 818cff8380b61382211e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.h_upd\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/819d955ef27b4a6f4fb7/interpretations.jsonl b/records/sd1/sweep/819d955ef27b4a6f4fb7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7798956c04cf763b28924600d1edccf48f44551d
--- /dev/null
+++ b/records/sd1/sweep/819d955ef27b4a6f4fb7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "819d955ef27b4a6f4fb7"}, "kind": "caption", "value": "Realistic young woman with a light brown bob and blue eyes, on a soft grey background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bd52125d23e2bbd1"}
+{"target": {"record": "819d955ef27b4a6f4fb7"}, "kind": "change", "value": "The style shifts from stylised illustration toward soft realism: more natural skin tones, lighter brown hair and a plain backdrop, with the paper frame gone.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "192c65009278d8cb"}
+{"target": {"record": "819d955ef27b4a6f4fb7"}, "kind": "keywords", "value": ["realistic portrait", "soft realism", "brown bob", "natural skin", "style shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9f7b74ab27114175"}
+{"target": {"record": "819d955ef27b4a6f4fb7"}, "kind": "effect_tags", "value": ["subject_kept", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:24:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6ad842ca29839281"}
diff --git a/records/sd1/sweep/819d955ef27b4a6f4fb7/measurements.json b/records/sd1/sweep/819d955ef27b4a6f4fb7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..03679a22125561e908911c716263f5075254532d
--- /dev/null
+++ b/records/sd1/sweep/819d955ef27b4a6f4fb7/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "819d955ef27b4a6f4fb7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.287792,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.308449,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.112503,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.048607,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.394,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00437,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 32.158,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9608,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1585,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/819d955ef27b4a6f4fb7/record.json b/records/sd1/sweep/819d955ef27b4a6f4fb7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..25a375e264dfbc5e3d6bf769f669a9e5bec008dd
--- /dev/null
+++ b/records/sd1/sweep/819d955ef27b4a6f4fb7/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "819d955ef27b4a6f4fb7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:10Z",
+ "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.2.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "mid",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/6efba8c54ae89e8d9ebf8a29089890f5cb5709d266a4ea7a69719a62b85bc90c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6efba8c54ae89e8d9ebf8a29089890f5cb5709d266a4ea7a69719a62b85bc90c",
+  "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": "d1db6c5f9b102ebcfacc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/819d955ef27b4a6f4fb7/workflow.json b/records/sd1/sweep/819d955ef27b4a6f4fb7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..92fceff58c7bb45255dd5f25475db378679f1519
--- /dev/null
+++ b/records/sd1/sweep/819d955ef27b4a6f4fb7/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 819d955ef27b4a6f4fb7)"
+  },
+  "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\": 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/81a6890c37622dcabaf9/measurements.json b/records/sd1/sweep/81a6890c37622dcabaf9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd5013dcc4e7c8d6a3eeddc7769903db1fcc044e
--- /dev/null
+++ b/records/sd1/sweep/81a6890c37622dcabaf9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "81a6890c37622dcabaf9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.066975,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.094488,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.026269,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023466,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.023,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00573,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.565,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.126,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81a6890c37622dcabaf9/record.json b/records/sd1/sweep/81a6890c37622dcabaf9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c8f7acd4e6e4203fd0700cd75b85e97f22a3947
--- /dev/null
+++ b/records/sd1/sweep/81a6890c37622dcabaf9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "81a6890c37622dcabaf9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:12Z",
+ "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": "middle_block.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "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/21274d7b3bb363cdf7b5473c25ee0fdab7fa6eb07ae72031ea22ab06a29dfa9b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "21274d7b3bb363cdf7b5473c25ee0fdab7fa6eb07ae72031ea22ab06a29dfa9b",
+  "type_path": "UNetModel.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": "0e8e41a5d5e68750246f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81a6890c37622dcabaf9/workflow.json b/records/sd1/sweep/81a6890c37622dcabaf9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc83d44d0f166345e7042a6eafb2e849407be29c
--- /dev/null
+++ b/records/sd1/sweep/81a6890c37622dcabaf9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 81a6890c37622dcabaf9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.attn1.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81a852dc57f9a94574b0/measurements.json b/records/sd1/sweep/81a852dc57f9a94574b0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d99dcd12e1ad1cb369c132cb88cf1b72d98ce14d
--- /dev/null
+++ b/records/sd1/sweep/81a852dc57f9a94574b0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "81a852dc57f9a94574b0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.172431,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.273201,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.123879,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.090874,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.039,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0051,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.21,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9432,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0625,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81a852dc57f9a94574b0/record.json b/records/sd1/sweep/81a852dc57f9a94574b0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..191521170caab0a1b6644638567df392231093c6
--- /dev/null
+++ b/records/sd1/sweep/81a852dc57f9a94574b0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "81a852dc57f9a94574b0",
+ "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": 1.5
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/56271b38c1ad10e96d24466e9f236d2f4a905c8a7744a271b9701a56d2414237.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "56271b38c1ad10e96d24466e9f236d2f4a905c8a7744a271b9701a56d2414237",
+  "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": "eb739f86c8382d30edf0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81a852dc57f9a94574b0/workflow.json b/records/sd1/sweep/81a852dc57f9a94574b0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b9f0f2fea9701bab85591a5f8bc720e87743a04
--- /dev/null
+++ b/records/sd1/sweep/81a852dc57f9a94574b0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 81a852dc57f9a94574b0)"
+  },
+  "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\": 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/81b4aecce0e32e6327ef/measurements.json b/records/sd1/sweep/81b4aecce0e32e6327ef/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..87f9e0207083f4268ce0f9584f40d5b4c6fc3281
--- /dev/null
+++ b/records/sd1/sweep/81b4aecce0e32e6327ef/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "81b4aecce0e32e6327ef",
+ "values": {
+  "cosine_distance": {
+   "value": 0.19114,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.232316,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.075789,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.042763,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.628,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00541,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.815,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9931,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.084,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81b4aecce0e32e6327ef/record.json b/records/sd1/sweep/81b4aecce0e32e6327ef/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..62fdaf6fa902df13ab00a3c09ea589439c2ebb15
--- /dev/null
+++ b/records/sd1/sweep/81b4aecce0e32e6327ef/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "81b4aecce0e32e6327ef",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:56Z",
+ "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": "middle_block.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1ee06e7cf667818cef96609796ff52c2c0e4013938875bc60e97930630aee926.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1ee06e7cf667818cef96609796ff52c2c0e4013938875bc60e97930630aee926",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.SiLU",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "49c2c833d6acc6d485bc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81b4aecce0e32e6327ef/workflow.json b/records/sd1/sweep/81b4aecce0e32e6327ef/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5dbcb92c612e7ed1ff8505688d2fbf40d3e84933
--- /dev/null
+++ b/records/sd1/sweep/81b4aecce0e32e6327ef/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 81b4aecce0e32e6327ef)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81be3a4c65a05bdf7842/measurements.json b/records/sd1/sweep/81be3a4c65a05bdf7842/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5c98263c3a79c7caa191fcb51a956a26dc12869
--- /dev/null
+++ b/records/sd1/sweep/81be3a4c65a05bdf7842/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "81be3a4c65a05bdf7842",
+ "values": {
+  "cosine_distance": {
+   "value": 0.102183,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.113101,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.097635,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.157112,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.091,
+   "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": 17.251,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.954,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2899,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81be3a4c65a05bdf7842/record.json b/records/sd1/sweep/81be3a4c65a05bdf7842/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..04e55e827fb811368c13cb2b202e59ea6d06f452
--- /dev/null
+++ b/records/sd1/sweep/81be3a4c65a05bdf7842/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "81be3a4c65a05bdf7842",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:46Z",
+ "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": "output_blocks.8.1.transformer_blocks.0.attn2.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Dropout",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/1486a4d875c9274d819663c0ebaf1f83bffdb23fab7472a6c51f114bd5361f96.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1486a4d875c9274d819663c0ebaf1f83bffdb23fab7472a6c51f114bd5361f96",
+  "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": "de9c1ee4ffc6591a7b47"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81be3a4c65a05bdf7842/workflow.json b/records/sd1/sweep/81be3a4c65a05bdf7842/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a9d84c87af547e86e626047a12de38b5b01c981
--- /dev/null
+++ b/records/sd1/sweep/81be3a4c65a05bdf7842/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 81be3a4c65a05bdf7842)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.transformer_blocks.0.attn2.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81be82838f29e2ed5b07/measurements.json b/records/sd1/sweep/81be82838f29e2ed5b07/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..22aba3d270869b3c64bcde1d06b841bfdd1e64f2
--- /dev/null
+++ b/records/sd1/sweep/81be82838f29e2ed5b07/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "81be82838f29e2ed5b07",
+ "values": {
+  "std": {
+   "value": 36.015,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01484,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.983,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.097938,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.306534,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.3456,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.144177,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0294,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.016,
+   "method": "CLIP zero-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/81be82838f29e2ed5b07/record.json b/records/sd1/sweep/81be82838f29e2ed5b07/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..270f6a7f2a97b53cae80deac4ccab38e4be797d1
--- /dev/null
+++ b/records/sd1/sweep/81be82838f29e2ed5b07/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "81be82838f29e2ed5b07",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:11:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a 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": "de1eccace874429a61dc",
+  "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": "506a6f7318f5a50a4d1b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81be82838f29e2ed5b07/workflow.json b/records/sd1/sweep/81be82838f29e2ed5b07/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..11594a7cf8c32fb88b75345065e3dda8aeef8dff
--- /dev/null
+++ b/records/sd1/sweep/81be82838f29e2ed5b07/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 81be82838f29e2ed5b07)"
+  },
+  "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": 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/81c3246513f80ce3ed0f/measurements.json b/records/sd1/sweep/81c3246513f80ce3ed0f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..952e1aae53b249b31540bc11beb1481360635dc5
--- /dev/null
+++ b/records/sd1/sweep/81c3246513f80ce3ed0f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "81c3246513f80ce3ed0f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.407611,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.369563,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.345272,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.255224,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.161,
+   "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": 42.41,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9781,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5372,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81c3246513f80ce3ed0f/record.json b/records/sd1/sweep/81c3246513f80ce3ed0f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..31e080e3566acf2f094935cfb593420fbf6b5a33
--- /dev/null
+++ b/records/sd1/sweep/81c3246513f80ce3ed0f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "81c3246513f80ce3ed0f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:37Z",
+ "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": "input_blocks.4.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/92abace5a70dab219df9a6ead97eb42d4b123785c02c9560f95d616c01cb3259.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "92abace5a70dab219df9a6ead97eb42d4b123785c02c9560f95d616c01cb3259",
+  "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": "94ea4dcb55f96a4edeff"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81c3246513f80ce3ed0f/workflow.json b/records/sd1/sweep/81c3246513f80ce3ed0f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..26edaf35914160534bde61d9c4700fe31e048032
--- /dev/null
+++ b/records/sd1/sweep/81c3246513f80ce3ed0f/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 81c3246513f80ce3ed0f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    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\": -0.5\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/81d82f007058affa8d5f/measurements.json b/records/sd1/sweep/81d82f007058affa8d5f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..32dbb90a2ded80b6986522603577568ed6250bf5
--- /dev/null
+++ b/records/sd1/sweep/81d82f007058affa8d5f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "81d82f007058affa8d5f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.327999,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.358594,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.222039,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.122723,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.329,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00413,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.027,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0288,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1322,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81d82f007058affa8d5f/record.json b/records/sd1/sweep/81d82f007058affa8d5f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..40c36e786f16395164fd44539998331f5dc60156
--- /dev/null
+++ b/records/sd1/sweep/81d82f007058affa8d5f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "81d82f007058affa8d5f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.10.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "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/95013dca4454761c93735e44aedea3eb6cadced6d320b1bf049da19d02ecf76c.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "95013dca4454761c93735e44aedea3eb6cadced6d320b1bf049da19d02ecf76c",
+  "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": "0657f0a04b587013300f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81d82f007058affa8d5f/workflow.json b/records/sd1/sweep/81d82f007058affa8d5f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6be22d3afde52203c5dba491d228391e233dbe4f
--- /dev/null
+++ b/records/sd1/sweep/81d82f007058affa8d5f/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 81d82f007058affa8d5f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/81d9790f3c0d1eb38eac/interpretations.jsonl b/records/sd1/sweep/81d9790f3c0d1eb38eac/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..7325cbd994fb9b74e8bf01d0005d67f62989c321
--- /dev/null
+++ b/records/sd1/sweep/81d9790f3c0d1eb38eac/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "81d9790f3c0d1eb38eac"}, "kind": "caption", "value": "Extreme close-up of an old man's deeply wrinkled face and long white beard", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:53:17Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e15c4a736c3c2469"}
+{"target": {"record": "81d9790f3c0d1eb38eac"}, "kind": "change", "value": "The camera zooms into a tight, hyper-detailed face crop. The hat and sea are gone, wrinkles and skin pores are exaggerated, and the background turns soft blue-grey.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:53:17Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "61c4bdedf8fd8cf6"}
+{"target": {"record": "81d9790f3c0d1eb38eac"}, "kind": "keywords", "value": ["extreme close-up", "wrinkled skin", "detailed face", "white beard", "zoomed portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:53:17Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0e3a7b1d12e915c3"}
+{"target": {"record": "81d9790f3c0d1eb38eac"}, "kind": "effect_tags", "value": ["zoom", "fine_detail", "texture", "new_composition", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:53:17Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "81c6a60b337ae47c"}
diff --git a/records/sd1/sweep/81d9790f3c0d1eb38eac/measurements.json b/records/sd1/sweep/81d9790f3c0d1eb38eac/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..55465aeedf9dfec6e4788fa0351afe8a8a32f82a
--- /dev/null
+++ b/records/sd1/sweep/81d9790f3c0d1eb38eac/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "81d9790f3c0d1eb38eac",
+ "values": {
+  "std": {
+   "value": 37.672,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02028,
+   "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": 33.224,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.445992,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.565969,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.187436,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.131752,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9475,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6097,
+   "method": "CLIP zero-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/81d9790f3c0d1eb38eac/record.json b/records/sd1/sweep/81d9790f3c0d1eb38eac/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fda65127681b6dcec7e9e91a829237702843201d
--- /dev/null
+++ b/records/sd1/sweep/81d9790f3c0d1eb38eac/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "81d9790f3c0d1eb38eac",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:26:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "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/81d9790f3c0d1eb38eac/workflow.json b/records/sd1/sweep/81d9790f3c0d1eb38eac/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..baf9165a30bd926623bd40fde31f540a79d33c07
--- /dev/null
+++ b/records/sd1/sweep/81d9790f3c0d1eb38eac/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 81d9790f3c0d1eb38eac)"
+  },
+  "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": "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/81e5d1fd0f5e44b8d299/measurements.json b/records/sd1/sweep/81e5d1fd0f5e44b8d299/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f79e126688be3cd60c8a3b16b5d2beb21c1f0e74
--- /dev/null
+++ b/records/sd1/sweep/81e5d1fd0f5e44b8d299/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "81e5d1fd0f5e44b8d299",
+ "values": {
+  "cosine_distance": {
+   "value": 0.166342,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.230617,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.117567,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036317,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.315,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00735,
+   "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": 23.418,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9637,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1382,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81e5d1fd0f5e44b8d299/record.json b/records/sd1/sweep/81e5d1fd0f5e44b8d299/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..313400f290c40790afe45c90fc00d0a138a833b5
--- /dev/null
+++ b/records/sd1/sweep/81e5d1fd0f5e44b8d299/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "81e5d1fd0f5e44b8d299",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:27Z",
+ "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.2.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.hi",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/cf9002f04557325199e8376766006562916bae0b0e41e042a6f881a95a904bcd.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "cf9002f04557325199e8376766006562916bae0b0e41e042a6f881a95a904bcd",
+  "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": "8aa7670ad0372a936e8e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81e5d1fd0f5e44b8d299/workflow.json b/records/sd1/sweep/81e5d1fd0f5e44b8d299/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..738378d418745b9673570d985aca683f5c120a40
--- /dev/null
+++ b/records/sd1/sweep/81e5d1fd0f5e44b8d299/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 81e5d1fd0f5e44b8d299)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.transformer_blocks.0.attn1.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/81e808c054d05a34f894/measurements.json b/records/sd1/sweep/81e808c054d05a34f894/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5cb9a9b8328dc80f6a904371eca603e3b06a4fb
--- /dev/null
+++ b/records/sd1/sweep/81e808c054d05a34f894/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "81e808c054d05a34f894",
+ "values": {
+  "cosine_distance": {
+   "value": 0.098217,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.116972,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.173395,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.089769,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.811,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00543,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.882,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9766,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1696,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81e808c054d05a34f894/record.json b/records/sd1/sweep/81e808c054d05a34f894/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5aa388421947ff55ef67b6432b22dae9827b273c
--- /dev/null
+++ b/records/sd1/sweep/81e808c054d05a34f894/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "81e808c054d05a34f894",
+ "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.10.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/bf82bf02dcdcd4cff5ffabe437adf12c7bf8abebdaf41a0ad97ecc1ee5216569.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "bf82bf02dcdcd4cff5ffabe437adf12c7bf8abebdaf41a0ad97ecc1ee5216569",
+  "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": "4f36db1fb32571769ec6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/81e808c054d05a34f894/workflow.json b/records/sd1/sweep/81e808c054d05a34f894/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f80ff26ca14d633e9f7c2fe56959b11fb260feba
--- /dev/null
+++ b/records/sd1/sweep/81e808c054d05a34f894/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 81e808c054d05a34f894)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/82006f5aaa8936c4e395/measurements.json b/records/sd1/sweep/82006f5aaa8936c4e395/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..28c6c1ed20fde934bb9124170b053e5bdcb93078
--- /dev/null
+++ b/records/sd1/sweep/82006f5aaa8936c4e395/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82006f5aaa8936c4e395",
+ "values": {
+  "cosine_distance": {
+   "value": 0.137673,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.183903,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.096109,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.047918,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.223,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00483,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.472,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0574,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1026,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82006f5aaa8936c4e395/record.json b/records/sd1/sweep/82006f5aaa8936c4e395/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3207e6da194ab61b9114551a7ac27d9128cc2883
--- /dev/null
+++ b/records/sd1/sweep/82006f5aaa8936c4e395/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82006f5aaa8936c4e395",
+ "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_out",
+   "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/9704b518053b7544ee360f73bf1040d5c53bc6a3a074967a1999ae6f4009bfb4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9704b518053b7544ee360f73bf1040d5c53bc6a3a074967a1999ae6f4009bfb4",
+  "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": "8ac4949f56429b224ff8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82006f5aaa8936c4e395/workflow.json b/records/sd1/sweep/82006f5aaa8936c4e395/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..86f5f961b8c9eccae77e43e38f22278254701a0c
--- /dev/null
+++ b/records/sd1/sweep/82006f5aaa8936c4e395/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82006f5aaa8936c4e395)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.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/8206758aad082b71b0ac/measurements.json b/records/sd1/sweep/8206758aad082b71b0ac/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca00169309a86782e447be3b03f3a77a621cb10e
--- /dev/null
+++ b/records/sd1/sweep/8206758aad082b71b0ac/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8206758aad082b71b0ac",
+ "values": {
+  "std": {
+   "value": 35.201,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02452,
+   "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": 30.676,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.318194,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.524038,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.409445,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.121941,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9966,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4406,
+   "method": "CLIP zero-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/8206758aad082b71b0ac/record.json b/records/sd1/sweep/8206758aad082b71b0ac/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..069aa7090cc30126bbe7bcd8afd2334996f1d64f
--- /dev/null
+++ b/records/sd1/sweep/8206758aad082b71b0ac/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8206758aad082b71b0ac",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:20:51Z",
+ "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 city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "zero",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "multiply",
+   0.0,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "58f513e6d7d2e4d2cb8e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8206758aad082b71b0ac/workflow.json b/records/sd1/sweep/8206758aad082b71b0ac/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9bdb9c2c4690cf47510484caefc638cb44b347a8
--- /dev/null
+++ b/records/sd1/sweep/8206758aad082b71b0ac/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8206758aad082b71b0ac)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/820ac84d3548a365820f/measurements.json b/records/sd1/sweep/820ac84d3548a365820f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ead25f9164663d5b239fb0cb4e2aef205acce48
--- /dev/null
+++ b/records/sd1/sweep/820ac84d3548a365820f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "820ac84d3548a365820f",
+ "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/820ac84d3548a365820f/record.json b/records/sd1/sweep/820ac84d3548a365820f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..34ce2268dcd5b78089de4e4fec48189c1eef74fb
--- /dev/null
+++ b/records/sd1/sweep/820ac84d3548a365820f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "820ac84d3548a365820f",
+ "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": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/6cb1afc6f7eb96296a7fa6ff78ab8ea10b474d37d505252c579884aed4dc8027.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6cb1afc6f7eb96296a7fa6ff78ab8ea10b474d37d505252c579884aed4dc8027",
+  "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": "a3658848642c176b997c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/820ac84d3548a365820f/workflow.json b/records/sd1/sweep/820ac84d3548a365820f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e787e2d63563ee3d3f15e2e8b7bdea681091c55
--- /dev/null
+++ b/records/sd1/sweep/820ac84d3548a365820f/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 820ac84d3548a365820f)"
+  },
+  "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\": 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/822164effd127b7578e9/measurements.json b/records/sd1/sweep/822164effd127b7578e9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2bbb924461f6cd44f39e394d72579d18aef45179
--- /dev/null
+++ b/records/sd1/sweep/822164effd127b7578e9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "822164effd127b7578e9",
+ "values": {
+  "std": {
+   "value": 59.214,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00986,
+   "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": 4.595,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.005653,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.014093,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.007142,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004047,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9902,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3243,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/822164effd127b7578e9/record.json b/records/sd1/sweep/822164effd127b7578e9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdf9779e0746dbbbb8d2c5025bbbc47e852fb67b
--- /dev/null
+++ b/records/sd1/sweep/822164effd127b7578e9/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "822164effd127b7578e9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:39:10Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "in.lo",
+   "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": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "c4b2c06862759b581c14"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/822164effd127b7578e9/workflow.json b/records/sd1/sweep/822164effd127b7578e9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f09fe754a2916b3f6db028bac385e914a322fe6
--- /dev/null
+++ b/records/sd1/sweep/822164effd127b7578e9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 822164effd127b7578e9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 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/82229ed26a444356bb7f/measurements.json b/records/sd1/sweep/82229ed26a444356bb7f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1677265efb68e373c7080870bc139eb3e0f2c4ed
--- /dev/null
+++ b/records/sd1/sweep/82229ed26a444356bb7f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82229ed26a444356bb7f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.074289,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.050597,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.037632,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.057984,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.992,
+   "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": 11.751,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9673,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2415,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82229ed26a444356bb7f/record.json b/records/sd1/sweep/82229ed26a444356bb7f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..41fb91269453a0edd07b0b69c4592391af9e645d
--- /dev/null
+++ b/records/sd1/sweep/82229ed26a444356bb7f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "82229ed26a444356bb7f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:59Z",
+ "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": "input_blocks.11.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/50e48dc860e7dc7589a35e655da004c2030d813e1d2c7847cc0c4e83c0591745.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "50e48dc860e7dc7589a35e655da004c2030d813e1d2c7847cc0c4e83c0591745",
+  "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": "1b4027e8dbe4a19d37da"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82229ed26a444356bb7f/workflow.json b/records/sd1/sweep/82229ed26a444356bb7f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c5aafe890b3815c2a62f27d356f8fae57c81be4e
--- /dev/null
+++ b/records/sd1/sweep/82229ed26a444356bb7f/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 82229ed26a444356bb7f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/8224a7d1162278df4f8a/measurements.json b/records/sd1/sweep/8224a7d1162278df4f8a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bce2d82e046849da714846fc5a1541d5548b6c97
--- /dev/null
+++ b/records/sd1/sweep/8224a7d1162278df4f8a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8224a7d1162278df4f8a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.167714,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.2045,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.122628,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.126874,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.472,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00312,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.084,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9848,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3409,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8224a7d1162278df4f8a/record.json b/records/sd1/sweep/8224a7d1162278df4f8a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..076a8a1077450736dfffde0a35d91fc0d5568fe2
--- /dev/null
+++ b/records/sd1/sweep/8224a7d1162278df4f8a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8224a7d1162278df4f8a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:28Z",
+ "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": "output_blocks.10.1.norm",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/05be0d8a24d2e54fd841508f04ecd6efbac6f732402dffe99e544974d0459046.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "05be0d8a24d2e54fd841508f04ecd6efbac6f732402dffe99e544974d0459046",
+  "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": "1f10a7ebaf711e9a28fe"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8224a7d1162278df4f8a/workflow.json b/records/sd1/sweep/8224a7d1162278df4f8a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b8ad8eb0f2db291465543431ac24425fcbee47af
--- /dev/null
+++ b/records/sd1/sweep/8224a7d1162278df4f8a/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 8224a7d1162278df4f8a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.norm\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/822c1d6ea98d41b06569/measurements.json b/records/sd1/sweep/822c1d6ea98d41b06569/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aab6a2af6b35e57eb2537b5fd307ff883d763004
--- /dev/null
+++ b/records/sd1/sweep/822c1d6ea98d41b06569/measurements.json
@@ -0,0 +1,59 @@
+{
+ "record": "822c1d6ea98d41b06569",
+ "values": {
+  "std": {
+   "value": 91.904,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00298,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "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": 80.851,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.717803,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.83177,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.840369,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.298957,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7937,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9524,
+   "method": "CLIP zero-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/822c1d6ea98d41b06569/record.json b/records/sd1/sweep/822c1d6ea98d41b06569/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..25715f165f0707a978534863b57fb0dce12fc877
--- /dev/null
+++ b/records/sd1/sweep/822c1d6ea98d41b06569/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "822c1d6ea98d41b06569",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:32:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "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": "81ef55543b3b3c080780",
+  "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": "8b15fda677257eb8aa1c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/822c1d6ea98d41b06569/workflow.json b/records/sd1/sweep/822c1d6ea98d41b06569/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..47db56e26582cc88c7b5400db176735fb51ed677
--- /dev/null
+++ b/records/sd1/sweep/822c1d6ea98d41b06569/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 822c1d6ea98d41b06569)"
+  },
+  "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 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/82307206ad4854a26778/interpretations.jsonl b/records/sd1/sweep/82307206ad4854a26778/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..34e7f1dd08e8cd084a585f103044e105735bbdf5
--- /dev/null
+++ b/records/sd1/sweep/82307206ad4854a26778/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "82307206ad4854a26778"}, "kind": "caption", "value": "Very blurry, faded sepia portrait of a woman with long wavy hair", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "534b632b0640df8c"}
+{"target": {"record": "82307206ad4854a26778"}, "kind": "change", "value": "The portrait becomes strongly blurred and washed into warm sepia, with a ghostly face.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "508fce37706cd35f"}
+{"target": {"record": "82307206ad4854a26778"}, "kind": "keywords", "value": ["blur", "sepia", "faded portrait", "ghostly"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2364e0fed01db22e"}
+{"target": {"record": "82307206ad4854a26778"}, "kind": "effect_tags", "value": ["blur", "washed_out", "colour_shift", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:49:02Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2017038ab1ee8c87"}
diff --git a/records/sd1/sweep/82307206ad4854a26778/measurements.json b/records/sd1/sweep/82307206ad4854a26778/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..442b65f192a55ea9d735e3529f574b2396c98c85
--- /dev/null
+++ b/records/sd1/sweep/82307206ad4854a26778/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82307206ad4854a26778",
+ "values": {
+  "cosine_distance": {
+   "value": 0.20568,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.564548,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.231035,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.171265,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.259,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00145,
+   "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": 23.156,
+   "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": 0.7835,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82307206ad4854a26778/record.json b/records/sd1/sweep/82307206ad4854a26778/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9da40da67c469cf50e3a7c8244f10bf5853cff91
--- /dev/null
+++ b/records/sd1/sweep/82307206ad4854a26778/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82307206ad4854a26778",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/f209ce8a861aaad93feda597406366615dd09b789888708e659f0555eef9f890.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "f209ce8a861aaad93feda597406366615dd09b789888708e659f0555eef9f890",
+  "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": "820cb606038fb482e77b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82307206ad4854a26778/workflow.json b/records/sd1/sweep/82307206ad4854a26778/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..daae8ed39acaff64bbbbfe6d856952b178a9aa3e
--- /dev/null
+++ b/records/sd1/sweep/82307206ad4854a26778/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82307206ad4854a26778)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/823ab61838a8fc4abfd6/measurements.json b/records/sd1/sweep/823ab61838a8fc4abfd6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a8ba41fdef6a367a2b659fd59b4cc2e7c736970
--- /dev/null
+++ b/records/sd1/sweep/823ab61838a8fc4abfd6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "823ab61838a8fc4abfd6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.144357,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.321777,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.229235,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.119818,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.239,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00353,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.26,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0323,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0423,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/823ab61838a8fc4abfd6/record.json b/records/sd1/sweep/823ab61838a8fc4abfd6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b98f854fdf9f5a86ca4994705c0c45c17fb8f908
--- /dev/null
+++ b/records/sd1/sweep/823ab61838a8fc4abfd6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "823ab61838a8fc4abfd6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:40Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d4e2e91c0b1f8c3b3367280430477b04c00268b686f0f7513e41c591a2fdf6b5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d4e2e91c0b1f8c3b3367280430477b04c00268b686f0f7513e41c591a2fdf6b5",
+  "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": "52a2d66d84098a5f107d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/823ab61838a8fc4abfd6/workflow.json b/records/sd1/sweep/823ab61838a8fc4abfd6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b276509db1736c725f664b21c15386900f55d9b9
--- /dev/null
+++ b/records/sd1/sweep/823ab61838a8fc4abfd6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 823ab61838a8fc4abfd6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/82418ded2c5ba9ea8733/measurements.json b/records/sd1/sweep/82418ded2c5ba9ea8733/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..37362cde0103da206c038a482000b61586010747
--- /dev/null
+++ b/records/sd1/sweep/82418ded2c5ba9ea8733/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82418ded2c5ba9ea8733",
+ "values": {
+  "cosine_distance": {
+   "value": 0.402133,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.806457,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.785433,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.36278,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 39.063,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00619,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 39.268,
+   "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": 0.1712,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82418ded2c5ba9ea8733/record.json b/records/sd1/sweep/82418ded2c5ba9ea8733/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b94c51067bd67bd1a63c077bf0995412e581ec7
--- /dev/null
+++ b/records/sd1/sweep/82418ded2c5ba9ea8733/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "82418ded2c5ba9ea8733",
+ "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.2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "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/5bcec65b71de10a40eb1b8a2376d42272feb7fb7bc2562fb7095bf774d4a0311.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "5bcec65b71de10a40eb1b8a2376d42272feb7fb7bc2562fb7095bf774d4a0311",
+  "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": "d4ddf40f2e339bb2cd4e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82418ded2c5ba9ea8733/workflow.json b/records/sd1/sweep/82418ded2c5ba9ea8733/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e673610f968b5a7dd6dcc4bdd7bb64303ddeef9f
--- /dev/null
+++ b/records/sd1/sweep/82418ded2c5ba9ea8733/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 82418ded2c5ba9ea8733)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/8243604a11085514a003/measurements.json b/records/sd1/sweep/8243604a11085514a003/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..49d8abcfc1b9162962242fa2fb906a53dabd22e1
--- /dev/null
+++ b/records/sd1/sweep/8243604a11085514a003/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8243604a11085514a003",
+ "values": {
+  "cosine_distance": {
+   "value": 0.008407,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.019406,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004515,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003929,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.029,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00578,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.066,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9848,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1476,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8243604a11085514a003/record.json b/records/sd1/sweep/8243604a11085514a003/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ef0d128fad861b35d26a8eb75b9ab24d0c9fa58
--- /dev/null
+++ b/records/sd1/sweep/8243604a11085514a003/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8243604a11085514a003",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:03Z",
+ "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.2.0.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/689b9ec61d8c8958e0e13cc78cdd0d076a4619daa4cfe16c8e30701899671aa2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "689b9ec61d8c8958e0e13cc78cdd0d076a4619daa4cfe16c8e30701899671aa2",
+  "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": "1344013ae476b48b0429"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8243604a11085514a003/workflow.json b/records/sd1/sweep/8243604a11085514a003/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..33ad7e70ffb11909f9ab117a9285d43ddafd3e35
--- /dev/null
+++ b/records/sd1/sweep/8243604a11085514a003/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8243604a11085514a003)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.out_layers.3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8248971776c6153ebbb4/measurements.json b/records/sd1/sweep/8248971776c6153ebbb4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3b3361bbec49230e6ca95b8bbab0c1f39b4c1752
--- /dev/null
+++ b/records/sd1/sweep/8248971776c6153ebbb4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8248971776c6153ebbb4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.403982,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.466967,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.229824,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.18568,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.785,
+   "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": 47.653,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9529,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2276,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8248971776c6153ebbb4/record.json b/records/sd1/sweep/8248971776c6153ebbb4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b063b5dbb67fd0b3449cfc363bbbb815999f50be
--- /dev/null
+++ b/records/sd1/sweep/8248971776c6153ebbb4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8248971776c6153ebbb4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.proj_in",
+   "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/9852d57e6cce568d682e0243921c84f7a3435490cb7b8ef55437f04569a38556.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9852d57e6cce568d682e0243921c84f7a3435490cb7b8ef55437f04569a38556",
+  "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": "7bf0f5eeae8e7a188f05"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8248971776c6153ebbb4/workflow.json b/records/sd1/sweep/8248971776c6153ebbb4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f892bffa48d6aa3c82f80837f0e6fa40e4041a6
--- /dev/null
+++ b/records/sd1/sweep/8248971776c6153ebbb4/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 8248971776c6153ebbb4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.proj_in\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/825188806eedaab3bed7/interpretations.jsonl b/records/sd1/sweep/825188806eedaab3bed7/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a3ef2e5f8bab087d8e9fdf3f7fc8951c6322ab74
--- /dev/null
+++ b/records/sd1/sweep/825188806eedaab3bed7/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "825188806eedaab3bed7"}, "kind": "caption", "value": "Young man with slicked-back dark hair in profile-leaning pose against plain grey backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "51993b1d30b874f4"}
+{"target": {"record": "825188806eedaab3bed7"}, "kind": "change", "value": "The woman becomes a young man with slicked-back hair and a warmer, ruddier skin tone, while the plain backdrop and dark top stay the same.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cf391a8dc9ce297b"}
+{"target": {"record": "825188806eedaab3bed7"}, "kind": "keywords", "value": ["gender swap", "male portrait", "slicked hair", "warm skin", "identity change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f1e9b51142776dd7"}
+{"target": {"record": "825188806eedaab3bed7"}, "kind": "effect_tags", "value": ["subject_kept", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7772ddc1c4a938ae"}
diff --git a/records/sd1/sweep/825188806eedaab3bed7/measurements.json b/records/sd1/sweep/825188806eedaab3bed7/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0b284c4bea0fd085880c39025a5adba25ac732cd
--- /dev/null
+++ b/records/sd1/sweep/825188806eedaab3bed7/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "825188806eedaab3bed7",
+ "values": {
+  "cosine_distance": {
+   "value": 0.173531,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.19989,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.165522,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.258542,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.653,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00339,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.956,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.1797,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/825188806eedaab3bed7/record.json b/records/sd1/sweep/825188806eedaab3bed7/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..97a517b486d3f95a3ba7a5b928ed29989b209053
--- /dev/null
+++ b/records/sd1/sweep/825188806eedaab3bed7/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "825188806eedaab3bed7",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:16Z",
+ "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.6.1.norm",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.mid",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f917f52aea17b2a0d0e97e7f9be0e43c30228323cfc1e21cc3b9b3225978c0e3.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f917f52aea17b2a0d0e97e7f9be0e43c30228323cfc1e21cc3b9b3225978c0e3",
+  "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": "216c4d0dc822a19c82fa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/825188806eedaab3bed7/workflow.json b/records/sd1/sweep/825188806eedaab3bed7/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..77048a8ccb5c7c6c367ae6a477447141f672b69c
--- /dev/null
+++ b/records/sd1/sweep/825188806eedaab3bed7/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 825188806eedaab3bed7)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.norm\",\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/82520d884b33a0177453/measurements.json b/records/sd1/sweep/82520d884b33a0177453/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa3a206d991596560853f82424a521282cb44b05
--- /dev/null
+++ b/records/sd1/sweep/82520d884b33a0177453/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82520d884b33a0177453",
+ "values": {
+  "cosine_distance": {
+   "value": 0.364107,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.414544,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.198171,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.097857,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.037,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00493,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.72,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0416,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2394,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82520d884b33a0177453/record.json b/records/sd1/sweep/82520d884b33a0177453/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..258813e075b6c4b987aae3ae0c22ad6281263ed2
--- /dev/null
+++ b/records/sd1/sweep/82520d884b33a0177453/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "82520d884b33a0177453",
+ "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.in_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "in.lo",
+   "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/07ac78b780e098c9ec571e3b4bb341996f7a8ab7e37d6cba6c5eb99b437b422f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "07ac78b780e098c9ec571e3b4bb341996f7a8ab7e37d6cba6c5eb99b437b422f",
+  "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": "88ebcb35b7cb07b6cb48"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82520d884b33a0177453/workflow.json b/records/sd1/sweep/82520d884b33a0177453/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2afcab4e595ac1ea90b69f95a5b73563ac4fc5b
--- /dev/null
+++ b/records/sd1/sweep/82520d884b33a0177453/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 82520d884b33a0177453)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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/825bc50caaf1b1a8c0bb/measurements.json b/records/sd1/sweep/825bc50caaf1b1a8c0bb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c29d2f367bfd86988da3a52fb3d44ac26fdb282
--- /dev/null
+++ b/records/sd1/sweep/825bc50caaf1b1a8c0bb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "825bc50caaf1b1a8c0bb",
+ "values": {
+  "std": {
+   "value": 57.359,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02155,
+   "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": 48.917,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.518071,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.609632,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.184299,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.061835,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.089,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0056,
+   "method": "CLIP zero-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/825bc50caaf1b1a8c0bb/record.json b/records/sd1/sweep/825bc50caaf1b1a8c0bb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cbf1204fb405e9e6bd0b773e24d8f2b444b74682
--- /dev/null
+++ b/records/sd1/sweep/825bc50caaf1b1a8c0bb/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "825bc50caaf1b1a8c0bb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:14:54Z",
+ "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": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "zero",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "acdaf444fe41d2bf7035"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/825bc50caaf1b1a8c0bb/workflow.json b/records/sd1/sweep/825bc50caaf1b1a8c0bb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b279cb893831c044f8f8f5b5230566392a019830
--- /dev/null
+++ b/records/sd1/sweep/825bc50caaf1b1a8c0bb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 825bc50caaf1b1a8c0bb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/827e472ce9b523759b9d/measurements.json b/records/sd1/sweep/827e472ce9b523759b9d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d714da47eb1907a7af093aaff26ab548e9f4227b
--- /dev/null
+++ b/records/sd1/sweep/827e472ce9b523759b9d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "827e472ce9b523759b9d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.559884,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.620548,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.255202,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.129992,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 26.399,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00276,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.979,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.014,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2525,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/827e472ce9b523759b9d/record.json b/records/sd1/sweep/827e472ce9b523759b9d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5fc1abcfdbcdf581fa7c8bcc195077df20409573
--- /dev/null
+++ b/records/sd1/sweep/827e472ce9b523759b9d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "827e472ce9b523759b9d",
+ "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.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "embed",
+   "kind": "time_embed",
+   "module_type": "Linear",
+   "container": "Sequential",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/a41e3ab3ecf9aeff5633b00b6ca282bca728452b10618ebc26edd8ed6c1eef70.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a41e3ab3ecf9aeff5633b00b6ca282bca728452b10618ebc26edd8ed6c1eef70",
+  "type_path": "UNetModel.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "e145516fed23269a4dbd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/827e472ce9b523759b9d/workflow.json b/records/sd1/sweep/827e472ce9b523759b9d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..173980bced7f5bc3bb79ccd4623840a982e0ffe6
--- /dev/null
+++ b/records/sd1/sweep/827e472ce9b523759b9d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 827e472ce9b523759b9d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"time_embed.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/827f07c664d7ada24a7d/measurements.json b/records/sd1/sweep/827f07c664d7ada24a7d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d339f9da979323e47a2df2a71c7d2535ca2184ca
--- /dev/null
+++ b/records/sd1/sweep/827f07c664d7ada24a7d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "827f07c664d7ada24a7d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.126068,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.175567,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.03766,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.017917,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.908,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00755,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 18.542,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9949,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1031,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/827f07c664d7ada24a7d/record.json b/records/sd1/sweep/827f07c664d7ada24a7d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..61bf604a26679ef40123952b5796c9c783289c2c
--- /dev/null
+++ b/records/sd1/sweep/827f07c664d7ada24a7d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "827f07c664d7ada24a7d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.1.1.transformer_blocks.0.ff.net.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "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/54e52e0bf0aeb9d65bef9b2a27a1bb30721e46e1ba05adc0185d70873bc365d5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "54e52e0bf0aeb9d65bef9b2a27a1bb30721e46e1ba05adc0185d70873bc365d5",
+  "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": "5f8d47370c9ecb010322"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/827f07c664d7ada24a7d/workflow.json b/records/sd1/sweep/827f07c664d7ada24a7d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bd5516562a105119f0970bfa064c5787f1675ec
--- /dev/null
+++ b/records/sd1/sweep/827f07c664d7ada24a7d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 827f07c664d7ada24a7d)"
+  },
+  "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\": 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/8289ea73a301a90ec7c1/measurements.json b/records/sd1/sweep/8289ea73a301a90ec7c1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7b2a02c231c9c1ccdb93129ebc6106267285b0c
--- /dev/null
+++ b/records/sd1/sweep/8289ea73a301a90ec7c1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8289ea73a301a90ec7c1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.56705,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.645265,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.50395,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.325718,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 35.111,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00205,
+   "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": 59.735,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9763,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2757,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8289ea73a301a90ec7c1/record.json b/records/sd1/sweep/8289ea73a301a90ec7c1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8cb7b7c7f9e3b6a2421ff045b7f3f6182ba41a50
--- /dev/null
+++ b/records/sd1/sweep/8289ea73a301a90ec7c1/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8289ea73a301a90ec7c1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.in_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e8e318ddb924f7b29c42bbd4a4b307c383ef4451677b20e25c0b4da4a337c016.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e8e318ddb924f7b29c42bbd4a4b307c383ef4451677b20e25c0b4da4a337c016",
+  "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": "79e26a8f365425e00ff5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8289ea73a301a90ec7c1/workflow.json b/records/sd1/sweep/8289ea73a301a90ec7c1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..befb7efc58fe2c9d92084aff3cdce9a1b274daec
--- /dev/null
+++ b/records/sd1/sweep/8289ea73a301a90ec7c1/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 8289ea73a301a90ec7c1)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/828fd6558d77051bc5ba/measurements.json b/records/sd1/sweep/828fd6558d77051bc5ba/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d73a43e09286988f15c16e32394534714158feff
--- /dev/null
+++ b/records/sd1/sweep/828fd6558d77051bc5ba/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "828fd6558d77051bc5ba",
+ "values": {
+  "cosine_distance": {
+   "value": 0.402223,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.461472,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.514421,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.205359,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.036,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00492,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 46.644,
+   "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": 0.3659,
+   "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/828fd6558d77051bc5ba/record.json b/records/sd1/sweep/828fd6558d77051bc5ba/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca85a66308d5be96a4e6df006201e3caa41e0bb5
--- /dev/null
+++ b/records/sd1/sweep/828fd6558d77051bc5ba/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "828fd6558d77051bc5ba",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:37Z",
+ "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": "input_blocks.4.1.transformer_blocks.0.attn1.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.mid",
+   "kind": "self_attn",
+   "module_type": "Dropout",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/2aaf91cc0049069eddb71ddcdb0a3679d39b561a8dd4efa36b48ec6ce88ec32b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2aaf91cc0049069eddb71ddcdb0a3679d39b561a8dd4efa36b48ec6ce88ec32b",
+  "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": "24e1844afd0cf7426fd2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/828fd6558d77051bc5ba/workflow.json b/records/sd1/sweep/828fd6558d77051bc5ba/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f058b9efbcdc10ae714eab47099d362099750ce1
--- /dev/null
+++ b/records/sd1/sweep/828fd6558d77051bc5ba/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 828fd6558d77051bc5ba)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.attn1.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82971c07e2c20658e4a9/interpretations.jsonl b/records/sd1/sweep/82971c07e2c20658e4a9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..60d5702183ebaedfb3639fc94a791f76f1fbe093
--- /dev/null
+++ b/records/sd1/sweep/82971c07e2c20658e4a9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "82971c07e2c20658e4a9"}, "kind": "caption", "value": "Old bearded man in a knitted brimmed hat under a darker stormy sky", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d7de46fca85e4887"}
+{"target": {"record": "82971c07e2c20658e4a9"}, "kind": "change", "value": "Mostly the same portrait, but the hat becomes coarser knit, the face is reshaped a little and the sky gets darker and moodier.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "503e790a4c4e68b7"}
+{"target": {"record": "82971c07e2c20658e4a9"}, "kind": "keywords", "value": ["old man portrait", "knitted hat", "stormy sky", "moodier light", "minor variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7e175d8525e107bf"}
+{"target": {"record": "82971c07e2c20658e4a9"}, "kind": "effect_tags", "value": ["subtle", "subject_kept", "dark"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:52:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2f95701c4ca7ff64"}
diff --git a/records/sd1/sweep/82971c07e2c20658e4a9/measurements.json b/records/sd1/sweep/82971c07e2c20658e4a9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6ead24b5e74bef4652ff03e272a5229671741be
--- /dev/null
+++ b/records/sd1/sweep/82971c07e2c20658e4a9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82971c07e2c20658e4a9",
+ "values": {
+  "std": {
+   "value": 65.034,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01062,
+   "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": 20.981,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.111842,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.155664,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.024108,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.020132,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0048,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3391,
+   "method": "CLIP zero-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/82971c07e2c20658e4a9/record.json b/records/sd1/sweep/82971c07e2c20658e4a9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a14aa29e04659ea1e415c68c8c5bb58fce2e9aa1
--- /dev/null
+++ b/records/sd1/sweep/82971c07e2c20658e4a9/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "82971c07e2c20658e4a9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:48:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": 2.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "boost",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   2.0,
+   "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": "ce8e71e2d4752f99dab6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82971c07e2c20658e4a9/workflow.json b/records/sd1/sweep/82971c07e2c20658e4a9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9c1e80306922c214aabe582e5021de9c2b87b8e
--- /dev/null
+++ b/records/sd1/sweep/82971c07e2c20658e4a9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82971c07e2c20658e4a9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 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/82b119128685e8632147/measurements.json b/records/sd1/sweep/82b119128685e8632147/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1f651fa9cb3c16a42db0f12e9bb323234d19bfa
--- /dev/null
+++ b/records/sd1/sweep/82b119128685e8632147/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82b119128685e8632147",
+ "values": {
+  "cosine_distance": {
+   "value": 0.258803,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.303004,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.124872,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.073567,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.429,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00652,
+   "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": 29.913,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9463,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0495,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82b119128685e8632147/record.json b/records/sd1/sweep/82b119128685e8632147/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..adf8320d49c2055cd46405bc7edcf5bf7f108412
--- /dev/null
+++ b/records/sd1/sweep/82b119128685e8632147/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82b119128685e8632147",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/416667dd8689948389bd5e6c439e51d54238c800e342cff27b71a9a494cfe3e2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "416667dd8689948389bd5e6c439e51d54238c800e342cff27b71a9a494cfe3e2",
+  "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": "d27635eb353abdea30d0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82b119128685e8632147/workflow.json b/records/sd1/sweep/82b119128685e8632147/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b168358cd0cd36eb929a4d35163d5a036df408c1
--- /dev/null
+++ b/records/sd1/sweep/82b119128685e8632147/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82b119128685e8632147)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/82b95d2833b74ff0d834/interpretations.jsonl b/records/sd1/sweep/82b95d2833b74ff0d834/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2389403b587ec384b328ec01cb5a3f616a81a666
--- /dev/null
+++ b/records/sd1/sweep/82b95d2833b74ff0d834/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "82b95d2833b74ff0d834"}, "kind": "caption", "value": "Blurry grey abstract of vertical streaks resembling a window or curtain", "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": "87665793d7f9b285"}
+{"target": {"record": "82b95d2833b74ff0d834"}, "kind": "change", "value": "The portrait vanishes entirely, replaced by a desaturated, blurred abstract of grey vertical bands like a misty window frame.", "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": "f221119c38094322"}
+{"target": {"record": "82b95d2833b74ff0d834"}, "kind": "keywords", "value": ["abstract grey", "vertical streaks", "subject vanished", "window blur", "monochrome"], "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": "d897f698da4bd04f"}
+{"target": {"record": "82b95d2833b74ff0d834"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "desaturated", "blur", "new_composition", "degenerate"], "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": "bd4fee93629862f2"}
diff --git a/records/sd1/sweep/82b95d2833b74ff0d834/measurements.json b/records/sd1/sweep/82b95d2833b74ff0d834/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..49720f3202088d0b949124319e763b158025c4ec
--- /dev/null
+++ b/records/sd1/sweep/82b95d2833b74ff0d834/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82b95d2833b74ff0d834",
+ "values": {
+  "cosine_distance": {
+   "value": 0.586524,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.885661,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.771392,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.397391,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 27.54,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00368,
+   "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": 49.905,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0113,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4598,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82b95d2833b74ff0d834/record.json b/records/sd1/sweep/82b95d2833b74ff0d834/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a46b3125f862f34f4dc45b020b46b17a4046889
--- /dev/null
+++ b/records/sd1/sweep/82b95d2833b74ff0d834/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82b95d2833b74ff0d834",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "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/116c00c4a54a9cb5bb9bf2c0bb09190373ddd527e8760e8afc17dd2727f390b9.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "116c00c4a54a9cb5bb9bf2c0bb09190373ddd527e8760e8afc17dd2727f390b9",
+  "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": "560b6a5e4a1ea2dd5221"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82b95d2833b74ff0d834/workflow.json b/records/sd1/sweep/82b95d2833b74ff0d834/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c909e419bdc3a46a5a86ed5a0692f05cee31adfd
--- /dev/null
+++ b/records/sd1/sweep/82b95d2833b74ff0d834/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82b95d2833b74ff0d834)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/interpretations.jsonl b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..9db2aea7df1da29f79339d8b8d6ff1485626d2d7
--- /dev/null
+++ b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "82c0d8ebdd9a5f60dbdd"}, "kind": "caption", "value": "Young man with slicked dark hair and stern expression, head turned, grey backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "088b666ed0d27ca7"}
+{"target": {"record": "82c0d8ebdd9a5f60dbdd"}, "kind": "change", "value": "The subject becomes a stern young man with slicked hair and slightly warmer skin, while the composition and backdrop stay the same.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "316e3d8bfaa2d676"}
+{"target": {"record": "82c0d8ebdd9a5f60dbdd"}, "kind": "keywords", "value": ["male portrait", "slicked hair", "stern face", "identity change", "warm skin"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c1dcfe1b595db8fc"}
+{"target": {"record": "82c0d8ebdd9a5f60dbdd"}, "kind": "effect_tags", "value": ["subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:21:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "241af4d339e6cff8"}
diff --git a/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/measurements.json b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5834c7f19d710d49c8b8c57db0fd0fcab85432f3
--- /dev/null
+++ b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82c0d8ebdd9a5f60dbdd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.173754,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.199698,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.167562,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.255117,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.685,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00339,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.097,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.1994,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/record.json b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9f1553884bece3970b619111cdc76936f4a8a046
--- /dev/null
+++ b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "82c0d8ebdd9a5f60dbdd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:16Z",
+ "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.6.1.proj_in",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/1a612215eda6fa94a67d9cd764627a021417e36cae49fae9047874c5b548b279.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1a612215eda6fa94a67d9cd764627a021417e36cae49fae9047874c5b548b279",
+  "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": "b0d0cb0faa36c15fcd16"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/workflow.json b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a301f875dd8743853976f57dce5a550dd24fdce
--- /dev/null
+++ b/records/sd1/sweep/82c0d8ebdd9a5f60dbdd/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 82c0d8ebdd9a5f60dbdd)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.proj_in\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c36ab49bd36906d95a/measurements.json b/records/sd1/sweep/82c36ab49bd36906d95a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f660c55c1f30e0300eb0c057545d67ffc1d54806
--- /dev/null
+++ b/records/sd1/sweep/82c36ab49bd36906d95a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82c36ab49bd36906d95a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.331728,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.350454,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.162554,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.062252,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.785,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00573,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 35.812,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9617,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0824,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c36ab49bd36906d95a/record.json b/records/sd1/sweep/82c36ab49bd36906d95a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3f7a70eb8b61c7d09bc0f4b175b4400c13a279be
--- /dev/null
+++ b/records/sd1/sweep/82c36ab49bd36906d95a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82c36ab49bd36906d95a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:35Z",
+ "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.5.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5f707a9b07e5e1596020596cfc1a7019967b005e0df49c4c6eb32d5f0ed037f7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5f707a9b07e5e1596020596cfc1a7019967b005e0df49c4c6eb32d5f0ed037f7",
+  "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": "9f7859b42e0775ac94e0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c36ab49bd36906d95a/workflow.json b/records/sd1/sweep/82c36ab49bd36906d95a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d79e8b966f47573d65ded3757fc9f4824dc08174
--- /dev/null
+++ b/records/sd1/sweep/82c36ab49bd36906d95a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82c36ab49bd36906d95a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.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": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c6f124164ee260877b/interpretations.jsonl b/records/sd1/sweep/82c6f124164ee260877b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e1b172ca40ac4e5df059c54fdaa1885fce810ce0
--- /dev/null
+++ b/records/sd1/sweep/82c6f124164ee260877b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "82c6f124164ee260877b"}, "kind": "caption", "value": "Painted portrait of a woman with blue-green eyes in aged beige frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "70e2ca9fd11136da"}
+{"target": {"record": "82c6f124164ee260877b"}, "kind": "change", "value": "Almost no visible change aside from a slightly softer face, bluer eyes and a warmer aged border.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e5ae0832b033afb1"}
+{"target": {"record": "82c6f124164ee260877b"}, "kind": "keywords", "value": ["minor variation", "blue eyes", "aged border", "painted portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "929d82a40eb986f6"}
+{"target": {"record": "82c6f124164ee260877b"}, "kind": "effect_tags", "value": ["subject_kept", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:35Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "040b4d473197cd0e"}
diff --git a/records/sd1/sweep/82c6f124164ee260877b/measurements.json b/records/sd1/sweep/82c6f124164ee260877b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..00009e3936f71b7890fd810c581a97cc00930fec
--- /dev/null
+++ b/records/sd1/sweep/82c6f124164ee260877b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82c6f124164ee260877b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.084954,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.128479,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.054895,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023379,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.54,
+   "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": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.021,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9434,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1815,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c6f124164ee260877b/record.json b/records/sd1/sweep/82c6f124164ee260877b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..df4b0ac293e1cab9a0833373423bffa56863a8db
--- /dev/null
+++ b/records/sd1/sweep/82c6f124164ee260877b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82c6f124164ee260877b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:24Z",
+ "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": "input_blocks.8.0.out_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/2a4ed8f4d95a917fd5ef69165b8c4799a71381464c5477e2ea192f422d242a2f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2a4ed8f4d95a917fd5ef69165b8c4799a71381464c5477e2ea192f422d242a2f",
+  "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": "18f5ab23dec21c0a7625"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82c6f124164ee260877b/workflow.json b/records/sd1/sweep/82c6f124164ee260877b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..94bed6b52989da12192526225f35995932824881
--- /dev/null
+++ b/records/sd1/sweep/82c6f124164ee260877b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82c6f124164ee260877b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.out_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82cb2a6c622d454c86f3/measurements.json b/records/sd1/sweep/82cb2a6c622d454c86f3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..da10e60760a013664eba3480f631fd693005f37d
--- /dev/null
+++ b/records/sd1/sweep/82cb2a6c622d454c86f3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82cb2a6c622d454c86f3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.110738,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.120791,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.057761,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.080129,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.097,
+   "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": 14.142,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9352,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3908,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82cb2a6c622d454c86f3/record.json b/records/sd1/sweep/82cb2a6c622d454c86f3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c02013c0c2a658b6003764e57ea1f0d58b86d6d5
--- /dev/null
+++ b/records/sd1/sweep/82cb2a6c622d454c86f3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "82cb2a6c622d454c86f3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.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/523eace8cf0cf720699b54c7b9719962870ec40ba66e7b7effbf2ab1933648d8.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "523eace8cf0cf720699b54c7b9719962870ec40ba66e7b7effbf2ab1933648d8",
+  "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": "8497b01dc7f20770c372"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82cb2a6c622d454c86f3/workflow.json b/records/sd1/sweep/82cb2a6c622d454c86f3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b6c3cd4e30c873ecd3fe3d739baf9171620379a
--- /dev/null
+++ b/records/sd1/sweep/82cb2a6c622d454c86f3/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 82cb2a6c622d454c86f3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_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/82ceabf4b33b8e473634/measurements.json b/records/sd1/sweep/82ceabf4b33b8e473634/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb02b3f05c937ac458c8c4b5be464a6708a8fb41
--- /dev/null
+++ b/records/sd1/sweep/82ceabf4b33b8e473634/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82ceabf4b33b8e473634",
+ "values": {
+  "std": {
+   "value": 40.694,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01495,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.873,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.203588,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.311706,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.136355,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.04727,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0338,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0027,
+   "method": "CLIP zero-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/82ceabf4b33b8e473634/record.json b/records/sd1/sweep/82ceabf4b33b8e473634/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e5e8b4726b7bddde9ad6c414e26ba694138b2aa
--- /dev/null
+++ b/records/sd1/sweep/82ceabf4b33b8e473634/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "82ceabf4b33b8e473634",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:02:09Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 1234
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "heavy",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_noise",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82ceabf4b33b8e473634/workflow.json b/records/sd1/sweep/82ceabf4b33b8e473634/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..643da378a3f53f9fea8b50bc2f53f2862f72ffb6
--- /dev/null
+++ b/records/sd1/sweep/82ceabf4b33b8e473634/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82ceabf4b33b8e473634)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 1234\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82e008a5fee43f239d4c/measurements.json b/records/sd1/sweep/82e008a5fee43f239d4c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbdd8022f689338a204a01df678de97329528f20
--- /dev/null
+++ b/records/sd1/sweep/82e008a5fee43f239d4c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82e008a5fee43f239d4c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.049106,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.050923,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.01697,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010738,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.26,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00449,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.568,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9998,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.238,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82e008a5fee43f239d4c/record.json b/records/sd1/sweep/82e008a5fee43f239d4c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0eab16c97d5609c72368feea2e159ba863f6983
--- /dev/null
+++ b/records/sd1/sweep/82e008a5fee43f239d4c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82e008a5fee43f239d4c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.hi",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/2324be0f2667857c78d2d115f366b811dccf52e7c17fa7366f9725afea614395.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2324be0f2667857c78d2d115f366b811dccf52e7c17fa7366f9725afea614395",
+  "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": "bf6834106ebcb3ebc4a9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82e008a5fee43f239d4c/workflow.json b/records/sd1/sweep/82e008a5fee43f239d4c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..17005baa1516396be697a7116ef43e10efa8e3df
--- /dev/null
+++ b/records/sd1/sweep/82e008a5fee43f239d4c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82e008a5fee43f239d4c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.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/82eeff53066e8b2f476c/measurements.json b/records/sd1/sweep/82eeff53066e8b2f476c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..52615ce76a3f03d9fd6f03ade6053443337a4a9e
--- /dev/null
+++ b/records/sd1/sweep/82eeff53066e8b2f476c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82eeff53066e8b2f476c",
+ "values": {
+  "std": {
+   "value": 50.86,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01042,
+   "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": 2.476,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.0067,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.040865,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.012265,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.005298,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0167,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0156,
+   "method": "CLIP zero-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/82eeff53066e8b2f476c/record.json b/records/sd1/sweep/82eeff53066e8b2f476c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..244eecc1e0541a77a639836ad02eb83db39980ce
--- /dev/null
+++ b/records/sd1/sweep/82eeff53066e8b2f476c/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "82eeff53066e8b2f476c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:43:38Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "medium"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "add_noise",
+   0.2,
+   "all"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for 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": "515e5985ba09a137a8ef"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82eeff53066e8b2f476c/workflow.json b/records/sd1/sweep/82eeff53066e8b2f476c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d841bbf27b90841bebbfd3871fa53f6501910010
--- /dev/null
+++ b/records/sd1/sweep/82eeff53066e8b2f476c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82eeff53066e8b2f476c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 1234\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/82f2d9162beb033bdc5f/measurements.json b/records/sd1/sweep/82f2d9162beb033bdc5f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a8104e08fb5bbf2b885331003f802483bbd5f5dd
--- /dev/null
+++ b/records/sd1/sweep/82f2d9162beb033bdc5f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "82f2d9162beb033bdc5f",
+ "values": {
+  "std": {
+   "value": 27.584,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00778,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.201,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.51318,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.58367,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.61182,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.086963,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0589,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0612,
+   "method": "CLIP zero-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/82f2d9162beb033bdc5f/record.json b/records/sd1/sweep/82f2d9162beb033bdc5f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..347f6beb18458d5181e21c664a5b1872f798d0d0
--- /dev/null
+++ b/records/sd1/sweep/82f2d9162beb033bdc5f/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "82f2d9162beb033bdc5f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:03:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a 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": 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": "13216623a742ccb32c5e",
+  "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": "d8bc09c8a552d7cba5ec"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82f2d9162beb033bdc5f/workflow.json b/records/sd1/sweep/82f2d9162beb033bdc5f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a45141fc83b0758c134a62531a69b6a6d0fc2203
--- /dev/null
+++ b/records/sd1/sweep/82f2d9162beb033bdc5f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82f2d9162beb033bdc5f)"
+  },
+  "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 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/82f6ebc5715911f28482/measurements.json b/records/sd1/sweep/82f6ebc5715911f28482/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0ad951210eb81f38ba1ca3ec33b52c95b966126e
--- /dev/null
+++ b/records/sd1/sweep/82f6ebc5715911f28482/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "82f6ebc5715911f28482",
+ "values": {
+  "cosine_distance": {
+   "value": 0.032944,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.042744,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.024788,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.009969,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.548,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00573,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.508,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9928,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1726,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82f6ebc5715911f28482/record.json b/records/sd1/sweep/82f6ebc5715911f28482/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0e3500556d22c38cd204b88c85baf1043adfe2f
--- /dev/null
+++ b/records/sd1/sweep/82f6ebc5715911f28482/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "82f6ebc5715911f28482",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:15Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "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/aff7a9f72e8ea539559e52f38311e3242d026581791284a935e8a698de5c8444.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "aff7a9f72e8ea539559e52f38311e3242d026581791284a935e8a698de5c8444",
+  "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": "59ec20f7bc7c4594ec4e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/82f6ebc5715911f28482/workflow.json b/records/sd1/sweep/82f6ebc5715911f28482/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c08821986aa5e556a5d0e075fe7ce19f6f941f71
--- /dev/null
+++ b/records/sd1/sweep/82f6ebc5715911f28482/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 82f6ebc5715911f28482)"
+  },
+  "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\": 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/8305c5d3d03d6a4ab2ea/measurements.json b/records/sd1/sweep/8305c5d3d03d6a4ab2ea/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ad5fcdf4e5fc4d7b1883c8bde98ca8db920a067
--- /dev/null
+++ b/records/sd1/sweep/8305c5d3d03d6a4ab2ea/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8305c5d3d03d6a4ab2ea",
+ "values": {
+  "cosine_distance": {
+   "value": 0.215797,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.507196,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.246269,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.143841,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.934,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00184,
+   "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": 24.756,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0587,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5533,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8305c5d3d03d6a4ab2ea/record.json b/records/sd1/sweep/8305c5d3d03d6a4ab2ea/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4a636832674792e0feffa03340a3a201ce68fc27
--- /dev/null
+++ b/records/sd1/sweep/8305c5d3d03d6a4ab2ea/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8305c5d3d03d6a4ab2ea",
+ "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": 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/9206503f304bf3ba10c449900a351ef95d669e2543327024844b45675d50e7d3.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9206503f304bf3ba10c449900a351ef95d669e2543327024844b45675d50e7d3",
+  "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": "dcc43218382e76379741"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8305c5d3d03d6a4ab2ea/workflow.json b/records/sd1/sweep/8305c5d3d03d6a4ab2ea/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5d65139e6dd35a810d58d5c452a047c60747fd4
--- /dev/null
+++ b/records/sd1/sweep/8305c5d3d03d6a4ab2ea/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8305c5d3d03d6a4ab2ea)"
+  },
+  "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\": 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/8305ce9dd9aae97256ab/measurements.json b/records/sd1/sweep/8305ce9dd9aae97256ab/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b988ebaf72239b1beb3c37b439c8a787dafd6d5c
--- /dev/null
+++ b/records/sd1/sweep/8305ce9dd9aae97256ab/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8305ce9dd9aae97256ab",
+ "values": {
+  "cosine_distance": {
+   "value": 0.02842,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.046463,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.010436,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.006432,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.502,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00588,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 6.947,
+   "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": 0.102,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8305ce9dd9aae97256ab/record.json b/records/sd1/sweep/8305ce9dd9aae97256ab/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a513fe3f6df23043e2c8fccdaf09578dfc30092
--- /dev/null
+++ b/records/sd1/sweep/8305ce9dd9aae97256ab/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8305ce9dd9aae97256ab",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:29Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/2a2600d37aff8b49fc716f053410e84ea96ebe8698aab7aaf392e4dc3158c690.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2a2600d37aff8b49fc716f053410e84ea96ebe8698aab7aaf392e4dc3158c690",
+  "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": "f1bc672c8929aa70d3cc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8305ce9dd9aae97256ab/workflow.json b/records/sd1/sweep/8305ce9dd9aae97256ab/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..01afa0091ab403d5d7d987ff7af744c385d40db2
--- /dev/null
+++ b/records/sd1/sweep/8305ce9dd9aae97256ab/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8305ce9dd9aae97256ab)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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/83115d306dda66c3e9cb/measurements.json b/records/sd1/sweep/83115d306dda66c3e9cb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c7ece3541ef8ba2fc8c54b007ea77523de12007
--- /dev/null
+++ b/records/sd1/sweep/83115d306dda66c3e9cb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83115d306dda66c3e9cb",
+ "values": {
+  "cosine_distance": {
+   "value": 0.381905,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.394049,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.386564,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.391274,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.284,
+   "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": 35.407,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.833,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3897,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83115d306dda66c3e9cb/record.json b/records/sd1/sweep/83115d306dda66c3e9cb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc48fe3c86343adf987fb5b62d8e9e07a36d8382
--- /dev/null
+++ b/records/sd1/sweep/83115d306dda66c3e9cb/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "83115d306dda66c3e9cb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/d633571ab5987a6cfe5de6fc48c55435ab7e405dd63e45e2bced0d9e7adeb2d4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d633571ab5987a6cfe5de6fc48c55435ab7e405dd63e45e2bced0d9e7adeb2d4",
+  "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": "db802fdde4963ad2f15e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83115d306dda66c3e9cb/workflow.json b/records/sd1/sweep/83115d306dda66c3e9cb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc076d23f318d9fd7b1598bdbc6507269b79f329
--- /dev/null
+++ b/records/sd1/sweep/83115d306dda66c3e9cb/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 83115d306dda66c3e9cb)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.transformer_blocks.0.attn2.to_k\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83121ab4831de90776f1/measurements.json b/records/sd1/sweep/83121ab4831de90776f1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9af3fe89bad8a909fa92e381a628d72c78ffc0ba
--- /dev/null
+++ b/records/sd1/sweep/83121ab4831de90776f1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "83121ab4831de90776f1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.134568,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.297254,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.221073,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.075052,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.873,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00375,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.459,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9992,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0851,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83121ab4831de90776f1/record.json b/records/sd1/sweep/83121ab4831de90776f1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c006ded2516d82c0431b651091079f23e2b350fc
--- /dev/null
+++ b/records/sd1/sweep/83121ab4831de90776f1/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "83121ab4831de90776f1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1c67795e74859fe631007ffb7d297dc15f2f80681a2bdf4931f08066627b2c5a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1c67795e74859fe631007ffb7d297dc15f2f80681a2bdf4931f08066627b2c5a",
+  "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": "a6486c42b881c0078039"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83121ab4831de90776f1/workflow.json b/records/sd1/sweep/83121ab4831de90776f1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5b05fb7f5cf1e998c75691e2d0b2e8bab5fcb159
--- /dev/null
+++ b/records/sd1/sweep/83121ab4831de90776f1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83121ab4831de90776f1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.out_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/832f5b295fdbe15017be/interpretations.jsonl b/records/sd1/sweep/832f5b295fdbe15017be/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..00b1564bcb90f8ec5c0ea0d0e6b25fe8bb690faa
--- /dev/null
+++ b/records/sd1/sweep/832f5b295fdbe15017be/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "832f5b295fdbe15017be"}, "kind": "caption", "value": "Small framed photo of a blonde hooded figure hung on a pale grey wall", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "15ad97464c3aba43"}
+{"target": {"record": "832f5b295fdbe15017be"}, "kind": "change", "value": "The portrait shrinks into a small framed print of a different blonde, hooded person, hung on a large empty grey wall with a soft shadow.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7fe5e120da815333"}
+{"target": {"record": "832f5b295fdbe15017be"}, "kind": "keywords", "value": ["picture in picture", "framed photo", "grey wall", "blonde hood", "negative space"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4c4a1fc59337844c"}
+{"target": {"record": "832f5b295fdbe15017be"}, "kind": "effect_tags", "value": ["framing", "new_composition", "subject_lost", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:13:37Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "21f10f701d391d31"}
diff --git a/records/sd1/sweep/832f5b295fdbe15017be/measurements.json b/records/sd1/sweep/832f5b295fdbe15017be/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fbb6b1b036e60a0109a7cede8ece648c0d2e4523
--- /dev/null
+++ b/records/sd1/sweep/832f5b295fdbe15017be/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "832f5b295fdbe15017be",
+ "values": {
+  "cosine_distance": {
+   "value": 0.518739,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.569458,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.635766,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.373372,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.49,
+   "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": 49.308,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.843,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1025,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/832f5b295fdbe15017be/record.json b/records/sd1/sweep/832f5b295fdbe15017be/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..00189f94077fd963f2b13f4f433c144f89827780
--- /dev/null
+++ b/records/sd1/sweep/832f5b295fdbe15017be/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "832f5b295fdbe15017be",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:41Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0.skip_connection",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "in.mid",
+   "kind": "skip",
+   "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/25599502f6da9491a8f4511602a0798a236aa3915fe1812e72c7af47f9995433.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "25599502f6da9491a8f4511602a0798a236aa3915fe1812e72c7af47f9995433",
+  "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": "0bcc689ca08160086c3e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/832f5b295fdbe15017be/workflow.json b/records/sd1/sweep/832f5b295fdbe15017be/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2dcb3d6c55b786f3015d07ff3363c84de6267d72
--- /dev/null
+++ b/records/sd1/sweep/832f5b295fdbe15017be/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 832f5b295fdbe15017be)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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/83326684ea7e565e578e/measurements.json b/records/sd1/sweep/83326684ea7e565e578e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0ee36ae3da35bb61c66856709c5660d53d38c1b
--- /dev/null
+++ b/records/sd1/sweep/83326684ea7e565e578e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83326684ea7e565e578e",
+ "values": {
+  "std": {
+   "value": 37.182,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01545,
+   "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": 5.538,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.021526,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.062293,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.021042,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00374,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0064,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0129,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83326684ea7e565e578e/record.json b/records/sd1/sweep/83326684ea7e565e578e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f76e9faebb65df45aaa3f7cf3043c55b57accf9c
--- /dev/null
+++ b/records/sd1/sweep/83326684ea7e565e578e/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "83326684ea7e565e578e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:10:24Z",
+ "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 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": 0.333,
+   "window": "early",
+   "bucket": "mid",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "add_scalar",
+   -0.8,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "282843b24771476a1b84"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83326684ea7e565e578e/workflow.json b/records/sd1/sweep/83326684ea7e565e578e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..983341f6a4c572c3df1ae9718e30e8307fd51ce8
--- /dev/null
+++ b/records/sd1/sweep/83326684ea7e565e578e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83326684ea7e565e578e)"
+  },
+  "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 \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8335d2f64becf540fe02/measurements.json b/records/sd1/sweep/8335d2f64becf540fe02/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..740d3581122db2957f39047d838cc55b0ce1d190
--- /dev/null
+++ b/records/sd1/sweep/8335d2f64becf540fe02/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8335d2f64becf540fe02",
+ "values": {
+  "cosine_distance": {
+   "value": 0.251477,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.299359,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.129009,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.064461,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.746,
+   "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": 31.856,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0056,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2482,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8335d2f64becf540fe02/record.json b/records/sd1/sweep/8335d2f64becf540fe02/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a41a6a65df4c6a789996bfedd309edb2b6b670e
--- /dev/null
+++ b/records/sd1/sweep/8335d2f64becf540fe02/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8335d2f64becf540fe02",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.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/77447c279fd280f2f595b6c7d0b295042c17c85563b3b7c5a7d33b87f6049bdd.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "77447c279fd280f2f595b6c7d0b295042c17c85563b3b7c5a7d33b87f6049bdd",
+  "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": "67ce254931fe9b016e74"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8335d2f64becf540fe02/workflow.json b/records/sd1/sweep/8335d2f64becf540fe02/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6b90367ed292ed76100165376e3d408ae0c21e10
--- /dev/null
+++ b/records/sd1/sweep/8335d2f64becf540fe02/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 8335d2f64becf540fe02)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/8337195fa1e5d7d4a6e6/measurements.json b/records/sd1/sweep/8337195fa1e5d7d4a6e6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4865b03b1bae0002f454358248705e2567cc70d7
--- /dev/null
+++ b/records/sd1/sweep/8337195fa1e5d7d4a6e6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8337195fa1e5d7d4a6e6",
+ "values": {
+  "std": {
+   "value": 31.547,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01525,
+   "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": 54.6,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.669367,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.669183,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.544703,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.132,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0439,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0639,
+   "method": "CLIP zero-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/8337195fa1e5d7d4a6e6/record.json b/records/sd1/sweep/8337195fa1e5d7d4a6e6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3632dfcb0d9dbd268d35a04d3c884fcdf9915b19
--- /dev/null
+++ b/records/sd1/sweep/8337195fa1e5d7d4a6e6/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8337195fa1e5d7d4a6e6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:27Z",
+ "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": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "half",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "rotate",
+   180.0,
+   "early"
+  ],
+  "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": "65f530d0775eaf129b4a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8337195fa1e5d7d4a6e6/workflow.json b/records/sd1/sweep/8337195fa1e5d7d4a6e6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee743fa71fca4b7d6d9a72585b126163d4b6fb97
--- /dev/null
+++ b/records/sd1/sweep/8337195fa1e5d7d4a6e6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8337195fa1e5d7d4a6e6)"
+  },
+  "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\": 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/8338225bcbd70f23fd9a/interpretations.jsonl b/records/sd1/sweep/8338225bcbd70f23fd9a/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ad46935f14458154262045dcb3feaf0706e4fb2d
--- /dev/null
+++ b/records/sd1/sweep/8338225bcbd70f23fd9a/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8338225bcbd70f23fd9a"}, "kind": "caption", "value": "Realistic painted portrait of a freckled young woman with brown bob on grey background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:18Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "465babcadbc691f0"}
+{"target": {"record": "8338225bcbd70f23fd9a"}, "kind": "change", "value": "The stylised face becomes a more naturalistic, slightly freckled young woman with brown hair, and the framed dark background turns into a lighter full-bleed grey.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:18Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "140a3dda266ab358"}
+{"target": {"record": "8338225bcbd70f23fd9a"}, "kind": "keywords", "value": ["naturalistic", "freckles", "brown bob", "grey background"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:18Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f363738549e86171"}
+{"target": {"record": "8338225bcbd70f23fd9a"}, "kind": "effect_tags", "value": ["subject_kept", "style_change", "framing", "painterly"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:18Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "59208e061775d740"}
diff --git a/records/sd1/sweep/8338225bcbd70f23fd9a/measurements.json b/records/sd1/sweep/8338225bcbd70f23fd9a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f815e0082ead41830c831afc4b7f61ae85e336d0
--- /dev/null
+++ b/records/sd1/sweep/8338225bcbd70f23fd9a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8338225bcbd70f23fd9a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.308118,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.353193,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.160996,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.067812,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.066,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00433,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 34.345,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9819,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1015,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8338225bcbd70f23fd9a/record.json b/records/sd1/sweep/8338225bcbd70f23fd9a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..589e8ca9a5f24689b29b6dd91190d7f06763bec9
--- /dev/null
+++ b/records/sd1/sweep/8338225bcbd70f23fd9a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8338225bcbd70f23fd9a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.transformer_blocks.0.norm1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1fc91b817aadce35b09af27a609418d6eef9a4513c09db8d0164407d5d2cedd2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1fc91b817aadce35b09af27a609418d6eef9a4513c09db8d0164407d5d2cedd2",
+  "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": "8052ac7a54123a5469c4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8338225bcbd70f23fd9a/workflow.json b/records/sd1/sweep/8338225bcbd70f23fd9a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..536706ea740bb8372ba035ef3549d61c93b706e5
--- /dev/null
+++ b/records/sd1/sweep/8338225bcbd70f23fd9a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8338225bcbd70f23fd9a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.norm1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/8341b186fc83819027e5/measurements.json b/records/sd1/sweep/8341b186fc83819027e5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..908bb7d0342e7f1874c42f05477578a1cea75da6
--- /dev/null
+++ b/records/sd1/sweep/8341b186fc83819027e5/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8341b186fc83819027e5",
+ "values": {
+  "cosine_distance": {
+   "value": 0.189626,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.379914,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.290464,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.187697,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.725,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00349,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.086,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9961,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1177,
+   "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/8341b186fc83819027e5/record.json b/records/sd1/sweep/8341b186fc83819027e5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..479ed195f80b8c82e6b38b2d19b1994b880f9a75
--- /dev/null
+++ b/records/sd1/sweep/8341b186fc83819027e5/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8341b186fc83819027e5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/1a9b611c4a18cfc13c18bb388dbe0e52ae66485ce4dbac460de599db4e56ee78.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "1a9b611c4a18cfc13c18bb388dbe0e52ae66485ce4dbac460de599db4e56ee78",
+  "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": "95b50ce75b9bd4c4691f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8341b186fc83819027e5/workflow.json b/records/sd1/sweep/8341b186fc83819027e5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1098535d673c03d70a56c4cffa73e03424595b92
--- /dev/null
+++ b/records/sd1/sweep/8341b186fc83819027e5/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 8341b186fc83819027e5)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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/8346f7c0233b5d4d90fc/measurements.json b/records/sd1/sweep/8346f7c0233b5d4d90fc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3c067e8b9e98184b4fe7e6a50e6728575610aa68
--- /dev/null
+++ b/records/sd1/sweep/8346f7c0233b5d4d90fc/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8346f7c0233b5d4d90fc",
+ "values": {
+  "std": {
+   "value": 51.11,
+   "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": 1.108,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000784,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.003022,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.001202,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000409,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0094,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0148,
+   "method": "CLIP zero-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/8346f7c0233b5d4d90fc/record.json b/records/sd1/sweep/8346f7c0233b5d4d90fc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd21d89094884f82f4118fe8cd60326d1810c095
--- /dev/null
+++ b/records/sd1/sweep/8346f7c0233b5d4d90fc/record.json
@@ -0,0 +1,86 @@
+{
+ "id": "8346f7c0233b5d4d90fc",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:40:55Z",
+ "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": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "medium",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_noise",
+   0.2,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "6f073a5d2156c5d18cb9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8346f7c0233b5d4d90fc/workflow.json b/records/sd1/sweep/8346f7c0233b5d4d90fc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b542c33decaa64809bf1b9b5c33b51b00ca208d
--- /dev/null
+++ b/records/sd1/sweep/8346f7c0233b5d4d90fc/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8346f7c0233b5d4d90fc)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 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/83482dc0f05db4165e36/measurements.json b/records/sd1/sweep/83482dc0f05db4165e36/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..95043fd0d87308145def4641f654f95cbfd6f997
--- /dev/null
+++ b/records/sd1/sweep/83482dc0f05db4165e36/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "83482dc0f05db4165e36",
+ "values": {
+  "cosine_distance": {
+   "value": 0.106797,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.175272,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.102215,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.10824,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.534,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00287,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 16.583,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0175,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4027,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83482dc0f05db4165e36/record.json b/records/sd1/sweep/83482dc0f05db4165e36/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8fdbc3c34d86a5e7cd54bf190fd67b55d078480
--- /dev/null
+++ b/records/sd1/sweep/83482dc0f05db4165e36/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "83482dc0f05db4165e36",
+ "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.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/43c46ecff7d860ee1dbb4209d5c7e3461a20eeb9ee36b151a4e254f31d89d4cf.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "43c46ecff7d860ee1dbb4209d5c7e3461a20eeb9ee36b151a4e254f31d89d4cf",
+  "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": "deb4c4840985db0118a9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83482dc0f05db4165e36/workflow.json b/records/sd1/sweep/83482dc0f05db4165e36/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d8671d58ea3e447e6b9f01a5f1aad3788ef6b233
--- /dev/null
+++ b/records/sd1/sweep/83482dc0f05db4165e36/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 83482dc0f05db4165e36)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/834e13f168a406609b3c/measurements.json b/records/sd1/sweep/834e13f168a406609b3c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8ded1eecdf1d2c4573f34e212b7503a519dfa487
--- /dev/null
+++ b/records/sd1/sweep/834e13f168a406609b3c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "834e13f168a406609b3c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.251067,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.405734,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.133812,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.104611,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 44.355,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00364,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.373,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0151,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6515,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/834e13f168a406609b3c/record.json b/records/sd1/sweep/834e13f168a406609b3c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..cd8baea6b40d2c3eeaebf98bfb389b7eeaf90a80
--- /dev/null
+++ b/records/sd1/sweep/834e13f168a406609b3c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "834e13f168a406609b3c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:08Z",
+ "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": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/8f085c510562b4f2a2e23f505fb895f61938ec77f37183fdee9e9f07ba6187b4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8f085c510562b4f2a2e23f505fb895f61938ec77f37183fdee9e9f07ba6187b4",
+  "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": "2f62da0dd9aa9505f31b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/834e13f168a406609b3c/workflow.json b/records/sd1/sweep/834e13f168a406609b3c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..366da19a36e8da179dc230f957a987f0f615ca8b
--- /dev/null
+++ b/records/sd1/sweep/834e13f168a406609b3c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 834e13f168a406609b3c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.in_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/834eedd8b484f6386ae0/measurements.json b/records/sd1/sweep/834eedd8b484f6386ae0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b017d4192f5bbb641fd98c116c19508c162cfa7
--- /dev/null
+++ b/records/sd1/sweep/834eedd8b484f6386ae0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "834eedd8b484f6386ae0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.191984,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.234886,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.100116,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.056529,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.168,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00578,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.583,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.975,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0457,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/834eedd8b484f6386ae0/record.json b/records/sd1/sweep/834eedd8b484f6386ae0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9379a1c6712a712b4d36d37b2622a32009a6a01a
--- /dev/null
+++ b/records/sd1/sweep/834eedd8b484f6386ae0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "834eedd8b484f6386ae0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:07Z",
+ "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": "input_blocks.7.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/0d2c05164ba49d38fb2c0a72e558d1fa2a8a9fb3021a99da4171cbefd0e38440.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0d2c05164ba49d38fb2c0a72e558d1fa2a8a9fb3021a99da4171cbefd0e38440",
+  "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": "f73cd8698a19f3b326e0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/834eedd8b484f6386ae0/workflow.json b/records/sd1/sweep/834eedd8b484f6386ae0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8b1016df0e352627c35c0a93337945a39f84c49d
--- /dev/null
+++ b/records/sd1/sweep/834eedd8b484f6386ae0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 834eedd8b484f6386ae0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.attn2.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83537511fda0f267fdc8/measurements.json b/records/sd1/sweep/83537511fda0f267fdc8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6176e1bbcc1dee50a29019b24ccf2b607aee4e74
--- /dev/null
+++ b/records/sd1/sweep/83537511fda0f267fdc8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83537511fda0f267fdc8",
+ "values": {
+  "std": {
+   "value": 36.541,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02321,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.8,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.342114,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.507513,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.174416,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.037011,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0357,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2377,
+   "method": "CLIP zero-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/83537511fda0f267fdc8/record.json b/records/sd1/sweep/83537511fda0f267fdc8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..edf030b9774a284f97eeeb0b9dfc60c05f04b4f6
--- /dev/null
+++ b/records/sd1/sweep/83537511fda0f267fdc8/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "83537511fda0f267fdc8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:20:47Z",
+ "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": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "zero",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "multiply",
+   0.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "dfdeb1e3d285fc0d75fd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83537511fda0f267fdc8/workflow.json b/records/sd1/sweep/83537511fda0f267fdc8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4cdfd31b8442b6228717a9457a9470bec5e491
--- /dev/null
+++ b/records/sd1/sweep/83537511fda0f267fdc8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83537511fda0f267fdc8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 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/8356b3fa432ada359be5/measurements.json b/records/sd1/sweep/8356b3fa432ada359be5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0cbb752570fe1b890310fd994bd08bc5485355e6
--- /dev/null
+++ b/records/sd1/sweep/8356b3fa432ada359be5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8356b3fa432ada359be5",
+ "values": {
+  "std": {
+   "value": 27.4,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02697,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.616,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.031835,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.113642,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.155576,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.021782,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0131,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3018,
+   "method": "CLIP zero-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/8356b3fa432ada359be5/record.json b/records/sd1/sweep/8356b3fa432ada359be5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1e53cdc26d5ff6f2c41090c6991374b7148678a
--- /dev/null
+++ b/records/sd1/sweep/8356b3fa432ada359be5/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8356b3fa432ada359be5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:55:56Z",
+ "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": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.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": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.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": "d52e4f36dfc4fd494db5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8356b3fa432ada359be5/workflow.json b/records/sd1/sweep/8356b3fa432ada359be5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5bc21b0bbf58b70fbeb6d517e5d2941ec9256bc
--- /dev/null
+++ b/records/sd1/sweep/8356b3fa432ada359be5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8356b3fa432ada359be5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/835bfe792299ff0fc8ae/measurements.json b/records/sd1/sweep/835bfe792299ff0fc8ae/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..35017ed6d1f347c7d3b43ba2b3f2a0aab24cfc0a
--- /dev/null
+++ b/records/sd1/sweep/835bfe792299ff0fc8ae/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "835bfe792299ff0fc8ae",
+ "values": {
+  "cosine_distance": {
+   "value": 0.110317,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.161069,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.054239,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.021932,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.652,
+   "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": 17.584,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9686,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1182,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/835bfe792299ff0fc8ae/record.json b/records/sd1/sweep/835bfe792299ff0fc8ae/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b40298c35626c192c82ea877107cc987dc003c6
--- /dev/null
+++ b/records/sd1/sweep/835bfe792299ff0fc8ae/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "835bfe792299ff0fc8ae",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/8c6ef0c4b21f3679e4fb098f483bf78b37e587102635ce971cfd2dd0953880da.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8c6ef0c4b21f3679e4fb098f483bf78b37e587102635ce971cfd2dd0953880da",
+  "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": "369c49fb29bace734655"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/835bfe792299ff0fc8ae/workflow.json b/records/sd1/sweep/835bfe792299ff0fc8ae/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2d01b00b6cbf3f5ca34fc9e9e3b18f777659e9a8
--- /dev/null
+++ b/records/sd1/sweep/835bfe792299ff0fc8ae/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 835bfe792299ff0fc8ae)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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/836277ebb257607551b6/interpretations.jsonl b/records/sd1/sweep/836277ebb257607551b6/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..f7daff5480ec287fb44ae58fbb2db1b43e40ab2b
--- /dev/null
+++ b/records/sd1/sweep/836277ebb257607551b6/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "836277ebb257607551b6"}, "kind": "caption", "value": "Short-haired young person staring forward with serious face in grey framed portrait", "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": "9e6f19f73cc44e00"}
+{"target": {"record": "836277ebb257607551b6"}, "kind": "change", "value": "The subject becomes more androgynous and frontal with a sterner expression, keeping the grey palette and frame.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5f1f91cb2392b633"}
+{"target": {"record": "836277ebb257607551b6"}, "kind": "keywords", "value": ["androgynous portrait", "serious expression", "frontal gaze", "grey backdrop"], "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": "81d2b3cc8fbfb512"}
+{"target": {"record": "836277ebb257607551b6"}, "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": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9b18a2125a75cc9d"}
diff --git a/records/sd1/sweep/836277ebb257607551b6/measurements.json b/records/sd1/sweep/836277ebb257607551b6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3ec30b2339b926d656beab6edcd70b59024bc895
--- /dev/null
+++ b/records/sd1/sweep/836277ebb257607551b6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "836277ebb257607551b6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.105533,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.107279,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.080923,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.125551,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.741,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00441,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.133,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9317,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.658,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/836277ebb257607551b6/record.json b/records/sd1/sweep/836277ebb257607551b6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6e43df0fb9c56f3b2d41d18bada908dac6fbcef0
--- /dev/null
+++ b/records/sd1/sweep/836277ebb257607551b6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "836277ebb257607551b6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:23Z",
+ "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.8.1.proj_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "shrink"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/38cdb5dcbf02d322b5539935dc5ba9221e91257cecdb36b67685e61f8745e12f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "38cdb5dcbf02d322b5539935dc5ba9221e91257cecdb36b67685e61f8745e12f",
+  "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": "69a1062cefc005affb3b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/836277ebb257607551b6/workflow.json b/records/sd1/sweep/836277ebb257607551b6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b2c6939fbc215827669a145da0e858d6bde13f11
--- /dev/null
+++ b/records/sd1/sweep/836277ebb257607551b6/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 836277ebb257607551b6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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/8367015def398e7bacae/measurements.json b/records/sd1/sweep/8367015def398e7bacae/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc70def5b81a3286cc4731085b468668755bd30e
--- /dev/null
+++ b/records/sd1/sweep/8367015def398e7bacae/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8367015def398e7bacae",
+ "values": {
+  "std": {
+   "value": 50.035,
+   "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"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.815,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.228735,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.590475,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.843495,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.363368,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.637,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8109,
+   "method": "CLIP zero-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/8367015def398e7bacae/record.json b/records/sd1/sweep/8367015def398e7bacae/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0395d0352e4d82a6ccb78591ab994e2cd06b2f16
--- /dev/null
+++ b/records/sd1/sweep/8367015def398e7bacae/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8367015def398e7bacae",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:23:17Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a 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": 90.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "quarter",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [0.0, 5.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ccad20fcd80e9efc8926"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8367015def398e7bacae/workflow.json b/records/sd1/sweep/8367015def398e7bacae/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..770ce96d01dedf8d0bcce41a79c63aaea1282384
--- /dev/null
+++ b/records/sd1/sweep/8367015def398e7bacae/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8367015def398e7bacae)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/837911bd1e256772afe6/measurements.json b/records/sd1/sweep/837911bd1e256772afe6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..06ccb188eec9fda8c9cecf7bb16fecb7323bde62
--- /dev/null
+++ b/records/sd1/sweep/837911bd1e256772afe6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "837911bd1e256772afe6",
+ "values": {
+  "std": {
+   "value": 41.621,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01506,
+   "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": 22.382,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.201936,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.325029,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.140154,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.040808,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0568,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0079,
+   "method": "CLIP zero-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/837911bd1e256772afe6/record.json b/records/sd1/sweep/837911bd1e256772afe6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8eea87d57e91aec29fdf156fa8a0984706a1868f
--- /dev/null
+++ b/records/sd1/sweep/837911bd1e256772afe6/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "837911bd1e256772afe6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:48:19Z",
+ "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": "input_blocks.2.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "heavy"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "add_noise",
+   0.8,
+   "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": "9751e4a407d460f4d756"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/837911bd1e256772afe6/workflow.json b/records/sd1/sweep/837911bd1e256772afe6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ab6ea78b6e0177602f0c989ea0dca612ebaea30
--- /dev/null
+++ b/records/sd1/sweep/837911bd1e256772afe6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 837911bd1e256772afe6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ]\n}",
+   "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/837fbe2fd8275174d4e9/interpretations.jsonl b/records/sd1/sweep/837fbe2fd8275174d4e9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..86fcb239d33dfd6f42051ca0c8899041607fad89
--- /dev/null
+++ b/records/sd1/sweep/837fbe2fd8275174d4e9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "837fbe2fd8275174d4e9"}, "kind": "caption", "value": "Illustrated portrait of dark-haired woman with slightly bolder eyes and darker hair", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e6e1729a91fe84a0"}
+{"target": {"record": "837fbe2fd8275174d4e9"}, "kind": "change", "value": "Almost no visible change; the eyes are a bit more defined and the hair a touch darker.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c424028fb1b589d6"}
+{"target": {"record": "837fbe2fd8275174d4e9"}, "kind": "keywords", "value": ["near identical", "defined eyes", "illustrated portrait"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4193a81f844b2e9d"}
+{"target": {"record": "837fbe2fd8275174d4e9"}, "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": "2026-10-05T21:27:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "87290ad4ffa5a4d3"}
diff --git a/records/sd1/sweep/837fbe2fd8275174d4e9/measurements.json b/records/sd1/sweep/837fbe2fd8275174d4e9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..32f8d02a567329ee5578d7697b21e5795b073797
--- /dev/null
+++ b/records/sd1/sweep/837fbe2fd8275174d4e9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "837fbe2fd8275174d4e9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.059918,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.107554,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.156912,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01515,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.321,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00776,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.105,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0189,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.136,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/837fbe2fd8275174d4e9/record.json b/records/sd1/sweep/837fbe2fd8275174d4e9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d7aae95429027b3e279108acc6df462225f4078
--- /dev/null
+++ b/records/sd1/sweep/837fbe2fd8275174d4e9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "837fbe2fd8275174d4e9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.out_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 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/e640a76823f9209b144244f4b62cc2dbe5e7e401e3dbeb181e00e0b4a0c4914d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e640a76823f9209b144244f4b62cc2dbe5e7e401e3dbeb181e00e0b4a0c4914d",
+  "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": "c52543aef7adefa0027a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/837fbe2fd8275174d4e9/workflow.json b/records/sd1/sweep/837fbe2fd8275174d4e9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c81428a73df5a40ee478bf8280f45b253db2b84
--- /dev/null
+++ b/records/sd1/sweep/837fbe2fd8275174d4e9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 837fbe2fd8275174d4e9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.out_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/8380bf2343ea8041f925/measurements.json b/records/sd1/sweep/8380bf2343ea8041f925/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..008e669a1cfeaaf70aca57cf3b207be22b87ae9b
--- /dev/null
+++ b/records/sd1/sweep/8380bf2343ea8041f925/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8380bf2343ea8041f925",
+ "values": {
+  "cosine_distance": {
+   "value": 0.244977,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.355937,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.098481,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.078454,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 42.43,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00506,
+   "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": 30.835,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9794,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3525,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8380bf2343ea8041f925/record.json b/records/sd1/sweep/8380bf2343ea8041f925/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a126e3f9054f316ed3037c83d9828d7b64cee24b
--- /dev/null
+++ b/records/sd1/sweep/8380bf2343ea8041f925/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8380bf2343ea8041f925",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:59Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9e1cdb3f519a4a67116451a06d9c5d7e5682c87f111ff98d001b7f9794de16de.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9e1cdb3f519a4a67116451a06d9c5d7e5682c87f111ff98d001b7f9794de16de",
+  "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": "d2b5782355a502737c62"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8380bf2343ea8041f925/workflow.json b/records/sd1/sweep/8380bf2343ea8041f925/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff017540ab99bdb535207664f12b8c0c7ac24489
--- /dev/null
+++ b/records/sd1/sweep/8380bf2343ea8041f925/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8380bf2343ea8041f925)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/83820d39ee0ddd135de1/measurements.json b/records/sd1/sweep/83820d39ee0ddd135de1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4651eeff8f4dfdd602692e931d59f5b319842db
--- /dev/null
+++ b/records/sd1/sweep/83820d39ee0ddd135de1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83820d39ee0ddd135de1",
+ "values": {
+  "std": {
+   "value": 26.741,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02217,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.532,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.050484,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.170686,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.427232,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.063675,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9387,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2591,
+   "method": "CLIP zero-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/83820d39ee0ddd135de1/record.json b/records/sd1/sweep/83820d39ee0ddd135de1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..acf0e7d821db369bd4a4a8346a067b4a76e13144
--- /dev/null
+++ b/records/sd1/sweep/83820d39ee0ddd135de1/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "83820d39ee0ddd135de1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:03:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.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": "1bbda6ca1f7efc631dfb",
+  "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": "2a73166dab49f0163a43"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83820d39ee0ddd135de1/workflow.json b/records/sd1/sweep/83820d39ee0ddd135de1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c2a25e5f73cecf5add655fe1eab716f42100012
--- /dev/null
+++ b/records/sd1/sweep/83820d39ee0ddd135de1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83820d39ee0ddd135de1)"
+  },
+  "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": "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/8383801959ec04f9c10a/measurements.json b/records/sd1/sweep/8383801959ec04f9c10a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7fcf941411be658322c7b10707f4a8271966a4a
--- /dev/null
+++ b/records/sd1/sweep/8383801959ec04f9c10a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8383801959ec04f9c10a",
+ "values": {
+  "std": {
+   "value": 13.748,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01776,
+   "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": 45.229,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.4936,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.678521,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.621123,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.271224,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8453,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6931,
+   "method": "CLIP zero-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/8383801959ec04f9c10a/record.json b/records/sd1/sweep/8383801959ec04f9c10a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0aee427fd7267834bb5bf4c74c7b326a7d7b94e1
--- /dev/null
+++ b/records/sd1/sweep/8383801959ec04f9c10a/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8383801959ec04f9c10a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:30Z",
+ "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": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "half",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "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": "78f47559a2e23cbc8d18"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8383801959ec04f9c10a/workflow.json b/records/sd1/sweep/8383801959ec04f9c10a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec58dd5166852b5a0b03ab26b296e1e7be21a29d
--- /dev/null
+++ b/records/sd1/sweep/8383801959ec04f9c10a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8383801959ec04f9c10a)"
+  },
+  "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 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/838dd326c62ee4dc344e/measurements.json b/records/sd1/sweep/838dd326c62ee4dc344e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08270aee16913ba66e03b7503b007031c57f2b9d
--- /dev/null
+++ b/records/sd1/sweep/838dd326c62ee4dc344e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "838dd326c62ee4dc344e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.419894,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.567575,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.528347,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.247773,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 35.974,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00349,
+   "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": 39.559,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0488,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2259,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/838dd326c62ee4dc344e/record.json b/records/sd1/sweep/838dd326c62ee4dc344e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1568e8faba0331dbadb6bccef14e25fdbb1a7838
--- /dev/null
+++ b/records/sd1/sweep/838dd326c62ee4dc344e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "838dd326c62ee4dc344e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:48Z",
+ "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": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9f8f3f29bcc8301bdf5660b7778e36ef1f1f149c537be8bd7fc75e31e1f4a68b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9f8f3f29bcc8301bdf5660b7778e36ef1f1f149c537be8bd7fc75e31e1f4a68b",
+  "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": "bf89c8605a2e8d278be7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/838dd326c62ee4dc344e/workflow.json b/records/sd1/sweep/838dd326c62ee4dc344e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..82936b01b661de661278f63c5c6874ef14bf7d2c
--- /dev/null
+++ b/records/sd1/sweep/838dd326c62ee4dc344e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 838dd326c62ee4dc344e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.out_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/838fd56cdf079efbd9d9/measurements.json b/records/sd1/sweep/838fd56cdf079efbd9d9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c12e8baf80fb4b25b3fc2b59970d1aceca48a8b8
--- /dev/null
+++ b/records/sd1/sweep/838fd56cdf079efbd9d9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "838fd56cdf079efbd9d9",
+ "values": {
+  "std": {
+   "value": 44.505,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01743,
+   "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": 3.641,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.004737,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.023051,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004467,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003434,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0025,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0551,
+   "method": "CLIP zero-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/838fd56cdf079efbd9d9/record.json b/records/sd1/sweep/838fd56cdf079efbd9d9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..dcb977f01ab1c87c098bed3ca9ae892f753dd1db
--- /dev/null
+++ b/records/sd1/sweep/838fd56cdf079efbd9d9/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "838fd56cdf079efbd9d9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:52:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "heavy",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "add_noise",
+   0.8,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "104adaef82b717ab363f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/838fd56cdf079efbd9d9/workflow.json b/records/sd1/sweep/838fd56cdf079efbd9d9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5cbbd3e78133281c16da424489cfb2d64786ee8
--- /dev/null
+++ b/records/sd1/sweep/838fd56cdf079efbd9d9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 838fd56cdf079efbd9d9)"
+  },
+  "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 \"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/83958a657cab794d37b0/measurements.json b/records/sd1/sweep/83958a657cab794d37b0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..754401ba7c69757638a30e3504ebd91645e39c00
--- /dev/null
+++ b/records/sd1/sweep/83958a657cab794d37b0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "83958a657cab794d37b0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.376882,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.375748,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.237359,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.208617,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.802,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00588,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 38.662,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.959,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2076,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83958a657cab794d37b0/record.json b/records/sd1/sweep/83958a657cab794d37b0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab61af039b64cc1b6ee5f32b0d3ddf950cc1a2fd
--- /dev/null
+++ b/records/sd1/sweep/83958a657cab794d37b0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "83958a657cab794d37b0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:42Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "mid",
+   "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/30dd2a1ec4c4b842be008c6e3cf120cb3e878814d8d0546cae6dfc454b4aa7eb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "30dd2a1ec4c4b842be008c6e3cf120cb3e878814d8d0546cae6dfc454b4aa7eb",
+  "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": "24bd57f7508bb4ed9fe0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83958a657cab794d37b0/workflow.json b/records/sd1/sweep/83958a657cab794d37b0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9c1a1b47abf914fd315166ec27aedbfbfb8264b
--- /dev/null
+++ b/records/sd1/sweep/83958a657cab794d37b0/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 83958a657cab794d37b0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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/83a26a66b86c40634776/interpretations.jsonl b/records/sd1/sweep/83a26a66b86c40634776/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..1fae28df9d7a824b7a17985a102f235caacd84c0
--- /dev/null
+++ b/records/sd1/sweep/83a26a66b86c40634776/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "83a26a66b86c40634776"}, "kind": "caption", "value": "Minimal still life of glass fruit bowl and leafy glass vase on a wooden table", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "421679f1da3c427c"}
+{"target": {"record": "83a26a66b86c40634776"}, "kind": "change", "value": "Fewer objects: the berry jars disappear, a clear glass bowl holds the fruit and a leafy vase stands behind, in a quieter, softer composition with more empty space.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c1cdd20879f73455"}
+{"target": {"record": "83a26a66b86c40634776"}, "kind": "keywords", "value": ["still life", "glass bowl", "fruit", "minimal composition", "leaves"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "be9b46c376f4873f"}
+{"target": {"record": "83a26a66b86c40634776"}, "kind": "effect_tags", "value": ["new_composition", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cef81652c5e088b7"}
diff --git a/records/sd1/sweep/83a26a66b86c40634776/measurements.json b/records/sd1/sweep/83a26a66b86c40634776/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..db2ee3384fc55686dc7dc9009c23fd853dda276f
--- /dev/null
+++ b/records/sd1/sweep/83a26a66b86c40634776/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83a26a66b86c40634776",
+ "values": {
+  "std": {
+   "value": 37.116,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00779,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.405,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.315033,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.500296,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.279996,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044448,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9706,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0437,
+   "method": "CLIP zero-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/83a26a66b86c40634776/record.json b/records/sd1/sweep/83a26a66b86c40634776/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe7cdb24135b57aa76f01a074068ab7dd550b1c9
--- /dev/null
+++ b/records/sd1/sweep/83a26a66b86c40634776/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "83a26a66b86c40634776",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:39:42Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "slight",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "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": "aea9109b253462ba1644"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83a26a66b86c40634776/workflow.json b/records/sd1/sweep/83a26a66b86c40634776/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..183984b81479b0e2df79056e8e4edf8a5ad317b5
--- /dev/null
+++ b/records/sd1/sweep/83a26a66b86c40634776/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83a26a66b86c40634776)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83a4c0249a611e4db55d/measurements.json b/records/sd1/sweep/83a4c0249a611e4db55d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..99d7623abdaba6112471c85fb6af5997922dcbb2
--- /dev/null
+++ b/records/sd1/sweep/83a4c0249a611e4db55d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83a4c0249a611e4db55d",
+ "values": {
+  "std": {
+   "value": 28.373,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02802,
+   "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": 8.855,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.031897,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.095601,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.112743,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.019791,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0012,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3263,
+   "method": "CLIP zero-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/83a4c0249a611e4db55d/record.json b/records/sd1/sweep/83a4c0249a611e4db55d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..45d183de019de0894d16ada2a75eadecadd95652
--- /dev/null
+++ b/records/sd1/sweep/83a4c0249a611e4db55d/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "83a4c0249a611e4db55d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:00:22Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "zero",
+   "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",
+   "multiply",
+   0.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "dedd73c94bb2e2347c11"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83a4c0249a611e4db55d/workflow.json b/records/sd1/sweep/83a4c0249a611e4db55d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa9cc8ed9a1fb3c1d30ccd2d88859912c9ad7dff
--- /dev/null
+++ b/records/sd1/sweep/83a4c0249a611e4db55d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83a4c0249a611e4db55d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 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/83af3cd320bdea456858/measurements.json b/records/sd1/sweep/83af3cd320bdea456858/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cc54f6bf7d062ed3fb1abef9b8293eefbfcc2d4
--- /dev/null
+++ b/records/sd1/sweep/83af3cd320bdea456858/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83af3cd320bdea456858",
+ "values": {
+  "cosine_distance": {
+   "value": 0.186442,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.242318,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.094773,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.033661,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.771,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00617,
+   "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": 26.104,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.947,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3843,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83af3cd320bdea456858/record.json b/records/sd1/sweep/83af3cd320bdea456858/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..40934b389cfce591138519705313bce49d4e9e57
--- /dev/null
+++ b/records/sd1/sweep/83af3cd320bdea456858/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "83af3cd320bdea456858",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/84e41199c2b36f001bfabdc0ba29ce0531e68cfe3e1f2488c1f05929858994e6.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "84e41199c2b36f001bfabdc0ba29ce0531e68cfe3e1f2488c1f05929858994e6",
+  "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": "a5e78e847c6ace5ab68f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83af3cd320bdea456858/workflow.json b/records/sd1/sweep/83af3cd320bdea456858/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f258e56b725fb2418bea877cd8bb302655868cc
--- /dev/null
+++ b/records/sd1/sweep/83af3cd320bdea456858/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83af3cd320bdea456858)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.in_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83b77562d221bf01bc15/measurements.json b/records/sd1/sweep/83b77562d221bf01bc15/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba1a405be01f98e3037660d48a8a45f5819d2016
--- /dev/null
+++ b/records/sd1/sweep/83b77562d221bf01bc15/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83b77562d221bf01bc15",
+ "values": {
+  "cosine_distance": {
+   "value": 0.214691,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.256375,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.088895,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.093506,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 47.461,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.015,
+   "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": 27.043,
+   "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": 0.0184,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83b77562d221bf01bc15/record.json b/records/sd1/sweep/83b77562d221bf01bc15/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3fa3fddc881b71c6c67194fa8ec06f725d2a32ee
--- /dev/null
+++ b/records/sd1/sweep/83b77562d221bf01bc15/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "83b77562d221bf01bc15",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "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/8c727e51029e71fe76cbda8fb31823740df1babfd29f6011b008982ba45e930d.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8c727e51029e71fe76cbda8fb31823740df1babfd29f6011b008982ba45e930d",
+  "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": "1bc2bc9498cf2f2a61ca"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83b77562d221bf01bc15/workflow.json b/records/sd1/sweep/83b77562d221bf01bc15/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4befbabb5edcb9b46a5b4bc61c3850b04954ab40
--- /dev/null
+++ b/records/sd1/sweep/83b77562d221bf01bc15/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83b77562d221bf01bc15)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.1.transformer_blocks.0.attn2.to_k\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\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/83c5a08e86a3fc9ae7ab/interpretations.jsonl b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..892329d0efda4079da8b9aef971f811aa88673b6
--- /dev/null
+++ b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "83c5a08e86a3fc9ae7ab"}, "kind": "caption", "value": "Still life of apples, lemons and a flower bouquet in bowls on a wooden table", "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": "0a6b195a237b3908"}
+{"target": {"record": "83c5a08e86a3fc9ae7ab"}, "kind": "change", "value": "Still a fruit still life, but recomposed: the glass vases are replaced by a flower-and-fruit arrangement, there are more bowls of yellow fruit, and the palette is lighter and greyer.", "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": "e20795efac006a47"}
+{"target": {"record": "83c5a08e86a3fc9ae7ab"}, "kind": "keywords", "value": ["still life", "fruit bowl", "flowers", "apples and lemons", "painting"], "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": "d003f2d45df7304d"}
+{"target": {"record": "83c5a08e86a3fc9ae7ab"}, "kind": "effect_tags", "value": ["subject_kept", "new_composition", "painterly", "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": "5116daea6dd9c4f7"}
diff --git a/records/sd1/sweep/83c5a08e86a3fc9ae7ab/measurements.json b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b968652500d1571f871abb356407a7f27dc41f99
--- /dev/null
+++ b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83c5a08e86a3fc9ae7ab",
+ "values": {
+  "std": {
+   "value": 32.152,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02071,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 42.422,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.544437,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.594181,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.408274,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.084932,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0434,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0934,
+   "method": "CLIP zero-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/83c5a08e86a3fc9ae7ab/record.json b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ffd3b9a3f1401b45ce69de5c5b148532589f1faa
--- /dev/null
+++ b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "83c5a08e86a3fc9ae7ab",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:28:00Z",
+ "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 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": 0.333,
+   "window": "early",
+   "bucket": "boost",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "multiply",
+   2.0,
+   "early"
+  ],
+  "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": "33c470031f3d14a9c74f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83c5a08e86a3fc9ae7ab/workflow.json b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..203fb6e8d56724cd03c5dba5f51745ce8ac69eae
--- /dev/null
+++ b/records/sd1/sweep/83c5a08e86a3fc9ae7ab/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83c5a08e86a3fc9ae7ab)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 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/83cc9ef76c3f1dda5280/measurements.json b/records/sd1/sweep/83cc9ef76c3f1dda5280/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3eeed0c1038a38cfa66429fff3da35d21659e97
--- /dev/null
+++ b/records/sd1/sweep/83cc9ef76c3f1dda5280/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "83cc9ef76c3f1dda5280",
+ "values": {
+  "cosine_distance": {
+   "value": 0.355916,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.438105,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.271956,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.077446,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 39.319,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00469,
+   "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": 39.802,
+   "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": 0.0313,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83cc9ef76c3f1dda5280/record.json b/records/sd1/sweep/83cc9ef76c3f1dda5280/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d888cf9cd19ea487943dce62105644441f59c3ab
--- /dev/null
+++ b/records/sd1/sweep/83cc9ef76c3f1dda5280/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "83cc9ef76c3f1dda5280",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:23Z",
+ "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.1.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/617e9ac7e6cd2f7e82599be92bb00b065555e9f2e17c3f9c7181838f815bdf96.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "617e9ac7e6cd2f7e82599be92bb00b065555e9f2e17c3f9c7181838f815bdf96",
+  "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": "3c10703d40e850f8ea64"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83cc9ef76c3f1dda5280/workflow.json b/records/sd1/sweep/83cc9ef76c3f1dda5280/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0e29fa607aeaae0e4d4c2b80e568f4dda31bcbf4
--- /dev/null
+++ b/records/sd1/sweep/83cc9ef76c3f1dda5280/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 83cc9ef76c3f1dda5280)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/83e77782f1a9303a08b1/measurements.json b/records/sd1/sweep/83e77782f1a9303a08b1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b7e9bcee68c2a50b6b99178f8fcac20541ca84df
--- /dev/null
+++ b/records/sd1/sweep/83e77782f1a9303a08b1/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "83e77782f1a9303a08b1",
+ "values": {
+  "cosine_distance": {
+   "value": 0.356053,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.483593,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.549885,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.235474,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.808,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0034,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 39.982,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0395,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/83e77782f1a9303a08b1/record.json b/records/sd1/sweep/83e77782f1a9303a08b1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5819edd6766dba29f82cb82fd7a78b1bcd1f1819
--- /dev/null
+++ b/records/sd1/sweep/83e77782f1a9303a08b1/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "83e77782f1a9303a08b1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:59Z",
+ "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.0.0.out_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/6e07d75baa28fbdb99f033e20bf6f02d10d8deccafd43d40c1cdcf737e5aeadd.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6e07d75baa28fbdb99f033e20bf6f02d10d8deccafd43d40c1cdcf737e5aeadd",
+  "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": "dd0bab6eabd1dbed0931"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/83e77782f1a9303a08b1/workflow.json b/records/sd1/sweep/83e77782f1a9303a08b1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..614a1cf04fc6ce52d92dfa18ef497ddc2b4d9e6f
--- /dev/null
+++ b/records/sd1/sweep/83e77782f1a9303a08b1/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 83e77782f1a9303a08b1)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84013ff2b63badaa9388/measurements.json b/records/sd1/sweep/84013ff2b63badaa9388/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2a0a8fbd965b6432586d7144ecf3a539640ead16
--- /dev/null
+++ b/records/sd1/sweep/84013ff2b63badaa9388/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84013ff2b63badaa9388",
+ "values": {
+  "cosine_distance": {
+   "value": 0.403532,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.45524,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.390044,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.216603,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 37.985,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00684,
+   "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": 39.336,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0297,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3062,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84013ff2b63badaa9388/record.json b/records/sd1/sweep/84013ff2b63badaa9388/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..88970ed376abfb79a40d220b34914eb392c84dfe
--- /dev/null
+++ b/records/sd1/sweep/84013ff2b63badaa9388/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84013ff2b63badaa9388",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:21Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.norm",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.lo",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/51b5be7a8c67bf9296e0c5942eb573a815ff92efb92885fad81c0381543182a0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "51b5be7a8c67bf9296e0c5942eb573a815ff92efb92885fad81c0381543182a0",
+  "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": "94ada16b476d1d6ce4da"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84013ff2b63badaa9388/workflow.json b/records/sd1/sweep/84013ff2b63badaa9388/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..39beb430c752cc34f13109f586c3ec3df5ae590c
--- /dev/null
+++ b/records/sd1/sweep/84013ff2b63badaa9388/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 84013ff2b63badaa9388)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.norm\",\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/84030a387c9b43581444/interpretations.jsonl b/records/sd1/sweep/84030a387c9b43581444/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..adf643956b5718e0c7da2f3226f279c46f5408fc
--- /dev/null
+++ b/records/sd1/sweep/84030a387c9b43581444/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "84030a387c9b43581444"}, "kind": "caption", "value": "Androgynous young man with swept dark hair, black top, pale beige vintage photo", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:06:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "03cfd5133c81fa8a"}
+{"target": {"record": "84030a387c9b43581444"}, "kind": "change", "value": "The woman becomes an androgynous young man with swept-back hair in the same pose, on a lighter beige vintage background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:06:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "54b0d650d281725a"}
+{"target": {"record": "84030a387c9b43581444"}, "kind": "keywords", "value": ["gender swap", "androgynous", "male portrait", "beige backdrop", "vintage photo"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:06:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cfb87359e5eb7674"}
+{"target": {"record": "84030a387c9b43581444"}, "kind": "effect_tags", "value": ["subject_kept", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:06:26Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "dda68b2d44ddcb16"}
diff --git a/records/sd1/sweep/84030a387c9b43581444/measurements.json b/records/sd1/sweep/84030a387c9b43581444/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..31eb4e37a41c618403ed84377fcc896995d3ff43
--- /dev/null
+++ b/records/sd1/sweep/84030a387c9b43581444/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84030a387c9b43581444",
+ "values": {
+  "cosine_distance": {
+   "value": 0.325947,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.331638,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.158331,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.178907,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.839,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00573,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 42.01,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0023,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1418,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84030a387c9b43581444/record.json b/records/sd1/sweep/84030a387c9b43581444/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f336d9084d61d289d07ae45d1d3c531c4301e488
--- /dev/null
+++ b/records/sd1/sweep/84030a387c9b43581444/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84030a387c9b43581444",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "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/b7100edf57c2a66fa9058e8e0f2e3ea2503048ca3eaa60c6009ebcf45953babd.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b7100edf57c2a66fa9058e8e0f2e3ea2503048ca3eaa60c6009ebcf45953babd",
+  "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": "4538c259bf3237f691fa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84030a387c9b43581444/workflow.json b/records/sd1/sweep/84030a387c9b43581444/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..304ffe6f029573fe24f4f55872fbe4074cc675ab
--- /dev/null
+++ b/records/sd1/sweep/84030a387c9b43581444/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 84030a387c9b43581444)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.out_layers.3\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/8403c7e7ee3a328a510b/measurements.json b/records/sd1/sweep/8403c7e7ee3a328a510b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..371f8f76b3e78709b2ca5c9fa7959ba5966ffe3a
--- /dev/null
+++ b/records/sd1/sweep/8403c7e7ee3a328a510b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8403c7e7ee3a328a510b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.371618,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.429182,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.206428,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.0969,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.329,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00516,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 43.337,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0346,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2418,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8403c7e7ee3a328a510b/record.json b/records/sd1/sweep/8403c7e7ee3a328a510b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5b897ed7b80837527f6156c1b121e663b5729d9
--- /dev/null
+++ b/records/sd1/sweep/8403c7e7ee3a328a510b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8403c7e7ee3a328a510b",
+ "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.in_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/0c470203c57493572a6ab10a09b8117f226addba5c739189d15a5f750ee80e76.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0c470203c57493572a6ab10a09b8117f226addba5c739189d15a5f750ee80e76",
+  "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": "1853243108412fc5385d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8403c7e7ee3a328a510b/workflow.json b/records/sd1/sweep/8403c7e7ee3a328a510b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8469d1ac774c823372d5bb3dce7f1aa27bc0256e
--- /dev/null
+++ b/records/sd1/sweep/8403c7e7ee3a328a510b/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 8403c7e7ee3a328a510b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.in_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/840960f9d075cf59dff5/measurements.json b/records/sd1/sweep/840960f9d075cf59dff5/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..96e493f2406f6ff6a0e242f56d771254de72db4b
--- /dev/null
+++ b/records/sd1/sweep/840960f9d075cf59dff5/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "840960f9d075cf59dff5",
+ "values": {
+  "std": {
+   "value": 58.143,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00982,
+   "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": 6.255,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.018113,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.025335,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.005639,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002098,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9958,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.313,
+   "method": "CLIP zero-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/840960f9d075cf59dff5/record.json b/records/sd1/sweep/840960f9d075cf59dff5/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1868198c68c42308a2b9c06a7a9adc8f0c815ee3
--- /dev/null
+++ b/records/sd1/sweep/840960f9d075cf59dff5/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "840960f9d075cf59dff5",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:55:07Z",
+ "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": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "mid",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_scalar",
+   0.8,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "b7539f7260075ef19def"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/840960f9d075cf59dff5/workflow.json b/records/sd1/sweep/840960f9d075cf59dff5/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4df99771668e6929aa5adf720857775c12cfae79
--- /dev/null
+++ b/records/sd1/sweep/840960f9d075cf59dff5/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 840960f9d075cf59dff5)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/840df48065465e7db299/measurements.json b/records/sd1/sweep/840df48065465e7db299/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1a338d7fb709af757c01ae3a7836c45ce49b197
--- /dev/null
+++ b/records/sd1/sweep/840df48065465e7db299/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "840df48065465e7db299",
+ "values": {
+  "cosine_distance": {
+   "value": 0.132019,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.141229,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.153387,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.189568,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.661,
+   "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": 15.521,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9715,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.521,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/840df48065465e7db299/record.json b/records/sd1/sweep/840df48065465e7db299/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..60a4bb89068860290c032143c7a177ec5b1680da
--- /dev/null
+++ b/records/sd1/sweep/840df48065465e7db299/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "840df48065465e7db299",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.emb_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/535cfcdbc0a4b11ff4bd34d8bbb5f18e7c3c7bf66cffd13d668ac4879ff010ba.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "535cfcdbc0a4b11ff4bd34d8bbb5f18e7c3c7bf66cffd13d668ac4879ff010ba",
+  "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": "6a2dc7bc43138f6fb675"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/840df48065465e7db299/workflow.json b/records/sd1/sweep/840df48065465e7db299/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7c59abb1486fc39d7f9dfcd7f842545cbc2b1ccd
--- /dev/null
+++ b/records/sd1/sweep/840df48065465e7db299/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 840df48065465e7db299)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.emb_layers.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84107abbfdc75784c54f/measurements.json b/records/sd1/sweep/84107abbfdc75784c54f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f5b3c7a504d2e33d4744b93c157316a57e05adc
--- /dev/null
+++ b/records/sd1/sweep/84107abbfdc75784c54f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84107abbfdc75784c54f",
+ "values": {
+  "std": {
+   "value": 30.197,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02729,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 26.084,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.237689,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.416088,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.131666,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.034456,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9987,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4192,
+   "method": "CLIP zero-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/84107abbfdc75784c54f/record.json b/records/sd1/sweep/84107abbfdc75784c54f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..87dfeae04af1a237ad7e05ad7b47a77e66a3e63a
--- /dev/null
+++ b/records/sd1/sweep/84107abbfdc75784c54f/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "84107abbfdc75784c54f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:08:07Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "grow",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "scale",
+   1.4,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.7, 1.3] for out.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "a93d59be3e5f02f91fec"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84107abbfdc75784c54f/workflow.json b/records/sd1/sweep/84107abbfdc75784c54f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c7dfd59d93efc543334fb649a10993da3f680f32
--- /dev/null
+++ b/records/sd1/sweep/84107abbfdc75784c54f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84107abbfdc75784c54f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/8413ad8d1d30fa169a7d/measurements.json b/records/sd1/sweep/8413ad8d1d30fa169a7d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3eacbb90c2e6ac326a13f0e8251d5e3468ad7a2
--- /dev/null
+++ b/records/sd1/sweep/8413ad8d1d30fa169a7d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8413ad8d1d30fa169a7d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.009918,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.016144,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.005009,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002214,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.339,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00566,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 3.746,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0298,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1335,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8413ad8d1d30fa169a7d/record.json b/records/sd1/sweep/8413ad8d1d30fa169a7d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4885971be205bd245df333c1ef4dd8ba7c455598
--- /dev/null
+++ b/records/sd1/sweep/8413ad8d1d30fa169a7d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8413ad8d1d30fa169a7d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ce23f94e5ba17ba04be0343b0a22465ca9be80997ffcf8e8859b58efd2e216c5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ce23f94e5ba17ba04be0343b0a22465ca9be80997ffcf8e8859b58efd2e216c5",
+  "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": "f51dcb68d7e38d95ec71"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8413ad8d1d30fa169a7d/workflow.json b/records/sd1/sweep/8413ad8d1d30fa169a7d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6a5b50ed21fec57b0beb84ea1c4caec597eafc30
--- /dev/null
+++ b/records/sd1/sweep/8413ad8d1d30fa169a7d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8413ad8d1d30fa169a7d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.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/84153300b759a227dc65/measurements.json b/records/sd1/sweep/84153300b759a227dc65/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6338badbaaa55d648b46f89b094130a08d363eb3
--- /dev/null
+++ b/records/sd1/sweep/84153300b759a227dc65/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84153300b759a227dc65",
+ "values": {
+  "std": {
+   "value": 55.483,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00361,
+   "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": 52.773,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.635871,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.731645,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.929768,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.390825,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.668,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7727,
+   "method": "CLIP zero-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/84153300b759a227dc65/record.json b/records/sd1/sweep/84153300b759a227dc65/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6697ae82fb49e7c1d6f131268c345768af9292d2
--- /dev/null
+++ b/records/sd1/sweep/84153300b759a227dc65/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "84153300b759a227dc65",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:17:12Z",
+ "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": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "boost",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "multiply",
+   2.0,
+   "middle"
+  ],
+  "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": "79ade7196a13842a3c95"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84153300b759a227dc65/workflow.json b/records/sd1/sweep/84153300b759a227dc65/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c1c9e5543dc7af7d2990596843a5c71ccac8f93
--- /dev/null
+++ b/records/sd1/sweep/84153300b759a227dc65/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84153300b759a227dc65)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 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/841a968be102d7d24721/measurements.json b/records/sd1/sweep/841a968be102d7d24721/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d657e6ccfcb6e1e7c13c32c8d8824d3e4831ae6
--- /dev/null
+++ b/records/sd1/sweep/841a968be102d7d24721/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "841a968be102d7d24721",
+ "values": {
+  "cosine_distance": {
+   "value": 0.058402,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.080262,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.028128,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.012228,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.084,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00467,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.12,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9801,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/841a968be102d7d24721/record.json b/records/sd1/sweep/841a968be102d7d24721/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..85da5bb105ffe3c8c162553a71c8bbb79446b75f
--- /dev/null
+++ b/records/sd1/sweep/841a968be102d7d24721/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "841a968be102d7d24721",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:05Z",
+ "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.10.1.transformer_blocks.0.attn1.to_q",
+   "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/4f15ae605d0565f1d3959d42e5a6412e0d2a0381034def6aa82e67b95a458277.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4f15ae605d0565f1d3959d42e5a6412e0d2a0381034def6aa82e67b95a458277",
+  "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": "6d170d3c718b868b5501"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/841a968be102d7d24721/workflow.json b/records/sd1/sweep/841a968be102d7d24721/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..027fbca7f007bf90c05c1cac4d30d17708037535
--- /dev/null
+++ b/records/sd1/sweep/841a968be102d7d24721/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 841a968be102d7d24721)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/842638706b5d25186216/measurements.json b/records/sd1/sweep/842638706b5d25186216/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aeaebd0bb3c394037bb9575fae994b9a6827e382
--- /dev/null
+++ b/records/sd1/sweep/842638706b5d25186216/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "842638706b5d25186216",
+ "values": {
+  "std": {
+   "value": 44.251,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01684,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 0.352,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 1.9e-05,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.000167,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000126,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000126,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9811,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0651,
+   "method": "CLIP zero-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/842638706b5d25186216/record.json b/records/sd1/sweep/842638706b5d25186216/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9272e958931a535c8282ceaa9b84fbfefeb70a23
--- /dev/null
+++ b/records/sd1/sweep/842638706b5d25186216/record.json
@@ -0,0 +1,86 @@
+{
+ "id": "842638706b5d25186216",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:54:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "medium",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_noise",
+   0.2,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "fbc1e6177723cba9b4bb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/842638706b5d25186216/workflow.json b/records/sd1/sweep/842638706b5d25186216/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d660996cc45e64e16114f25e9dde44fdb7241fd2
--- /dev/null
+++ b/records/sd1/sweep/842638706b5d25186216/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 842638706b5d25186216)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 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/843270a6643934ef48bb/measurements.json b/records/sd1/sweep/843270a6643934ef48bb/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..104e6321fd5dace0289173d42db5909c7d09f13a
--- /dev/null
+++ b/records/sd1/sweep/843270a6643934ef48bb/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "843270a6643934ef48bb",
+ "values": {
+  "std": {
+   "value": 49.937,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01336,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 2.472,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001814,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.007621,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.005127,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001454,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0043,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0829,
+   "method": "CLIP zero-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/843270a6643934ef48bb/record.json b/records/sd1/sweep/843270a6643934ef48bb/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6929c35e78a628896abce5095a691c67a2840b04
--- /dev/null
+++ b/records/sd1/sweep/843270a6643934ef48bb/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "843270a6643934ef48bb",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:25:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "mid",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_scalar",
+   -0.8,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "a0f6fd84100fb29f84b3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/843270a6643934ef48bb/workflow.json b/records/sd1/sweep/843270a6643934ef48bb/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c13eeb41a12c7303a745b9d7c2bddebbee507d61
--- /dev/null
+++ b/records/sd1/sweep/843270a6643934ef48bb/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 843270a6643934ef48bb)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/8436a251c175b1e7c39b/measurements.json b/records/sd1/sweep/8436a251c175b1e7c39b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..83e574591ffd1168e3e46653902656e3c0216d49
--- /dev/null
+++ b/records/sd1/sweep/8436a251c175b1e7c39b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8436a251c175b1e7c39b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.259013,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.344433,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.234845,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.099455,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.356,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00303,
+   "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": 27.702,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9633,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2975,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8436a251c175b1e7c39b/record.json b/records/sd1/sweep/8436a251c175b1e7c39b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0f9e3c57fd32988ce812fa24cc0f08e24164ef61
--- /dev/null
+++ b/records/sd1/sweep/8436a251c175b1e7c39b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8436a251c175b1e7c39b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/03068fbed148c217bcf122bdb0d1c61bf89cf1ec7aa3f3f4471d3b9087466a67.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "03068fbed148c217bcf122bdb0d1c61bf89cf1ec7aa3f3f4471d3b9087466a67",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "606de447d7445a40765b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8436a251c175b1e7c39b/workflow.json b/records/sd1/sweep/8436a251c175b1e7c39b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..64177966f3467f343a6c792151bfbfb62fb685b7
--- /dev/null
+++ b/records/sd1/sweep/8436a251c175b1e7c39b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8436a251c175b1e7c39b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84376af822034201e624/measurements.json b/records/sd1/sweep/84376af822034201e624/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..505f69347a3b6a4e5b42be45fe143575ddca8603
--- /dev/null
+++ b/records/sd1/sweep/84376af822034201e624/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84376af822034201e624",
+ "values": {
+  "cosine_distance": {
+   "value": 0.20246,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.313094,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.220897,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.07206,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 50.59,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00756,
+   "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": 26.913,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.022,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0603,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84376af822034201e624/record.json b/records/sd1/sweep/84376af822034201e624/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e0299103c17fe63e6210a779d080ebc4ea61a41
--- /dev/null
+++ b/records/sd1/sweep/84376af822034201e624/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "84376af822034201e624",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.proj_in",
+   "op": "multiply",
+   "args": {
+    "scalar": 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/169a736dae56bf79c8814a32b58fe96d2a1b2c5882bbfb406aae765502d5196f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "169a736dae56bf79c8814a32b58fe96d2a1b2c5882bbfb406aae765502d5196f",
+  "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": "32a47d104edd176c8a31"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84376af822034201e624/workflow.json b/records/sd1/sweep/84376af822034201e624/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab246eaea061f19ea5440e941c343df8beca32b3
--- /dev/null
+++ b/records/sd1/sweep/84376af822034201e624/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84376af822034201e624)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.proj_in\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/843b6d24843ea4be9394/measurements.json b/records/sd1/sweep/843b6d24843ea4be9394/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c89a1e90f28ab91682d96b4a769fbe22c31ea2a0
--- /dev/null
+++ b/records/sd1/sweep/843b6d24843ea4be9394/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "843b6d24843ea4be9394",
+ "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/843b6d24843ea4be9394/record.json b/records/sd1/sweep/843b6d24843ea4be9394/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7885e3887c2981173ea39af78527a18f22b160f7
--- /dev/null
+++ b/records/sd1/sweep/843b6d24843ea4be9394/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "843b6d24843ea4be9394",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:50Z",
+ "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": "output_blocks.10.0.h_upd",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9d6c5698e6ab1b0eca7bef52baba16379cd659798c8658f3473887ecd8415017.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9d6c5698e6ab1b0eca7bef52baba16379cd659798c8658f3473887ecd8415017",
+  "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": "ce86cab885a3bb812c57"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/843b6d24843ea4be9394/workflow.json b/records/sd1/sweep/843b6d24843ea4be9394/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6838dc5cc2f5d0980c0ffcc125db40bd3df0746d
--- /dev/null
+++ b/records/sd1/sweep/843b6d24843ea4be9394/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 843b6d24843ea4be9394)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.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/843bb62f087dcee4f816/measurements.json b/records/sd1/sweep/843bb62f087dcee4f816/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..62d315d7e365d858da07a9e75a0fcfbd1af22df7
--- /dev/null
+++ b/records/sd1/sweep/843bb62f087dcee4f816/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "843bb62f087dcee4f816",
+ "values": {
+  "std": {
+   "value": 38.827,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01647,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.564,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.455405,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.508543,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.402518,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.059172,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9407,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1516,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/843bb62f087dcee4f816/record.json b/records/sd1/sweep/843bb62f087dcee4f816/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..08113fb8a33622723be3ec63bfd6e2e3de60cd08
--- /dev/null
+++ b/records/sd1/sweep/843bb62f087dcee4f816/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "843bb62f087dcee4f816",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:14:00Z",
+ "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": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "slight",
+   "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.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": "0d4e9ff440d3645a6171"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/843bb62f087dcee4f816/workflow.json b/records/sd1/sweep/843bb62f087dcee4f816/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..35e77222dfb6b2ace9afdc39045323791acdbe84
--- /dev/null
+++ b/records/sd1/sweep/843bb62f087dcee4f816/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 843bb62f087dcee4f816)"
+  },
+  "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 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/845a15ac0a1939f3a101/measurements.json b/records/sd1/sweep/845a15ac0a1939f3a101/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6d15f976a44f468eccf62e3dbc23d49f40af50f2
--- /dev/null
+++ b/records/sd1/sweep/845a15ac0a1939f3a101/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "845a15ac0a1939f3a101",
+ "values": {
+  "cosine_distance": {
+   "value": 0.296944,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.269315,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.213595,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.136482,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.295,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00436,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 32.93,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9464,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5855,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/845a15ac0a1939f3a101/record.json b/records/sd1/sweep/845a15ac0a1939f3a101/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7369bd83e1ccd4d5ad57f48fb349d2e003db1020
--- /dev/null
+++ b/records/sd1/sweep/845a15ac0a1939f3a101/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "845a15ac0a1939f3a101",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:34Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/151d6bd2deb5038c6bebac5e07e1bce60eefab321e9b8a9de37f8f835a230b92.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "151d6bd2deb5038c6bebac5e07e1bce60eefab321e9b8a9de37f8f835a230b92",
+  "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": "d55a42718e7b2c1d72e0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/845a15ac0a1939f3a101/workflow.json b/records/sd1/sweep/845a15ac0a1939f3a101/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc5dad308661b24b8581c4a1493a0d8959df1548
--- /dev/null
+++ b/records/sd1/sweep/845a15ac0a1939f3a101/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 845a15ac0a1939f3a101)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.transformer_blocks.0.attn2.to_k\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -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/8464672a5c65a2970b06/measurements.json b/records/sd1/sweep/8464672a5c65a2970b06/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc36b70d445e73b43b0369f5d18a498a6d8dfdf5
--- /dev/null
+++ b/records/sd1/sweep/8464672a5c65a2970b06/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8464672a5c65a2970b06",
+ "values": {
+  "std": {
+   "value": 58.193,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01643,
+   "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": 27.59,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.225584,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.308614,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.167922,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.060203,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0042,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0236,
+   "method": "CLIP zero-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/8464672a5c65a2970b06/record.json b/records/sd1/sweep/8464672a5c65a2970b06/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6dcf31eb41cccc72a72c69ad5f2cec24d70d554a
--- /dev/null
+++ b/records/sd1/sweep/8464672a5c65a2970b06/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8464672a5c65a2970b06",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:45:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   2.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7fb853e51dd18c4f79d9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8464672a5c65a2970b06/workflow.json b/records/sd1/sweep/8464672a5c65a2970b06/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9a2deb47e84e839a0ffc270af04a0228ee6d3a39
--- /dev/null
+++ b/records/sd1/sweep/8464672a5c65a2970b06/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8464672a5c65a2970b06)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/8470712958a9f3043a20/measurements.json b/records/sd1/sweep/8470712958a9f3043a20/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9fc1aed6a686d1440bd9747a1050be073ad02610
--- /dev/null
+++ b/records/sd1/sweep/8470712958a9f3043a20/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8470712958a9f3043a20",
+ "values": {
+  "cosine_distance": {
+   "value": 0.19238,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.265911,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.214029,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.128811,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.334,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00313,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.596,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.048,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0861,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8470712958a9f3043a20/record.json b/records/sd1/sweep/8470712958a9f3043a20/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..02f3fa7fd793c593efb3ac3f74c1b6bf6968a768
--- /dev/null
+++ b/records/sd1/sweep/8470712958a9f3043a20/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8470712958a9f3043a20",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/9c7b95ac08621c653684ba356cbc8f1a0a97eb49fd93c81ef2af1bac0bf24d7b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9c7b95ac08621c653684ba356cbc8f1a0a97eb49fd93c81ef2af1bac0bf24d7b",
+  "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": "0f271e48cb2bee198c72"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8470712958a9f3043a20/workflow.json b/records/sd1/sweep/8470712958a9f3043a20/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..92e3c7028a0ac91e909ad053ff0e431c04ad1169
--- /dev/null
+++ b/records/sd1/sweep/8470712958a9f3043a20/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 8470712958a9f3043a20)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/8470b16641abbea7681f/interpretations.jsonl b/records/sd1/sweep/8470b16641abbea7681f/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..cc22b7ed854410e3fc16bad2d52cb2484ef8e603
--- /dev/null
+++ b/records/sd1/sweep/8470b16641abbea7681f/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8470b16641abbea7681f"}, "kind": "caption", "value": "Soft muted portrait of a pale woman with brown bob and black collared shirt", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9a1c2c3ce22d431d"}
+{"target": {"record": "8470b16641abbea7681f"}, "kind": "change", "value": "The portrait becomes a quieter, desaturated sepia-grey rendering with a smooth brown bob, collared shirt and plain beige backdrop replacing the dark framed background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "db2432eaef29ca33"}
+{"target": {"record": "8470b16641abbea7681f"}, "kind": "keywords", "value": ["muted portrait", "brown bob", "sepia", "soft rendering", "minimal background"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "42dc500b2ea7fbdd"}
+{"target": {"record": "8470b16641abbea7681f"}, "kind": "effect_tags", "value": ["subject_kept", "desaturated", "washed_out", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:13Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "78a9882b0daed5c4"}
diff --git a/records/sd1/sweep/8470b16641abbea7681f/measurements.json b/records/sd1/sweep/8470b16641abbea7681f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..959b1aeb3ce591210cd0d357731f85738f70dbc4
--- /dev/null
+++ b/records/sd1/sweep/8470b16641abbea7681f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8470b16641abbea7681f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.27993,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.364814,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.267188,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.085487,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.844,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00372,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 30.45,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0276,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.39,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8470b16641abbea7681f/record.json b/records/sd1/sweep/8470b16641abbea7681f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c038fdda263fdc8277c35dcfcd86ff0bdb8bcbc3
--- /dev/null
+++ b/records/sd1/sweep/8470b16641abbea7681f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8470b16641abbea7681f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.mid",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d6c24a0bf3d2c6414985114a5e8157122b96b01a37297810f2a991179cfadf7f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d6c24a0bf3d2c6414985114a5e8157122b96b01a37297810f2a991179cfadf7f",
+  "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": "3d507049bb5a821a18cd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8470b16641abbea7681f/workflow.json b/records/sd1/sweep/8470b16641abbea7681f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e93a3a856c4da0f21ac86a48c86c80387e99c1e3
--- /dev/null
+++ b/records/sd1/sweep/8470b16641abbea7681f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8470b16641abbea7681f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.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": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84855c122ec1c3952e46/interpretations.jsonl b/records/sd1/sweep/84855c122ec1c3952e46/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a52b4fb958ad2ed44a1e2258c80ec56edb3aab53
--- /dev/null
+++ b/records/sd1/sweep/84855c122ec1c3952e46/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "84855c122ec1c3952e46"}, "kind": "caption", "value": "Painterly portrait of a woman with darker, wavier brown bob on a pale grey-green background", "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": "2041f20d4eba8257"}
+{"target": {"record": "84855c122ec1c3952e46"}, "kind": "change", "value": "The hair becomes darker brown with looser waves, the background brightens to a pale grey-green with a worn edge, and the rendering becomes more painterly and less linear.", "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": "e970b9fd2132d634"}
+{"target": {"record": "84855c122ec1c3952e46"}, "kind": "keywords", "value": ["painterly portrait", "wavy brown hair", "grey-green background", "vintage painting"], "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": "959e31050db62535"}
+{"target": {"record": "84855c122ec1c3952e46"}, "kind": "effect_tags", "value": ["subject_kept", "painterly", "colour_shift", "style_change"], "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": "e3e6bd5d940d5302"}
diff --git a/records/sd1/sweep/84855c122ec1c3952e46/measurements.json b/records/sd1/sweep/84855c122ec1c3952e46/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..42f983eba12cca1b35757d4c459fba4506dd1061
--- /dev/null
+++ b/records/sd1/sweep/84855c122ec1c3952e46/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84855c122ec1c3952e46",
+ "values": {
+  "cosine_distance": {
+   "value": 0.27927,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.380975,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.187245,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.092892,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.425,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00646,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 33.269,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9692,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0559,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84855c122ec1c3952e46/record.json b/records/sd1/sweep/84855c122ec1c3952e46/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..00f58dc3ab57ae32a49f2f1b66c2b3233fb47b36
--- /dev/null
+++ b/records/sd1/sweep/84855c122ec1c3952e46/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "84855c122ec1c3952e46",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:19Z",
+ "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": "input_blocks.5.0.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/43f7845888ffe3df2eb05cdacf57ef23f0f8c2b45df221eb4187f4e94ef28285.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "43f7845888ffe3df2eb05cdacf57ef23f0f8c2b45df221eb4187f4e94ef28285",
+  "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": "7b2eae0855b046c301bd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84855c122ec1c3952e46/workflow.json b/records/sd1/sweep/84855c122ec1c3952e46/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..916c24229747d858d5a3e24d2ff11c5f257fb28e
--- /dev/null
+++ b/records/sd1/sweep/84855c122ec1c3952e46/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84855c122ec1c3952e46)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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/84956df07a213713d05a/measurements.json b/records/sd1/sweep/84956df07a213713d05a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d03707882efffe57ef73482818ff5114c67624a
--- /dev/null
+++ b/records/sd1/sweep/84956df07a213713d05a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84956df07a213713d05a",
+ "values": {
+  "std": {
+   "value": 34.633,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01917,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.891,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.016778,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.068754,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.031917,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.010796,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0275,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2417,
+   "method": "CLIP zero-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/84956df07a213713d05a/record.json b/records/sd1/sweep/84956df07a213713d05a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bbcf1b4ad3f99f220d7a0a2b6f30e88fbb6add9b
--- /dev/null
+++ b/records/sd1/sweep/84956df07a213713d05a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "84956df07a213713d05a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:27:52Z",
+ "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.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": "81ef55543b3b3c080780",
+  "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": "1e440b3148a69add11c3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84956df07a213713d05a/workflow.json b/records/sd1/sweep/84956df07a213713d05a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ae49652c8705ebda72c12c775b6928af8e9e8ea
--- /dev/null
+++ b/records/sd1/sweep/84956df07a213713d05a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84956df07a213713d05a)"
+  },
+  "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": "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/849c3e21896e3ef04587/measurements.json b/records/sd1/sweep/849c3e21896e3ef04587/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8d8ff54153152e317782aa3fe16efd769526748
--- /dev/null
+++ b/records/sd1/sweep/849c3e21896e3ef04587/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "849c3e21896e3ef04587",
+ "values": {
+  "std": {
+   "value": 43.365,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01239,
+   "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": 17.599,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.133348,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.324838,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.152258,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.109153,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9755,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2421,
+   "method": "CLIP zero-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/849c3e21896e3ef04587/record.json b/records/sd1/sweep/849c3e21896e3ef04587/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5c3b13690bcc04f6d97be385d25d34ade2c1b573
--- /dev/null
+++ b/records/sd1/sweep/849c3e21896e3ef04587/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "849c3e21896e3ef04587",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:29:17Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "grow",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "scale",
+   1.4,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.7, 1.3] for out.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0422d908258b1e747828"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/849c3e21896e3ef04587/workflow.json b/records/sd1/sweep/849c3e21896e3ef04587/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e3785f9942b5bb82e756cdfde8674de5758050ab
--- /dev/null
+++ b/records/sd1/sweep/849c3e21896e3ef04587/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 849c3e21896e3ef04587)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/84a09c98683f8fa735b4/measurements.json b/records/sd1/sweep/84a09c98683f8fa735b4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..55898714d0ffb8575a9c98784f51489d74e0fa34
--- /dev/null
+++ b/records/sd1/sweep/84a09c98683f8fa735b4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84a09c98683f8fa735b4",
+ "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/84a09c98683f8fa735b4/record.json b/records/sd1/sweep/84a09c98683f8fa735b4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7eddefea65f96e06de318ef7df3c91eb82d3f3e
--- /dev/null
+++ b/records/sd1/sweep/84a09c98683f8fa735b4/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84a09c98683f8fa735b4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:40Z",
+ "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.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "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/de2122ba03c27dc117c91e238a9bfb54094df95ebed06b77399d47b54e712e7f.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "de2122ba03c27dc117c91e238a9bfb54094df95ebed06b77399d47b54e712e7f",
+  "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": "3bd8afc70dee90ccfd88"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84a09c98683f8fa735b4/workflow.json b/records/sd1/sweep/84a09c98683f8fa735b4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3e286749d68d1fd5e35a4965c705c25a616725a4
--- /dev/null
+++ b/records/sd1/sweep/84a09c98683f8fa735b4/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 84a09c98683f8fa735b4)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0.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/84a32bc41485bf88f01f/measurements.json b/records/sd1/sweep/84a32bc41485bf88f01f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6814d8e9bf416f20c522fbfa272c84a2a0d04045
--- /dev/null
+++ b/records/sd1/sweep/84a32bc41485bf88f01f/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84a32bc41485bf88f01f",
+ "values": {
+  "std": {
+   "value": 43.517,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01413,
+   "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": 23.638,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.189513,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.300351,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.128377,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.025815,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0132,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.01,
+   "method": "CLIP zero-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/84a32bc41485bf88f01f/record.json b/records/sd1/sweep/84a32bc41485bf88f01f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b945a70fd12f94cafe132202d2acfda950a936d5
--- /dev/null
+++ b/records/sd1/sweep/84a32bc41485bf88f01f/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "84a32bc41485bf88f01f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:02:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": 0.8,
+    "seed": 1234
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "heavy"
+  }
+ ],
+ "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",
+   0.8,
+   "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": "8226f428a522b7e20089"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84a32bc41485bf88f01f/workflow.json b/records/sd1/sweep/84a32bc41485bf88f01f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1b6ef8594b70f2d1ffac1e1a8f5e44bd21d9a01
--- /dev/null
+++ b/records/sd1/sweep/84a32bc41485bf88f01f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84a32bc41485bf88f01f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 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/84a39ec1c5304dc81120/interpretations.jsonl b/records/sd1/sweep/84a39ec1c5304dc81120/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ffdbaa30806c87cc2703688dad116100f586d26f
--- /dev/null
+++ b/records/sd1/sweep/84a39ec1c5304dc81120/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "84a39ec1c5304dc81120"}, "kind": "caption", "value": "Neon street at night with red-orange signs and a blue sky", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7ac4b01d473d030a"}
+{"target": {"record": "84a39ec1c5304dc81120"}, "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": "2026-10-05T04:54:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "97568e99fbb83638"}
+{"target": {"record": "84a39ec1c5304dc81120"}, "kind": "keywords", "value": ["neon street", "night city", "no change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:54:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a16be59399c370a0"}
+{"target": {"record": "84a39ec1c5304dc81120"}, "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": "2026-10-05T04:54:04Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "49a9b6741c7a70e8"}
diff --git a/records/sd1/sweep/84a39ec1c5304dc81120/measurements.json b/records/sd1/sweep/84a39ec1c5304dc81120/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ebd3eeda8d0b880a861d6467d126e8d4fe2d19f5
--- /dev/null
+++ b/records/sd1/sweep/84a39ec1c5304dc81120/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84a39ec1c5304dc81120",
+ "values": {
+  "std": {
+   "value": 27.986,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02685,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.413,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.013058,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.062065,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.067971,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.018574,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9315,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3631,
+   "method": "CLIP zero-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/84a39ec1c5304dc81120/record.json b/records/sd1/sweep/84a39ec1c5304dc81120/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f32b7fc51d267b778447315d136a01bc4335248f
--- /dev/null
+++ b/records/sd1/sweep/84a39ec1c5304dc81120/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "84a39ec1c5304dc81120",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:06:41Z",
+ "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 city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "5f87952b82bb496a0a5e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84a39ec1c5304dc81120/workflow.json b/records/sd1/sweep/84a39ec1c5304dc81120/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..746d42c845566bf6b97db39eba8dd39269f716d1
--- /dev/null
+++ b/records/sd1/sweep/84a39ec1c5304dc81120/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84a39ec1c5304dc81120)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/84a78b791e8087872052/measurements.json b/records/sd1/sweep/84a78b791e8087872052/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c561a0aa9d2bd91298ba282a57a85eb9045c87fe
--- /dev/null
+++ b/records/sd1/sweep/84a78b791e8087872052/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84a78b791e8087872052",
+ "values": {
+  "cosine_distance": {
+   "value": 0.28795,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.309687,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.234866,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.106915,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.341,
+   "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": 39.425,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0257,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1457,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84a78b791e8087872052/record.json b/records/sd1/sweep/84a78b791e8087872052/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3934487f03494f93a26673cf46c947db3fb4dc9b
--- /dev/null
+++ b/records/sd1/sweep/84a78b791e8087872052/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84a78b791e8087872052",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/940c122792900e16e2fc3bd2560cf7dff67e6d8aadb27ca86a886b6f5d04abd0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "940c122792900e16e2fc3bd2560cf7dff67e6d8aadb27ca86a886b6f5d04abd0",
+  "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": "5802cd5d75014e470e03"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84a78b791e8087872052/workflow.json b/records/sd1/sweep/84a78b791e8087872052/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e33f534067b791837c69dc18a543eb3dde30057
--- /dev/null
+++ b/records/sd1/sweep/84a78b791e8087872052/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 84a78b791e8087872052)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84af78bba77215a7510d/measurements.json b/records/sd1/sweep/84af78bba77215a7510d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac6149d629f6872e4b9d92eb843f022fd5588aec
--- /dev/null
+++ b/records/sd1/sweep/84af78bba77215a7510d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84af78bba77215a7510d",
+ "values": {
+  "std": {
+   "value": 50.863,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01063,
+   "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": 3.208,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.005975,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.028582,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.02024,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.011748,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0153,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0261,
+   "method": "CLIP zero-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/84af78bba77215a7510d/record.json b/records/sd1/sweep/84af78bba77215a7510d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a419b9dc0bdd80bd1b67ea232dc39d473d1cae84
--- /dev/null
+++ b/records/sd1/sweep/84af78bba77215a7510d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "84af78bba77215a7510d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:38:36Z",
+ "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": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.lo",
+   "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": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "39e0667aa65aeca06845"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84af78bba77215a7510d/workflow.json b/records/sd1/sweep/84af78bba77215a7510d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..61d6cb33d3cd5648063f3132b43efda5e677d09e
--- /dev/null
+++ b/records/sd1/sweep/84af78bba77215a7510d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84af78bba77215a7510d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 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/84b3d4574261a8601d07/measurements.json b/records/sd1/sweep/84b3d4574261a8601d07/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bc80a4b6f2d7ac21e607d0cbe482ece20990821
--- /dev/null
+++ b/records/sd1/sweep/84b3d4574261a8601d07/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84b3d4574261a8601d07",
+ "values": {
+  "std": {
+   "value": 23.707,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02589,
+   "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": 27.594,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.273491,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.476726,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.424724,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.097389,
+   "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": 0.2919,
+   "method": "CLIP zero-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/84b3d4574261a8601d07/record.json b/records/sd1/sweep/84b3d4574261a8601d07/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d0a186002850794f8ce413de429be8f07005380
--- /dev/null
+++ b/records/sd1/sweep/84b3d4574261a8601d07/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "84b3d4574261a8601d07",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:17:30Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "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": "52f6b476d7247f108a43"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84b3d4574261a8601d07/workflow.json b/records/sd1/sweep/84b3d4574261a8601d07/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..af9318ef6f5bc35294feddd646606f47da4d8fad
--- /dev/null
+++ b/records/sd1/sweep/84b3d4574261a8601d07/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84b3d4574261a8601d07)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\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 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/84c3d02c7b7d21f59dbf/measurements.json b/records/sd1/sweep/84c3d02c7b7d21f59dbf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..745b5868f7020d25546077aeaecce3461a78e31d
--- /dev/null
+++ b/records/sd1/sweep/84c3d02c7b7d21f59dbf/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84c3d02c7b7d21f59dbf",
+ "values": {
+  "std": {
+   "value": 56.378,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00899,
+   "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": 18.856,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.117697,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.311935,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.189301,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.103301,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9951,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4685,
+   "method": "CLIP zero-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/84c3d02c7b7d21f59dbf/record.json b/records/sd1/sweep/84c3d02c7b7d21f59dbf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e2f88e8860e2dd61f15b5724f179a73d737b7d4
--- /dev/null
+++ b/records/sd1/sweep/84c3d02c7b7d21f59dbf/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "84c3d02c7b7d21f59dbf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:50:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "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": "c244b02b7723b34ac353",
+  "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": "dad60d8796d57799a7c3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84c3d02c7b7d21f59dbf/workflow.json b/records/sd1/sweep/84c3d02c7b7d21f59dbf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..db12bb3f036ba0d4c090860be04d27ee77ed6ef8
--- /dev/null
+++ b/records/sd1/sweep/84c3d02c7b7d21f59dbf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84c3d02c7b7d21f59dbf)"
+  },
+  "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": "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/84c6ac791e8176310acf/measurements.json b/records/sd1/sweep/84c6ac791e8176310acf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..72283efdbda287bcf5b75c9dfbbbd70dc5c76cab
--- /dev/null
+++ b/records/sd1/sweep/84c6ac791e8176310acf/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84c6ac791e8176310acf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.019007,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.029393,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.015195,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.005436,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.898,
+   "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": 5.895,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9862,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1649,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84c6ac791e8176310acf/record.json b/records/sd1/sweep/84c6ac791e8176310acf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d07762ba24863157d75cbdcc7792e0b6af5599c
--- /dev/null
+++ b/records/sd1/sweep/84c6ac791e8176310acf/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "84c6ac791e8176310acf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/59ec3a5fe1163eff042665b90e2a3043eec395805d691a006a24da0e53ef9760.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "59ec3a5fe1163eff042665b90e2a3043eec395805d691a006a24da0e53ef9760",
+  "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": "ce0b35489726fe416f34"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84c6ac791e8176310acf/workflow.json b/records/sd1/sweep/84c6ac791e8176310acf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e7b37136cec7e0fbe4919e3f100625b836a692e
--- /dev/null
+++ b/records/sd1/sweep/84c6ac791e8176310acf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84c6ac791e8176310acf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.transformer_blocks.0.attn2.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/84cb5dc2e5de4a5dff9c/measurements.json b/records/sd1/sweep/84cb5dc2e5de4a5dff9c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b1497575dc2109a70a67861264700c36174727e
--- /dev/null
+++ b/records/sd1/sweep/84cb5dc2e5de4a5dff9c/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84cb5dc2e5de4a5dff9c",
+ "values": {
+  "std": {
+   "value": 50.97,
+   "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": 0.983,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000654,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.003434,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004846,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000413,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9971,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0137,
+   "method": "CLIP zero-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/84cb5dc2e5de4a5dff9c/record.json b/records/sd1/sweep/84cb5dc2e5de4a5dff9c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6659eaf02f766652aa20210c73e059d8b1a9cce2
--- /dev/null
+++ b/records/sd1/sweep/84cb5dc2e5de4a5dff9c/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "84cb5dc2e5de4a5dff9c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:40:59Z",
+ "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": 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": 0.2,
+    "seed": 1234
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "medium",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_noise",
+   0.2,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84cb5dc2e5de4a5dff9c/workflow.json b/records/sd1/sweep/84cb5dc2e5de4a5dff9c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3eda5036416c69794836dc84c1e5efe574319195
--- /dev/null
+++ b/records/sd1/sweep/84cb5dc2e5de4a5dff9c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84cb5dc2e5de4a5dff9c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 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/84cd8db61cf3a3c886a9/measurements.json b/records/sd1/sweep/84cd8db61cf3a3c886a9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..844d61fbdd50ed629c4a172520cc37b9b878b184
--- /dev/null
+++ b/records/sd1/sweep/84cd8db61cf3a3c886a9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84cd8db61cf3a3c886a9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.170756,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.190125,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.101065,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.050745,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.971,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00519,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 20.012,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9789,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.109,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84cd8db61cf3a3c886a9/record.json b/records/sd1/sweep/84cd8db61cf3a3c886a9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..34ace9032461861ae9ec94c16a0e29fb29401c0d
--- /dev/null
+++ b/records/sd1/sweep/84cd8db61cf3a3c886a9/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "84cd8db61cf3a3c886a9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.10.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/8a2c6a8e1e8240f56a8b4c3d010b149662d1caf6112cc520072fc7ffc6f94e4f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "8a2c6a8e1e8240f56a8b4c3d010b149662d1caf6112cc520072fc7ffc6f94e4f",
+  "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": "f59002c273f067ec6c1e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84cd8db61cf3a3c886a9/workflow.json b/records/sd1/sweep/84cd8db61cf3a3c886a9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..19b973c1c5e73a010a18ba15f244cce0bebf814d
--- /dev/null
+++ b/records/sd1/sweep/84cd8db61cf3a3c886a9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84cd8db61cf3a3c886a9)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/84d4c126d9d4c385c0b0/measurements.json b/records/sd1/sweep/84d4c126d9d4c385c0b0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..40a1edc405cd33d9b133811ee8223852bb453f3b
--- /dev/null
+++ b/records/sd1/sweep/84d4c126d9d4c385c0b0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84d4c126d9d4c385c0b0",
+ "values": {
+  "std": {
+   "value": 49.369,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01338,
+   "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": 5.987,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.015543,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.04956,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.075549,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008617,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0263,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0785,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/84d4c126d9d4c385c0b0/record.json b/records/sd1/sweep/84d4c126d9d4c385c0b0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e1244bf61e1562aaaec6d9200142d5f3f83e32e9
--- /dev/null
+++ b/records/sd1/sweep/84d4c126d9d4c385c0b0/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "84d4c126d9d4c385c0b0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:17:59Z",
+ "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.6.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "damp",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "multiply",
+   0.5,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "8a050caa8bc5486e0e2c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84d4c126d9d4c385c0b0/workflow.json b/records/sd1/sweep/84d4c126d9d4c385c0b0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee4a9e1356b1cff25b1b35f28302b560495283a0
--- /dev/null
+++ b/records/sd1/sweep/84d4c126d9d4c385c0b0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84d4c126d9d4c385c0b0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/84d7d5cade97e6345121/measurements.json b/records/sd1/sweep/84d7d5cade97e6345121/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a6142afd7c82e7178cdfe0f025270b847e76837e
--- /dev/null
+++ b/records/sd1/sweep/84d7d5cade97e6345121/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84d7d5cade97e6345121",
+ "values": {
+  "cosine_distance": {
+   "value": 0.215586,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.25007,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.219144,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.085046,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.399,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00465,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.478,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9842,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0871,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84d7d5cade97e6345121/record.json b/records/sd1/sweep/84d7d5cade97e6345121/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec484895a0b6f39f05f55fa8d6cc690757cc9503
--- /dev/null
+++ b/records/sd1/sweep/84d7d5cade97e6345121/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "84d7d5cade97e6345121",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:21Z",
+ "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.3.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/c61b3f2d89e27723e5d3e5d25e194557435e5549b2fddbe7b68131831548ec5b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c61b3f2d89e27723e5d3e5d25e194557435e5549b2fddbe7b68131831548ec5b",
+  "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": "886e70c71268f3f6061b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84d7d5cade97e6345121/workflow.json b/records/sd1/sweep/84d7d5cade97e6345121/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f9f7bd758a344306132da33d3f5ee03c04a78e4
--- /dev/null
+++ b/records/sd1/sweep/84d7d5cade97e6345121/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84d7d5cade97e6345121)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1.transformer_blocks.0.attn1.to_out.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/84da5cae34a5990d0899/measurements.json b/records/sd1/sweep/84da5cae34a5990d0899/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..706828dca7bbcdca2bde0221592c9ece510c235a
--- /dev/null
+++ b/records/sd1/sweep/84da5cae34a5990d0899/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "84da5cae34a5990d0899",
+ "values": {
+  "cosine_distance": {
+   "value": 0.343181,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.35872,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.152183,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.078993,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.165,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00647,
+   "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": 38.432,
+   "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": 0.0571,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84da5cae34a5990d0899/record.json b/records/sd1/sweep/84da5cae34a5990d0899/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b98e42e58f7e4f620f511c055384f61120db06f9
--- /dev/null
+++ b/records/sd1/sweep/84da5cae34a5990d0899/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "84da5cae34a5990d0899",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ee85a648da2067ec6884a112e51af5acb7a6c398d677622b9df648e852af041e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ee85a648da2067ec6884a112e51af5acb7a6c398d677622b9df648e852af041e",
+  "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": "ff354cf084f7219d053f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84da5cae34a5990d0899/workflow.json b/records/sd1/sweep/84da5cae34a5990d0899/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0e15b09567c7d8521a990d9951fed170b9bfdd2
--- /dev/null
+++ b/records/sd1/sweep/84da5cae34a5990d0899/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 84da5cae34a5990d0899)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/84ec82fac870a7de09a9/measurements.json b/records/sd1/sweep/84ec82fac870a7de09a9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bb6a6c4c93fa0b858f43dfd79ddd96a23a33fdb
--- /dev/null
+++ b/records/sd1/sweep/84ec82fac870a7de09a9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84ec82fac870a7de09a9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.270137,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.295464,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.195265,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.191112,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 68.999,
+   "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": 28.113,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9628,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4055,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84ec82fac870a7de09a9/record.json b/records/sd1/sweep/84ec82fac870a7de09a9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1fe4121cc2c7f6a35d6f176bb381d0afb2232857
--- /dev/null
+++ b/records/sd1/sweep/84ec82fac870a7de09a9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84ec82fac870a7de09a9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:46Z",
+ "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": "output_blocks.8.1.transformer_blocks.0.attn2.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/796d631005ad94a0b1e8c627620ec317930e2fe797d10b75f69c844e2ab8e7e0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "796d631005ad94a0b1e8c627620ec317930e2fe797d10b75f69c844e2ab8e7e0",
+  "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": "66bd93938b46c296ba30"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84ec82fac870a7de09a9/workflow.json b/records/sd1/sweep/84ec82fac870a7de09a9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c11d962acdcfdd2324640ccd54308775b351ed11
--- /dev/null
+++ b/records/sd1/sweep/84ec82fac870a7de09a9/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 84ec82fac870a7de09a9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_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/84f089cc082c9de3b8bc/measurements.json b/records/sd1/sweep/84f089cc082c9de3b8bc/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..03c422876556216069bdb1004aca0494712143c3
--- /dev/null
+++ b/records/sd1/sweep/84f089cc082c9de3b8bc/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84f089cc082c9de3b8bc",
+ "values": {
+  "cosine_distance": {
+   "value": 0.140706,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.159215,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.153805,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.169115,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.989,
+   "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": 16.627,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9491,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3148,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84f089cc082c9de3b8bc/record.json b/records/sd1/sweep/84f089cc082c9de3b8bc/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..90c2bd389387c7ca92cc03bba48237d44a96383f
--- /dev/null
+++ b/records/sd1/sweep/84f089cc082c9de3b8bc/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84f089cc082c9de3b8bc",
+ "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.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.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/4904172d53a6e2060ec61de579b2e8995fa254743ea427926e38952058240cc2.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "4904172d53a6e2060ec61de579b2e8995fa254743ea427926e38952058240cc2",
+  "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": "432eccb4a0bb26581f83"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84f089cc082c9de3b8bc/workflow.json b/records/sd1/sweep/84f089cc082c9de3b8bc/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..93afadf273b75bba535f0dd620f414efc9769763
--- /dev/null
+++ b/records/sd1/sweep/84f089cc082c9de3b8bc/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 84f089cc082c9de3b8bc)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.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/84fd4b2888b80756aef0/measurements.json b/records/sd1/sweep/84fd4b2888b80756aef0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..431bca6b8fbc44aace31ff5c11575033672fa8d9
--- /dev/null
+++ b/records/sd1/sweep/84fd4b2888b80756aef0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "84fd4b2888b80756aef0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.318749,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.363843,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.274389,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.31814,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 80.764,
+   "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": 45.771,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.969,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1811,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84fd4b2888b80756aef0/record.json b/records/sd1/sweep/84fd4b2888b80756aef0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..07f09f9a22558ff9bfe592a31d0b62a3c24e9e99
--- /dev/null
+++ b/records/sd1/sweep/84fd4b2888b80756aef0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "84fd4b2888b80756aef0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/3cb032cc950b02483bb52b63822d23e1612ab8f2291f3cbfc423817536202ff0.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3cb032cc950b02483bb52b63822d23e1612ab8f2291f3cbfc423817536202ff0",
+  "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": "66f11e1cf374ff712014"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/84fd4b2888b80756aef0/workflow.json b/records/sd1/sweep/84fd4b2888b80756aef0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2eee534b525437601e04e7f60570a01de40c626
--- /dev/null
+++ b/records/sd1/sweep/84fd4b2888b80756aef0/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 84fd4b2888b80756aef0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85064650a47e3b388039/measurements.json b/records/sd1/sweep/85064650a47e3b388039/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4b7c5e5372ffcd7a7f24e4f99559415724124d
--- /dev/null
+++ b/records/sd1/sweep/85064650a47e3b388039/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85064650a47e3b388039",
+ "values": {
+  "cosine_distance": {
+   "value": 0.288605,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.454879,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.267368,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.076765,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 73.929,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01138,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 37.459,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0043,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0404,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85064650a47e3b388039/record.json b/records/sd1/sweep/85064650a47e3b388039/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..da838933e34f5127b206f8e4cae30ab0538d7f86
--- /dev/null
+++ b/records/sd1/sweep/85064650a47e3b388039/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85064650a47e3b388039",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.hi",
+   "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/6f9477b91074df04e467e9f737cbbcc7158e31aeaa0546eb0aae1d4a965d8b31.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6f9477b91074df04e467e9f737cbbcc7158e31aeaa0546eb0aae1d4a965d8b31",
+  "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": "15fe0a24ab9eded79a26"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85064650a47e3b388039/workflow.json b/records/sd1/sweep/85064650a47e3b388039/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e64e47f1303a0c445ea2c70723072216c1e5bfe7
--- /dev/null
+++ b/records/sd1/sweep/85064650a47e3b388039/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85064650a47e3b388039)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/850966c6fd7d14bb1f68/measurements.json b/records/sd1/sweep/850966c6fd7d14bb1f68/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ad9d142290212029e3898f45820157781886a53
--- /dev/null
+++ b/records/sd1/sweep/850966c6fd7d14bb1f68/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "850966c6fd7d14bb1f68",
+ "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/850966c6fd7d14bb1f68/record.json b/records/sd1/sweep/850966c6fd7d14bb1f68/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a0ab5aa74f3e166e7bd7df2877b1e2b1a55368a
--- /dev/null
+++ b/records/sd1/sweep/850966c6fd7d14bb1f68/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "850966c6fd7d14bb1f68",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:50Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.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": "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/c1b9e7d3cf3b922912b1bb3b3d74eeb69aa062488296f69c1e651c86c395f5d5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c1b9e7d3cf3b922912b1bb3b3d74eeb69aa062488296f69c1e651c86c395f5d5",
+  "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": "eae44b9e1532ce483a6b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/850966c6fd7d14bb1f68/workflow.json b/records/sd1/sweep/850966c6fd7d14bb1f68/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..bce253871b2dca17c97d889026ccacdea3915262
--- /dev/null
+++ b/records/sd1/sweep/850966c6fd7d14bb1f68/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 850966c6fd7d14bb1f68)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.11.0.h_upd\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/851005d6bc0bd3b8eb33/measurements.json b/records/sd1/sweep/851005d6bc0bd3b8eb33/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c9c9427eb466aeee5171bd8fc577d411fc36bfc
--- /dev/null
+++ b/records/sd1/sweep/851005d6bc0bd3b8eb33/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "851005d6bc0bd3b8eb33",
+ "values": {
+  "std": {
+   "value": 39.469,
+   "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"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.021,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.1396,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.292089,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.170333,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.035619,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "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": 0.136,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/851005d6bc0bd3b8eb33/record.json b/records/sd1/sweep/851005d6bc0bd3b8eb33/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f62a663fba7da2ed97df4421224da3f8fcec4c82
--- /dev/null
+++ b/records/sd1/sweep/851005d6bc0bd3b8eb33/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "851005d6bc0bd3b8eb33",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:30:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "damp",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "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/851005d6bc0bd3b8eb33/workflow.json b/records/sd1/sweep/851005d6bc0bd3b8eb33/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1a92c2214cc5e6222f151b2bde7da9ef0167a00f
--- /dev/null
+++ b/records/sd1/sweep/851005d6bc0bd3b8eb33/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 851005d6bc0bd3b8eb33)"
+  },
+  "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 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/8515a7ab542991edbb6a/measurements.json b/records/sd1/sweep/8515a7ab542991edbb6a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cae3f8431ef115087c0fe6f6dcf21310bc1829c
--- /dev/null
+++ b/records/sd1/sweep/8515a7ab542991edbb6a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8515a7ab542991edbb6a",
+ "values": {
+  "std": {
+   "value": 44.911,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01346,
+   "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": 15.543,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.095929,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.228814,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.151776,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.083192,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9806,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4033,
+   "method": "CLIP zero-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/8515a7ab542991edbb6a/record.json b/records/sd1/sweep/8515a7ab542991edbb6a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..91793a023b3915546d6d8bc411d8a15fef324fe3
--- /dev/null
+++ b/records/sd1/sweep/8515a7ab542991edbb6a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8515a7ab542991edbb6a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:22:18Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "grow",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "2976534844fe80be26b4",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "scale",
+   1.4,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "4487b10d90ab28cbe9f2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8515a7ab542991edbb6a/workflow.json b/records/sd1/sweep/8515a7ab542991edbb6a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5b099fb37c8729ddffeaa5e39865b4831e3f335
--- /dev/null
+++ b/records/sd1/sweep/8515a7ab542991edbb6a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8515a7ab542991edbb6a)"
+  },
+  "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 \"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/85198fd0dc51ba65114e/measurements.json b/records/sd1/sweep/85198fd0dc51ba65114e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7fa3982b7e26fbe0c6c27be249c019e6308bef3d
--- /dev/null
+++ b/records/sd1/sweep/85198fd0dc51ba65114e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85198fd0dc51ba65114e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.189626,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.379914,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.290464,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.187697,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.725,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00349,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 25.086,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9961,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1177,
+   "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/85198fd0dc51ba65114e/record.json b/records/sd1/sweep/85198fd0dc51ba65114e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5bab6cd3b05d837960a61b5102188d57172726f2
--- /dev/null
+++ b/records/sd1/sweep/85198fd0dc51ba65114e/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "85198fd0dc51ba65114e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "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/3699fe45ae1fcbe0b8a2f8f71f7f4784e3816290a2966b137dbd1402b87d0d0e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "3699fe45ae1fcbe0b8a2f8f71f7f4784e3816290a2966b137dbd1402b87d0d0e",
+  "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": "173165e61ecd13314c5c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85198fd0dc51ba65114e/workflow.json b/records/sd1/sweep/85198fd0dc51ba65114e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..590c93b195b85daed4877f5013f0870d8e97c30e
--- /dev/null
+++ b/records/sd1/sweep/85198fd0dc51ba65114e/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 85198fd0dc51ba65114e)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.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": [
+    "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/851ef9f256e3da8a1eca/measurements.json b/records/sd1/sweep/851ef9f256e3da8a1eca/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cdc4628e846317b7fbe6dd9312699a07d7061b45
--- /dev/null
+++ b/records/sd1/sweep/851ef9f256e3da8a1eca/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "851ef9f256e3da8a1eca",
+ "values": {
+  "std": {
+   "value": 27.613,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02351,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.858,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.073036,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.190293,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.071248,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.026472,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9799,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2055,
+   "method": "CLIP zero-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/851ef9f256e3da8a1eca/record.json b/records/sd1/sweep/851ef9f256e3da8a1eca/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..13280bb92d82403e013de5a79f3893c262f0f41f
--- /dev/null
+++ b/records/sd1/sweep/851ef9f256e3da8a1eca/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "851ef9f256e3da8a1eca",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:55:29Z",
+ "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": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "heavy",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.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": "bc3cbb5f8ae37bb0c5d4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/851ef9f256e3da8a1eca/workflow.json b/records/sd1/sweep/851ef9f256e3da8a1eca/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5f70de0617d9e7d99c8a1fa0ca2aec15e42c423
--- /dev/null
+++ b/records/sd1/sweep/851ef9f256e3da8a1eca/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 851ef9f256e3da8a1eca)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 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/85200dd0b012e276184a/measurements.json b/records/sd1/sweep/85200dd0b012e276184a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8a02a2cb38bf5ef5563ce4fd1442a318c8b0ba8
--- /dev/null
+++ b/records/sd1/sweep/85200dd0b012e276184a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85200dd0b012e276184a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.128148,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.187672,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.149,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.1898,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.107,
+   "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": 16.642,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9489,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5386,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85200dd0b012e276184a/record.json b/records/sd1/sweep/85200dd0b012e276184a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..941e4f8b3d96186d556d67bbde9df5703a0e2db5
--- /dev/null
+++ b/records/sd1/sweep/85200dd0b012e276184a/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "85200dd0b012e276184a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/489dd663ca193913ca2d80fa58bea8f57f226508eee59855b91dd111061b03d6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "489dd663ca193913ca2d80fa58bea8f57f226508eee59855b91dd111061b03d6",
+  "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": "c09007434d56419ec27e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85200dd0b012e276184a/workflow.json b/records/sd1/sweep/85200dd0b012e276184a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..811ff62ada525f860dee087bf960c0afc9c3ce9d
--- /dev/null
+++ b/records/sd1/sweep/85200dd0b012e276184a/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 85200dd0b012e276184a)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8525924d6de4d6e9e05d/measurements.json b/records/sd1/sweep/8525924d6de4d6e9e05d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca7f3ccee16aec5dd3bc229c94320fe34ef3c9d0
--- /dev/null
+++ b/records/sd1/sweep/8525924d6de4d6e9e05d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8525924d6de4d6e9e05d",
+ "values": {
+  "std": {
+   "value": 50.499,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01497,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.677,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.012415,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.038909,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.022725,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023162,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0282,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0267,
+   "method": "CLIP zero-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/8525924d6de4d6e9e05d/record.json b/records/sd1/sweep/8525924d6de4d6e9e05d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6cfedf4a148b3c15336895bf1cafa9532525644f
--- /dev/null
+++ b/records/sd1/sweep/8525924d6de4d6e9e05d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8525924d6de4d6e9e05d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:06:51Z",
+ "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": "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": "13216623a742ccb32c5e",
+  "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": "9e737e3ff6c5c75d0fa2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8525924d6de4d6e9e05d/workflow.json b/records/sd1/sweep/8525924d6de4d6e9e05d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..16b9ecf10d543d3a1f231b062da70cd607bfdf4f
--- /dev/null
+++ b/records/sd1/sweep/8525924d6de4d6e9e05d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8525924d6de4d6e9e05d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/85488ff2363fcffe6540/measurements.json b/records/sd1/sweep/85488ff2363fcffe6540/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f91ef42f54f504e23edf905e76d427aaf1fa28c
--- /dev/null
+++ b/records/sd1/sweep/85488ff2363fcffe6540/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85488ff2363fcffe6540",
+ "values": {
+  "cosine_distance": {
+   "value": 0.291657,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.418671,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.354664,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.113129,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 46.45,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00326,
+   "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": 33.713,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9519,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.179,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85488ff2363fcffe6540/record.json b/records/sd1/sweep/85488ff2363fcffe6540/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..6010780c4f5a7acb54740e623a08700c1e9e88cf
--- /dev/null
+++ b/records/sd1/sweep/85488ff2363fcffe6540/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85488ff2363fcffe6540",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:08Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.skip_connection",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "skip",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/ff1a3c06a0bb94d7024998e95f89198e977802a2b497c5ee2f06bed8efc7aeb0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "ff1a3c06a0bb94d7024998e95f89198e977802a2b497c5ee2f06bed8efc7aeb0",
+  "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": "07775dc7c389c9c518bd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85488ff2363fcffe6540/workflow.json b/records/sd1/sweep/85488ff2363fcffe6540/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8db60d179870d585ea27b104dfdc06eb92657f53
--- /dev/null
+++ b/records/sd1/sweep/85488ff2363fcffe6540/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85488ff2363fcffe6540)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.skip_connection\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8549ea14f6db3812652e/measurements.json b/records/sd1/sweep/8549ea14f6db3812652e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aa1fbd05c2816d690ffd4b9304fef6fd5fe94339
--- /dev/null
+++ b/records/sd1/sweep/8549ea14f6db3812652e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8549ea14f6db3812652e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.06774,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.101912,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.050951,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.021065,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.911,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00594,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.249,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0162,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1319,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8549ea14f6db3812652e/record.json b/records/sd1/sweep/8549ea14f6db3812652e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..12dc16784e91f759ff29244c2b32cc412e92793d
--- /dev/null
+++ b/records/sd1/sweep/8549ea14f6db3812652e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8549ea14f6db3812652e",
+ "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": 1.5
+   },
+   "group": "mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/9b0f431d9b586f9ecdb4cb445b6b6612fbac2d19c76dbe70d99f58feb4ed6637.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "9b0f431d9b586f9ecdb4cb445b6b6612fbac2d19c76dbe70d99f58feb4ed6637",
+  "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": "c3cfe2c9798c9dc52646"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8549ea14f6db3812652e/workflow.json b/records/sd1/sweep/8549ea14f6db3812652e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a2f515a8beb6f1c088e1ae2d4558f3ae7286c580
--- /dev/null
+++ b/records/sd1/sweep/8549ea14f6db3812652e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8549ea14f6db3812652e)"
+  },
+  "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\": 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/8551892c011c9ff6a122/measurements.json b/records/sd1/sweep/8551892c011c9ff6a122/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..08cfe58cfdf31b66d35ae67fabf5afad2724c718
--- /dev/null
+++ b/records/sd1/sweep/8551892c011c9ff6a122/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8551892c011c9ff6a122",
+ "values": {
+  "cosine_distance": {
+   "value": 0.40696,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.582743,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.51687,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.316723,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 40.893,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00452,
+   "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": 42.855,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9802,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/8551892c011c9ff6a122/record.json b/records/sd1/sweep/8551892c011c9ff6a122/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..39fc0060c54c4f298e643314ec3433b58f077698
--- /dev/null
+++ b/records/sd1/sweep/8551892c011c9ff6a122/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8551892c011c9ff6a122",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:48Z",
+ "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.1.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/2399b79282f21842424afc75d2d493350801069f60837235ecfa1508c6501b89.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "2399b79282f21842424afc75d2d493350801069f60837235ecfa1508c6501b89",
+  "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": "1e34f0583bd72229ce68"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8551892c011c9ff6a122/workflow.json b/records/sd1/sweep/8551892c011c9ff6a122/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e31f489cddfbc250c3c221881da2973666ac9ee
--- /dev/null
+++ b/records/sd1/sweep/8551892c011c9ff6a122/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 8551892c011c9ff6a122)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/8551a78cba4e4b9b22f4/measurements.json b/records/sd1/sweep/8551a78cba4e4b9b22f4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..83462125d8111f846b28fc3d2d1d85405f3d9bc2
--- /dev/null
+++ b/records/sd1/sweep/8551a78cba4e4b9b22f4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8551a78cba4e4b9b22f4",
+ "values": {
+  "std": {
+   "value": 41.07,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01781,
+   "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": 12.289,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.068271,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.138878,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.035622,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023375,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0229,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0088,
+   "method": "CLIP zero-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/8551a78cba4e4b9b22f4/record.json b/records/sd1/sweep/8551a78cba4e4b9b22f4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..588f5b29d7b26d640d50f3aa1dcf2b26a148b180
--- /dev/null
+++ b/records/sd1/sweep/8551a78cba4e4b9b22f4/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8551a78cba4e4b9b22f4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:00:14Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "heavy",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_noise",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8551a78cba4e4b9b22f4/workflow.json b/records/sd1/sweep/8551a78cba4e4b9b22f4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a9678960bfc4a58e5af9bd8ff866c443e5855ba1
--- /dev/null
+++ b/records/sd1/sweep/8551a78cba4e4b9b22f4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8551a78cba4e4b9b22f4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 1,\n \"steps_max\": 4\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/8552c9350ae02b625732/measurements.json b/records/sd1/sweep/8552c9350ae02b625732/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4090d66625a1300022027bf21bbf5b38179638b9
--- /dev/null
+++ b/records/sd1/sweep/8552c9350ae02b625732/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8552c9350ae02b625732",
+ "values": {
+  "cosine_distance": {
+   "value": 0.29715,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.269671,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.210274,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.138864,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 67.338,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00436,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 32.968,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.4982,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8552c9350ae02b625732/record.json b/records/sd1/sweep/8552c9350ae02b625732/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2c4f0478939c3c418a808edac078c8297e98f0e
--- /dev/null
+++ b/records/sd1/sweep/8552c9350ae02b625732/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8552c9350ae02b625732",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/d8bddf0571bc3f937cce2ddfa2c4d425bd75ae53b9ff319541f251a432c5e780.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d8bddf0571bc3f937cce2ddfa2c4d425bd75ae53b9ff319541f251a432c5e780",
+  "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": "c72fa8db1883da95269a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8552c9350ae02b625732/workflow.json b/records/sd1/sweep/8552c9350ae02b625732/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..29022bdf3668bc756107cca82e241ba3684dce50
--- /dev/null
+++ b/records/sd1/sweep/8552c9350ae02b625732/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 8552c9350ae02b625732)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.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/85536fbffb8a590edf69/measurements.json b/records/sd1/sweep/85536fbffb8a590edf69/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..de3c05fbd73d98c0ba522c72b4d8540755a76e77
--- /dev/null
+++ b/records/sd1/sweep/85536fbffb8a590edf69/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85536fbffb8a590edf69",
+ "values": {
+  "cosine_distance": {
+   "value": 0.012021,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.024714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.005704,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00177,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.304,
+   "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": 4.616,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0174,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1244,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85536fbffb8a590edf69/record.json b/records/sd1/sweep/85536fbffb8a590edf69/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7bbdbfbb106d50bff576867a988034f7f275e73a
--- /dev/null
+++ b/records/sd1/sweep/85536fbffb8a590edf69/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85536fbffb8a590edf69",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:16Z",
+ "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": "middle_block.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7be37df11b2cd9d155a4a7937fb734572c40b7cba402029c8a5baf8ac42729dd.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7be37df11b2cd9d155a4a7937fb734572c40b7cba402029c8a5baf8ac42729dd",
+  "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": "6f90802be6c5868b8221"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85536fbffb8a590edf69/workflow.json b/records/sd1/sweep/85536fbffb8a590edf69/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..33a64af1ddd97f856f6a8774ec08db95092e6af8
--- /dev/null
+++ b/records/sd1/sweep/85536fbffb8a590edf69/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85536fbffb8a590edf69)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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/8555286a8c13401486d6/measurements.json b/records/sd1/sweep/8555286a8c13401486d6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..06a1fc6797a448a22aee9b18261b1103210e7983
--- /dev/null
+++ b/records/sd1/sweep/8555286a8c13401486d6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8555286a8c13401486d6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.456241,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.689646,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.502816,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.378856,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.828,
+   "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": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.966,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8835,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.7153,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8555286a8c13401486d6/record.json b/records/sd1/sweep/8555286a8c13401486d6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..310add09b3048482b5f9412f765fa8a567e038df
--- /dev/null
+++ b/records/sd1/sweep/8555286a8c13401486d6/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8555286a8c13401486d6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.out_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e320bda875cda64c63c9da56f08251f00ecf5a3cbec13d777b69db283ddfe347.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e320bda875cda64c63c9da56f08251f00ecf5a3cbec13d777b69db283ddfe347",
+  "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": "11831302afda01e9b8dd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8555286a8c13401486d6/workflow.json b/records/sd1/sweep/8555286a8c13401486d6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..82ec5056c3199b9801cef07fcf476578f6dd035b
--- /dev/null
+++ b/records/sd1/sweep/8555286a8c13401486d6/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 8555286a8c13401486d6)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.out_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855741fb309f1e29c64a/measurements.json b/records/sd1/sweep/855741fb309f1e29c64a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a764462871c38c16b7a11218583482958dfec14d
--- /dev/null
+++ b/records/sd1/sweep/855741fb309f1e29c64a/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "855741fb309f1e29c64a",
+ "values": {
+  "std": {
+   "value": 46.045,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01621,
+   "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": 52.985,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.562118,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.577282,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.208344,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.073426,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0356,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0111,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/855741fb309f1e29c64a/record.json b/records/sd1/sweep/855741fb309f1e29c64a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8978ceea8bd0dfeeeb736888e501cf79fb261cd0
--- /dev/null
+++ b/records/sd1/sweep/855741fb309f1e29c64a/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "855741fb309f1e29c64a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:53:36Z",
+ "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 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": 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": [
+   "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": "e4a67d07b88bc6ea7272"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855741fb309f1e29c64a/workflow.json b/records/sd1/sweep/855741fb309f1e29c64a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f9224f546dbedc7e1893802ade5e18a3b3f4100b
--- /dev/null
+++ b/records/sd1/sweep/855741fb309f1e29c64a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 855741fb309f1e29c64a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180.0\n   }\n  }\n ],\n \"steps_min\": 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/855754ae165e22a80ca4/measurements.json b/records/sd1/sweep/855754ae165e22a80ca4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c7222e94fe476c862ccf64e49b4ee9bdbbd0a57
--- /dev/null
+++ b/records/sd1/sweep/855754ae165e22a80ca4/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "855754ae165e22a80ca4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.384544,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.675962,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.338812,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.18117,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.862,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00235,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.716,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0175,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3712,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855754ae165e22a80ca4/record.json b/records/sd1/sweep/855754ae165e22a80ca4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b85b7c4b86ea3b54ad5db91c8ceabec207884bcb
--- /dev/null
+++ b/records/sd1/sweep/855754ae165e22a80ca4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "855754ae165e22a80ca4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:23Z",
+ "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.11.0.out_layers.3",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "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/2448e58b9556c792ca4c6fdd0f216b5e548f242fa67656c76db83ddcbc884ce2.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "2448e58b9556c792ca4c6fdd0f216b5e548f242fa67656c76db83ddcbc884ce2",
+  "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": "a036ac207425665e598f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855754ae165e22a80ca4/workflow.json b/records/sd1/sweep/855754ae165e22a80ca4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..87f621dc31b75a876a217d939641a0706e5947b7
--- /dev/null
+++ b/records/sd1/sweep/855754ae165e22a80ca4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 855754ae165e22a80ca4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.0.out_layers.3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/855c31b94d51aa7ed301/measurements.json b/records/sd1/sweep/855c31b94d51aa7ed301/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..de29b45c72e23e01406b92c88d50c44dcea6037a
--- /dev/null
+++ b/records/sd1/sweep/855c31b94d51aa7ed301/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "855c31b94d51aa7ed301",
+ "values": {
+  "std": {
+   "value": 46.344,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03679,
+   "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": 52.118,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.340959,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.527847,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.149838,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.07838,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9285,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0606,
+   "method": "CLIP zero-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/855c31b94d51aa7ed301/record.json b/records/sd1/sweep/855c31b94d51aa7ed301/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3bdd139161c15ad636f262ba90d3140436eeb523
--- /dev/null
+++ b/records/sd1/sweep/855c31b94d51aa7ed301/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "855c31b94d51aa7ed301",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:15:10Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   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": "0dca5b14fe31a93fc22e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855c31b94d51aa7ed301/workflow.json b/records/sd1/sweep/855c31b94d51aa7ed301/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..65f419aaea1214f28e30afe958cbf2e538f742d5
--- /dev/null
+++ b/records/sd1/sweep/855c31b94d51aa7ed301/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 855c31b94d51aa7ed301)"
+  },
+  "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": "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/855e9fb3e0c9299b6270/measurements.json b/records/sd1/sweep/855e9fb3e0c9299b6270/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bb6cc7647e0d651a57f701358f87932d57a52dfe
--- /dev/null
+++ b/records/sd1/sweep/855e9fb3e0c9299b6270/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "855e9fb3e0c9299b6270",
+ "values": {
+  "cosine_distance": {
+   "value": 0.187291,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.305607,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.432069,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.284264,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.561,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00322,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.825,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9939,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3866,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855e9fb3e0c9299b6270/record.json b/records/sd1/sweep/855e9fb3e0c9299b6270/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..07ebe8e109c51d36bf2fd3c35babaf74553ea44b
--- /dev/null
+++ b/records/sd1/sweep/855e9fb3e0c9299b6270/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "855e9fb3e0c9299b6270",
+ "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.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/c874fd5bcec0c2dddc918519005a393f451f916eba848b7094d9169fe5044f2b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "c874fd5bcec0c2dddc918519005a393f451f916eba848b7094d9169fe5044f2b",
+  "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": "553a8b6ec2154dfd0864"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855e9fb3e0c9299b6270/workflow.json b/records/sd1/sweep/855e9fb3e0c9299b6270/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ece4431078a7ec355e1b9999090a3839b0eb18cc
--- /dev/null
+++ b/records/sd1/sweep/855e9fb3e0c9299b6270/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 855e9fb3e0c9299b6270)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    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\": 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/855f11c4a62093947cf0/measurements.json b/records/sd1/sweep/855f11c4a62093947cf0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1426766e8bb7c9c29b46daf08474904c6c156c7a
--- /dev/null
+++ b/records/sd1/sweep/855f11c4a62093947cf0/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "855f11c4a62093947cf0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.134271,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.197016,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.052279,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.030051,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.302,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00554,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 17.623,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9432,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = 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/855f11c4a62093947cf0/record.json b/records/sd1/sweep/855f11c4a62093947cf0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b52527283c5a89044614127a9b026842fce69b0a
--- /dev/null
+++ b/records/sd1/sweep/855f11c4a62093947cf0/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "855f11c4a62093947cf0",
+ "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": 0.5
+   },
+   "group": "mid",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/71acb35c500c868d5ca0b62062821d0a3592b6f85fb209774ea9e1d20861edcb.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "71acb35c500c868d5ca0b62062821d0a3592b6f85fb209774ea9e1d20861edcb",
+  "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": "c98115de0a8de46320c3"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/855f11c4a62093947cf0/workflow.json b/records/sd1/sweep/855f11c4a62093947cf0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a4c97d6de8a782db9ba2eb6ae8f64aef771ee5c
--- /dev/null
+++ b/records/sd1/sweep/855f11c4a62093947cf0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 855f11c4a62093947cf0)"
+  },
+  "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\": 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/8566eb24268f4e2859f3/measurements.json b/records/sd1/sweep/8566eb24268f4e2859f3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e7e95090598ff68e61934b0959d8939dbc7c9ee
--- /dev/null
+++ b/records/sd1/sweep/8566eb24268f4e2859f3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8566eb24268f4e2859f3",
+ "values": {
+  "std": {
+   "value": 50.882,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01518,
+   "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": 3.851,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.008821,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.029443,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.01267,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.008605,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0035,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1285,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/8566eb24268f4e2859f3/record.json b/records/sd1/sweep/8566eb24268f4e2859f3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eaf33fc253ca4f6dc5fd3c445604f4e083e9f533
--- /dev/null
+++ b/records/sd1/sweep/8566eb24268f4e2859f3/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8566eb24268f4e2859f3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:09:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "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": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "mid",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "3c323fba458c52194488"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8566eb24268f4e2859f3/workflow.json b/records/sd1/sweep/8566eb24268f4e2859f3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d7fec8f72eefff02a9e351ccb1b1723897f16d1d
--- /dev/null
+++ b/records/sd1/sweep/8566eb24268f4e2859f3/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8566eb24268f4e2859f3)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 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/8571955eca531c487c72/interpretations.jsonl b/records/sd1/sweep/8571955eca531c487c72/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..04a7d23a3ecddfce965c23a74757b94b7d1962d5
--- /dev/null
+++ b/records/sd1/sweep/8571955eca531c487c72/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8571955eca531c487c72"}, "kind": "caption", "value": "Three-quarter portrait of a young man with slicked dark hair on pale vintage card", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "351147bc50042d0d"}
+{"target": {"record": "8571955eca531c487c72"}, "kind": "change", "value": "The woman becomes a young man with slicked-back hair in nearly the same pose, lighting and muted vintage card framing.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "76bae6dbeb4bee5e"}
+{"target": {"record": "8571955eca531c487c72"}, "kind": "keywords", "value": ["young man", "gender swap", "slicked hair", "same pose", "vintage card"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f3517bdce083278a"}
+{"target": {"record": "8571955eca531c487c72"}, "kind": "effect_tags", "value": ["subject_kept", "fine_detail", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:26:12Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fd81f8758f23be41"}
diff --git a/records/sd1/sweep/8571955eca531c487c72/measurements.json b/records/sd1/sweep/8571955eca531c487c72/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..39ab49eb8bf6cc7b82962a39c16af3a7c250f59f
--- /dev/null
+++ b/records/sd1/sweep/8571955eca531c487c72/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8571955eca531c487c72",
+ "values": {
+  "cosine_distance": {
+   "value": 0.154579,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.181638,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.183177,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.172033,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.345,
+   "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": 18.9,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9518,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6354,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8571955eca531c487c72/record.json b/records/sd1/sweep/8571955eca531c487c72/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7ae4e80ad2b8b8c33ede8b5b485e941423206a18
--- /dev/null
+++ b/records/sd1/sweep/8571955eca531c487c72/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8571955eca531c487c72",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.2.out_layers.3",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "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/0e7cd16a1f31d44eb69dbc6e301a4c8ac31a2e5211de1f860151dec62f9cc7f5.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "0e7cd16a1f31d44eb69dbc6e301a4c8ac31a2e5211de1f860151dec62f9cc7f5",
+  "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": "8cba0889e57032c7f9f8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8571955eca531c487c72/workflow.json b/records/sd1/sweep/8571955eca531c487c72/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6708bfcb6de0d71552f79ab5b65efd7c0985c6f4
--- /dev/null
+++ b/records/sd1/sweep/8571955eca531c487c72/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 8571955eca531c487c72)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.out_layers.3\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85736a2fdf0bd1fd0e01/measurements.json b/records/sd1/sweep/85736a2fdf0bd1fd0e01/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..89aebe3007d41b6a5b5d82ed7d630ee560826f31
--- /dev/null
+++ b/records/sd1/sweep/85736a2fdf0bd1fd0e01/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85736a2fdf0bd1fd0e01",
+ "values": {
+  "std": {
+   "value": 35.516,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0177,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.345,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.113634,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.209993,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.107971,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.013043,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.987,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3637,
+   "method": "CLIP zero-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/85736a2fdf0bd1fd0e01/record.json b/records/sd1/sweep/85736a2fdf0bd1fd0e01/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..003a50c3d5339721a5452ab61bec98c1b50a1708
--- /dev/null
+++ b/records/sd1/sweep/85736a2fdf0bd1fd0e01/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "85736a2fdf0bd1fd0e01",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:24:38Z",
+ "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 city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "boost",
+   "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",
+   "multiply",
+   2.0,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "0ae3ef62d52158964931"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85736a2fdf0bd1fd0e01/workflow.json b/records/sd1/sweep/85736a2fdf0bd1fd0e01/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..48ce091182f15ff4bd866a1d95ad4dad495778de
--- /dev/null
+++ b/records/sd1/sweep/85736a2fdf0bd1fd0e01/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85736a2fdf0bd1fd0e01)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/8580aa3ebc8b11ddedae/interpretations.jsonl b/records/sd1/sweep/8580aa3ebc8b11ddedae/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..0903d8671164f7ee33528feee9e6c60dc9d055cc
--- /dev/null
+++ b/records/sd1/sweep/8580aa3ebc8b11ddedae/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8580aa3ebc8b11ddedae"}, "kind": "caption", "value": "Woman with a centre-parted auburn bob and black top on a rounded pale green card", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4cef0a3e36fa7a68"}
+{"target": {"record": "8580aa3ebc8b11ddedae"}, "kind": "change", "value": "The hair becomes reddish-brown with a centre parting, the pose turns more frontal, and the background is a soft sage card with rounded corners.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c8301f3829aec1c3"}
+{"target": {"record": "8580aa3ebc8b11ddedae"}, "kind": "keywords", "value": ["auburn bob", "sage background", "frontal portrait", "rounded corners"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a03dbc64756c05d1"}
+{"target": {"record": "8580aa3ebc8b11ddedae"}, "kind": "effect_tags", "value": ["subject_kept", "colour_shift", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:08:33Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "7126ede718d7b834"}
diff --git a/records/sd1/sweep/8580aa3ebc8b11ddedae/measurements.json b/records/sd1/sweep/8580aa3ebc8b11ddedae/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2b0ef4a1c88571ebffa1916a36b93135df6940db
--- /dev/null
+++ b/records/sd1/sweep/8580aa3ebc8b11ddedae/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8580aa3ebc8b11ddedae",
+ "values": {
+  "cosine_distance": {
+   "value": 0.381704,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.451302,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.217621,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.162779,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.034,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00382,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.719,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.023,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1393,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8580aa3ebc8b11ddedae/record.json b/records/sd1/sweep/8580aa3ebc8b11ddedae/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f1ec3ee2147014ca40ed54f6d66afa818ecb6592
--- /dev/null
+++ b/records/sd1/sweep/8580aa3ebc8b11ddedae/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8580aa3ebc8b11ddedae",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.norm",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.lo",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f3e1a416bd9358b5a89087480a2fcf725b3832114545a90f9bcca21e307c29d1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f3e1a416bd9358b5a89087480a2fcf725b3832114545a90f9bcca21e307c29d1",
+  "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": "26ffcbb86092c6b25fe8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8580aa3ebc8b11ddedae/workflow.json b/records/sd1/sweep/8580aa3ebc8b11ddedae/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d18be39506b8e26ea64be4f3a4c4338573235746
--- /dev/null
+++ b/records/sd1/sweep/8580aa3ebc8b11ddedae/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 8580aa3ebc8b11ddedae)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.norm\",\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/85860e50ee8484444f1d/measurements.json b/records/sd1/sweep/85860e50ee8484444f1d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..45ab30f0f54b59a7fd759df09e9191cd22bd2ff5
--- /dev/null
+++ b/records/sd1/sweep/85860e50ee8484444f1d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85860e50ee8484444f1d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.391135,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.477228,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.467662,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.160254,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 63.525,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00478,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 47.196,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0347,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0309,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85860e50ee8484444f1d/record.json b/records/sd1/sweep/85860e50ee8484444f1d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c002efc1564e9740816e447b5b62fe94daec5e1b
--- /dev/null
+++ b/records/sd1/sweep/85860e50ee8484444f1d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "85860e50ee8484444f1d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:17Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 60
+   },
+   "group": "in.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/da40f546ef83d0d852a96c99153c6cf38781bedde75b1e07511aae785beabbab.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "da40f546ef83d0d852a96c99153c6cf38781bedde75b1e07511aae785beabbab",
+  "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": "f4c7edbaf81bf9526118"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85860e50ee8484444f1d/workflow.json b/records/sd1/sweep/85860e50ee8484444f1d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c87b8233cce9e6b26b562a69bf4f02a059f3c6f1
--- /dev/null
+++ b/records/sd1/sweep/85860e50ee8484444f1d/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 85860e50ee8484444f1d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_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/858fa75ab0c43a08bd9d/measurements.json b/records/sd1/sweep/858fa75ab0c43a08bd9d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ac975f46ad728e70381007749bca9cce1e881a96
--- /dev/null
+++ b/records/sd1/sweep/858fa75ab0c43a08bd9d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "858fa75ab0c43a08bd9d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.457417,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.533796,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.349633,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.142158,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 71.713,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0049,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 49.309,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9578,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3257,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/858fa75ab0c43a08bd9d/record.json b/records/sd1/sweep/858fa75ab0c43a08bd9d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..58d5f53ad01b58a786cffaf61f731b76e27efc65
--- /dev/null
+++ b/records/sd1/sweep/858fa75ab0c43a08bd9d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "858fa75ab0c43a08bd9d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.1.0.out_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/11fc421a48ddc19d2856693cf6150f4a7881ba3f1235ee0a90650534bab37fb6.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "11fc421a48ddc19d2856693cf6150f4a7881ba3f1235ee0a90650534bab37fb6",
+  "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": "19f8b1de9d40d3e03c5a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/858fa75ab0c43a08bd9d/workflow.json b/records/sd1/sweep/858fa75ab0c43a08bd9d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..52cc494e20d26436aa39f206f303d0730dc6348d
--- /dev/null
+++ b/records/sd1/sweep/858fa75ab0c43a08bd9d/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 858fa75ab0c43a08bd9d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.1.0.out_layers.1\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/859f8bc2bd357aa70817/measurements.json b/records/sd1/sweep/859f8bc2bd357aa70817/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff07533573b4b51d73c399f1e7bdec4ad2187996
--- /dev/null
+++ b/records/sd1/sweep/859f8bc2bd357aa70817/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "859f8bc2bd357aa70817",
+ "values": {
+  "std": {
+   "value": 85.591,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00348,
+   "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": 75.826,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.722998,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.750126,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.389891,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.225866,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8978,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2186,
+   "method": "CLIP zero-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/859f8bc2bd357aa70817/record.json b/records/sd1/sweep/859f8bc2bd357aa70817/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e045a685b8a6b429e7c0b68c6a5d67dc17b1f71d
--- /dev/null
+++ b/records/sd1/sweep/859f8bc2bd357aa70817/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "859f8bc2bd357aa70817",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:15:51Z",
+ "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 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": 0.333,
+   "window": "early",
+   "bucket": "shrink",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "scale",
+   0.7,
+   "early"
+  ],
+  "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": "6b75935d6148c0ae4159"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/859f8bc2bd357aa70817/workflow.json b/records/sd1/sweep/859f8bc2bd357aa70817/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f179fd4aa06b7497dddc4781b344aef807e1aca
--- /dev/null
+++ b/records/sd1/sweep/859f8bc2bd357aa70817/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 859f8bc2bd357aa70817)"
+  },
+  "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\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85a0f4a2036ea3fc2878/measurements.json b/records/sd1/sweep/85a0f4a2036ea3fc2878/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc930881c943c30e44c6921e2bb141156a1d4325
--- /dev/null
+++ b/records/sd1/sweep/85a0f4a2036ea3fc2878/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85a0f4a2036ea3fc2878",
+ "values": {
+  "cosine_distance": {
+   "value": 0.080743,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.102615,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.067972,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02976,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.599,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00709,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.691,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0066,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.142,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85a0f4a2036ea3fc2878/record.json b/records/sd1/sweep/85a0f4a2036ea3fc2878/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9e4175b1c5464eeb36ebdfae9e566c7491b4060
--- /dev/null
+++ b/records/sd1/sweep/85a0f4a2036ea3fc2878/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85a0f4a2036ea3fc2878",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/76fce630d578dd6c1fdfcf03b4c4ad2347475fd2696bc183881dcef3da5fd562.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "76fce630d578dd6c1fdfcf03b4c4ad2347475fd2696bc183881dcef3da5fd562",
+  "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": "90d26972265fd1f3eac5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85a0f4a2036ea3fc2878/workflow.json b/records/sd1/sweep/85a0f4a2036ea3fc2878/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f96e8fb37d63ab34572b96563e24ca03b76f8540
--- /dev/null
+++ b/records/sd1/sweep/85a0f4a2036ea3fc2878/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85a0f4a2036ea3fc2878)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.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/85a179b0b9fb0f547912/measurements.json b/records/sd1/sweep/85a179b0b9fb0f547912/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d16de9aad2a9c2b7e05b2d7b2667c2c97b590cc4
--- /dev/null
+++ b/records/sd1/sweep/85a179b0b9fb0f547912/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85a179b0b9fb0f547912",
+ "values": {
+  "cosine_distance": {
+   "value": 0.04275,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.0714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.026343,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.009578,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.099,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00549,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 9.034,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0041,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1666,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85a179b0b9fb0f547912/record.json b/records/sd1/sweep/85a179b0b9fb0f547912/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..30909975d4dc64ca67e5a9c316446e26e7be110d
--- /dev/null
+++ b/records/sd1/sweep/85a179b0b9fb0f547912/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85a179b0b9fb0f547912",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:27Z",
+ "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": "middle_block.2.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/164ae2d635bf978f6a64e60fe383d15bd7a9ff659537d872032ec1bda34a8e6b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "164ae2d635bf978f6a64e60fe383d15bd7a9ff659537d872032ec1bda34a8e6b",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Sequential.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "11eb53ad60e486287fc9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85a179b0b9fb0f547912/workflow.json b/records/sd1/sweep/85a179b0b9fb0f547912/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..cc070eb287b82c9e568e2c1e421e35e6d71d1a6c
--- /dev/null
+++ b/records/sd1/sweep/85a179b0b9fb0f547912/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85a179b0b9fb0f547912)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.in_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85ad772758120053fbd6/measurements.json b/records/sd1/sweep/85ad772758120053fbd6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f40662960efd1b12b53a44e5c845b6e973f3d92a
--- /dev/null
+++ b/records/sd1/sweep/85ad772758120053fbd6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85ad772758120053fbd6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.297686,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.470301,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.263106,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.179858,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 38.966,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00445,
+   "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": 35.275,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0159,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3169,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85ad772758120053fbd6/record.json b/records/sd1/sweep/85ad772758120053fbd6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe26b1da5563f3e8bb421d01df104bd5fd44c5d1
--- /dev/null
+++ b/records/sd1/sweep/85ad772758120053fbd6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85ad772758120053fbd6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:46:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.transformer_blocks.0.attn1.to_out.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/c7dd41cdc1acb7a2481537d848e60d39180947c0fc07f3fa7f17e5f5f6983142.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "c7dd41cdc1acb7a2481537d848e60d39180947c0fc07f3fa7f17e5f5f6983142",
+  "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": "d186598f86b8ee47b736"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85ad772758120053fbd6/workflow.json b/records/sd1/sweep/85ad772758120053fbd6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..72a3de3ddb4b3c26eb69f8fd9214aded8bebaf17
--- /dev/null
+++ b/records/sd1/sweep/85ad772758120053fbd6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85ad772758120053fbd6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.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/85b1edb646ec4ce624e0/interpretations.jsonl b/records/sd1/sweep/85b1edb646ec4ce624e0/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6a8d151b775f6a196cccca9067447deb7b2d1c6f
--- /dev/null
+++ b/records/sd1/sweep/85b1edb646ec4ce624e0/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "85b1edb646ec4ce624e0"}, "kind": "caption", "value": "Chaotic mosaic of green, yellow and black blotches with a faint head shape at top", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e7eac3951b03b85b"}
+{"target": {"record": "85b1edb646ec4ce624e0"}, "kind": "change", "value": "The portrait shatters into a dense camouflage-like patchwork of saturated green, yellow and black blobs, where only a vague head outline survives.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a76f92c4914e512f"}
+{"target": {"record": "85b1edb646ec4ce624e0"}, "kind": "keywords", "value": ["camouflage blotches", "abstract mosaic", "green yellow", "dissolved portrait", "noise patches"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1a04357375badbc7"}
+{"target": {"record": "85b1edb646ec4ce624e0"}, "kind": "effect_tags", "value": ["abstract", "fragmented", "saturated", "subject_lost", "degenerate", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b909ae65b473985c"}
diff --git a/records/sd1/sweep/85b1edb646ec4ce624e0/measurements.json b/records/sd1/sweep/85b1edb646ec4ce624e0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3eef0d49f21c7f442b3ff0389e624e2754a6448
--- /dev/null
+++ b/records/sd1/sweep/85b1edb646ec4ce624e0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85b1edb646ec4ce624e0",
+ "values": {
+  "cosine_distance": {
+   "value": 0.592085,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.74283,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.82715,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.45795,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 92.585,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01426,
+   "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": 79.4,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9798,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.9498,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85b1edb646ec4ce624e0/record.json b/records/sd1/sweep/85b1edb646ec4ce624e0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b5880dfd3860945de747710562a665bae15abb58
--- /dev/null
+++ b/records/sd1/sweep/85b1edb646ec4ce624e0/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "85b1edb646ec4ce624e0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:24Z",
+ "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.1.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "in.hi",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/9feed5c5050870b06ca185eee3c74f1ce7c8dc32bd6169e8a8b542fa7f9eeccf.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "9feed5c5050870b06ca185eee3c74f1ce7c8dc32bd6169e8a8b542fa7f9eeccf",
+  "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": "2585cb8bd7c3895e7263"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85b1edb646ec4ce624e0/workflow.json b/records/sd1/sweep/85b1edb646ec4ce624e0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a0fdf43ce54dd149287d5f2a8569df98b6b82ade
--- /dev/null
+++ b/records/sd1/sweep/85b1edb646ec4ce624e0/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 85b1edb646ec4ce624e0)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.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/85b3ca2a122290e40baa/measurements.json b/records/sd1/sweep/85b3ca2a122290e40baa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7fef307316e4a28b1cb53f6338034543e5112cc7
--- /dev/null
+++ b/records/sd1/sweep/85b3ca2a122290e40baa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85b3ca2a122290e40baa",
+ "values": {
+  "std": {
+   "value": 48.119,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00638,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.297,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.075285,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.262542,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.467973,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.105828,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.1144,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0157,
+   "method": "CLIP zero-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/85b3ca2a122290e40baa/record.json b/records/sd1/sweep/85b3ca2a122290e40baa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe0ad768e5dd0b6a0cd4f149477871b0beb20bda
--- /dev/null
+++ b/records/sd1/sweep/85b3ca2a122290e40baa/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "85b3ca2a122290e40baa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:34:09Z",
+ "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": "input_blocks.5.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "quarter",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "rotate",
+   90.0,
+   "late"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 90.0,
+    "applied": 90.0,
+    "reason": "outside safe range [5.0, 45.0] for in.mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "292e437c18fb2c38394d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85b3ca2a122290e40baa/workflow.json b/records/sd1/sweep/85b3ca2a122290e40baa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3e8921d6b486c244a02b9ecab65060caeb6630e
--- /dev/null
+++ b/records/sd1/sweep/85b3ca2a122290e40baa/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85b3ca2a122290e40baa)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90.0\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/85bb7263492c073ccbda/measurements.json b/records/sd1/sweep/85bb7263492c073ccbda/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..fdaff55c2249c823c0c5448165bbd440e277e00d
--- /dev/null
+++ b/records/sd1/sweep/85bb7263492c073ccbda/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85bb7263492c073ccbda",
+ "values": {
+  "std": {
+   "value": 26.871,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02259,
+   "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": 13.458,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.074585,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.270052,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.45821,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.131527,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.906,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2707,
+   "method": "CLIP zero-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/85bb7263492c073ccbda/record.json b/records/sd1/sweep/85bb7263492c073ccbda/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5f987710a6319fbbcc8f9186225132496214e25
--- /dev/null
+++ b/records/sd1/sweep/85bb7263492c073ccbda/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "85bb7263492c073ccbda",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:07:59Z",
+ "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 city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "shrink",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "scale",
+   0.7,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "ae3ec7a5648c77857666"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85bb7263492c073ccbda/workflow.json b/records/sd1/sweep/85bb7263492c073ccbda/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..902c1d8f70691786016b098503e02307ff0bbc4d
--- /dev/null
+++ b/records/sd1/sweep/85bb7263492c073ccbda/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85bb7263492c073ccbda)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/85c1f25301d8ac66f25a/measurements.json b/records/sd1/sweep/85c1f25301d8ac66f25a/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9650e677be2eff1d03aa7876f1abba1124de7b84
--- /dev/null
+++ b/records/sd1/sweep/85c1f25301d8ac66f25a/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85c1f25301d8ac66f25a",
+ "values": {
+  "cosine_distance": {
+   "value": 0.003952,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.00851,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.003047,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001253,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.089,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00552,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 2.493,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0202,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1414,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85c1f25301d8ac66f25a/record.json b/records/sd1/sweep/85c1f25301d8ac66f25a/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..148bf30d587820840f42968289ceac28d264626b
--- /dev/null
+++ b/records/sd1/sweep/85c1f25301d8ac66f25a/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85c1f25301d8ac66f25a",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.in_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/662a0c06f1823794f770cf587503902e955ad44b76635ac0a3d374909ee95ed8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "662a0c06f1823794f770cf587503902e955ad44b76635ac0a3d374909ee95ed8",
+  "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": "20b8005e9ed812a76f77"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85c1f25301d8ac66f25a/workflow.json b/records/sd1/sweep/85c1f25301d8ac66f25a/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..889cfb43be3a082706ee151103d145bfdf85ff06
--- /dev/null
+++ b/records/sd1/sweep/85c1f25301d8ac66f25a/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85c1f25301d8ac66f25a)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.in_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85c3b659804e587c0c86/measurements.json b/records/sd1/sweep/85c3b659804e587c0c86/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4f3c3bc1a9377dab0c5c955b6f8d29c5ffadfd5b
--- /dev/null
+++ b/records/sd1/sweep/85c3b659804e587c0c86/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85c3b659804e587c0c86",
+ "values": {
+  "cosine_distance": {
+   "value": 0.347094,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.485562,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.287106,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.18198,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.587,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00414,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 54.285,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0353,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1346,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85c3b659804e587c0c86/record.json b/records/sd1/sweep/85c3b659804e587c0c86/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..440696a18aa2c7bda7356779cabfb07e1c343481
--- /dev/null
+++ b/records/sd1/sweep/85c3b659804e587c0c86/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "85c3b659804e587c0c86",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.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/6bb5c4e16ddae4c75c925dd8491c169821a8285b9df179d3dafa2e97ec32ff6e.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6bb5c4e16ddae4c75c925dd8491c169821a8285b9df179d3dafa2e97ec32ff6e",
+  "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": "d58d068db3e6a346872c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85c3b659804e587c0c86/workflow.json b/records/sd1/sweep/85c3b659804e587c0c86/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2e70ed39b7d9852a929b9f6beb3b824bfef8a3f
--- /dev/null
+++ b/records/sd1/sweep/85c3b659804e587c0c86/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 85c3b659804e587c0c86)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.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/85c91214abede97beaf8/interpretations.jsonl b/records/sd1/sweep/85c91214abede97beaf8/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..bdd0461fac20c0464036e4b0069d1686542ea6ef
--- /dev/null
+++ b/records/sd1/sweep/85c91214abede97beaf8/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "85c91214abede97beaf8"}, "kind": "caption", "value": "Close-up of neon-lit building facades covered in glowing glyph signs at night", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2c6653e03e6303f0"}
+{"target": {"record": "85c91214abede97beaf8"}, "kind": "change", "value": "The deep street view with people becomes a closer, flatter view of building fronts packed with neon signs, losing the street and the crowd.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bb51493f032e97c2"}
+{"target": {"record": "85c91214abede97beaf8"}, "kind": "keywords", "value": ["neon signs", "building facade", "night city", "zoomed in", "signage"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5aac09da388692c7"}
+{"target": {"record": "85c91214abede97beaf8"}, "kind": "effect_tags", "value": ["zoom", "new_composition", "saturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:43Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c596b4ac15355538"}
diff --git a/records/sd1/sweep/85c91214abede97beaf8/measurements.json b/records/sd1/sweep/85c91214abede97beaf8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..219abb74d5e7394c56e9feab682ee5314ba0b0de
--- /dev/null
+++ b/records/sd1/sweep/85c91214abede97beaf8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85c91214abede97beaf8",
+ "values": {
+  "std": {
+   "value": 35.887,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01627,
+   "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": 28.984,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.273854,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.4496,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.381061,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.069672,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9601,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2467,
+   "method": "CLIP zero-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/85c91214abede97beaf8/record.json b/records/sd1/sweep/85c91214abede97beaf8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..889c496bc76c78d16fd68ba311fde486c994851b
--- /dev/null
+++ b/records/sd1/sweep/85c91214abede97beaf8/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "85c91214abede97beaf8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:22:04Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": "rotate",
+   "args": {
+    "angle_degrees": 90.0
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "quarter",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "81ef55543b3b3c080780",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "rotate",
+   90.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c3f97046a73cda0a04af"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85c91214abede97beaf8/workflow.json b/records/sd1/sweep/85c91214abede97beaf8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..f8381859be7bbd2e98d0ba4c9832e73b17434e4c
--- /dev/null
+++ b/records/sd1/sweep/85c91214abede97beaf8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85c91214abede97beaf8)"
+  },
+  "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\": 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/85d3cfbb57b3460b0dd6/measurements.json b/records/sd1/sweep/85d3cfbb57b3460b0dd6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ca7701c32301890d03fa6ed21b2974e8ed2b7c55
--- /dev/null
+++ b/records/sd1/sweep/85d3cfbb57b3460b0dd6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85d3cfbb57b3460b0dd6",
+ "values": {
+  "std": {
+   "value": 41.071,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01157,
+   "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": 35.538,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.397909,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.585219,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.339446,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.171683,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9711,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0056,
+   "method": "CLIP zero-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/85d3cfbb57b3460b0dd6/record.json b/records/sd1/sweep/85d3cfbb57b3460b0dd6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec40d3cee5e7f91292f7bff70775179ebb789e3d
--- /dev/null
+++ b/records/sd1/sweep/85d3cfbb57b3460b0dd6/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "85d3cfbb57b3460b0dd6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:43:31Z",
+ "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": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.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/85d3cfbb57b3460b0dd6/workflow.json b/records/sd1/sweep/85d3cfbb57b3460b0dd6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..50c185e95a4766a404d25241e1d141166bd18e5f
--- /dev/null
+++ b/records/sd1/sweep/85d3cfbb57b3460b0dd6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85d3cfbb57b3460b0dd6)"
+  },
+  "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 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/85d599fdf5db5b8f7171/measurements.json b/records/sd1/sweep/85d599fdf5db5b8f7171/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..671ab70a0786869d119e852c81ae303873477d08
--- /dev/null
+++ b/records/sd1/sweep/85d599fdf5db5b8f7171/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85d599fdf5db5b8f7171",
+ "values": {
+  "cosine_distance": {
+   "value": 0.102066,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.14653,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.045003,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.056034,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.747,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00506,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.482,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "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": 0.0459,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85d599fdf5db5b8f7171/record.json b/records/sd1/sweep/85d599fdf5db5b8f7171/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..00caad1637e866ab486048ffd9e1ece33adfdace
--- /dev/null
+++ b/records/sd1/sweep/85d599fdf5db5b8f7171/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85d599fdf5db5b8f7171",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:10Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.1.transformer_blocks.0.attn1.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "zero"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/a335ed884116eb7e460014d08b7fbf6e43853fafd571ea0aab55a99d1d0d0c16.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a335ed884116eb7e460014d08b7fbf6e43853fafd571ea0aab55a99d1d0d0c16",
+  "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": "c463dd83c85ba5db6393"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85d599fdf5db5b8f7171/workflow.json b/records/sd1/sweep/85d599fdf5db5b8f7171/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8231265d31f09db5fd0240b252ee05f4db9e10f9
--- /dev/null
+++ b/records/sd1/sweep/85d599fdf5db5b8f7171/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85d599fdf5db5b8f7171)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.1.transformer_blocks.0.attn1.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/85da90f7581bb2febe93/measurements.json b/records/sd1/sweep/85da90f7581bb2febe93/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1523f260b5edb68bbd9a9583e06bbb4ffdc5b645
--- /dev/null
+++ b/records/sd1/sweep/85da90f7581bb2febe93/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85da90f7581bb2febe93",
+ "values": {
+  "cosine_distance": {
+   "value": 0.12713,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.187496,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038324,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.019669,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.116,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00715,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.745,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9802,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2152,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85da90f7581bb2febe93/record.json b/records/sd1/sweep/85da90f7581bb2febe93/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fb11c828c1a42ded797cd6b868f1186bac4b0b7c
--- /dev/null
+++ b/records/sd1/sweep/85da90f7581bb2febe93/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85da90f7581bb2febe93",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/51eec0b41b55df9f43fc58f9f67faa6f84cf0d8da012608f0697a12b5514cbcc.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "51eec0b41b55df9f43fc58f9f67faa6f84cf0d8da012608f0697a12b5514cbcc",
+  "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": "66e74b6832b89293bcfa"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85da90f7581bb2febe93/workflow.json b/records/sd1/sweep/85da90f7581bb2febe93/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e572c1b1c395691d311eb1337b085e499a20b10
--- /dev/null
+++ b/records/sd1/sweep/85da90f7581bb2febe93/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85da90f7581bb2febe93)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85efdd186c8fe979928d/measurements.json b/records/sd1/sweep/85efdd186c8fe979928d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d80112dc0c8bb55fa91eb816dd31bdc4954e1123
--- /dev/null
+++ b/records/sd1/sweep/85efdd186c8fe979928d/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85efdd186c8fe979928d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.094788,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.103488,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.026453,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.079402,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.601,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00341,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 13.865,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9493,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4049,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85efdd186c8fe979928d/record.json b/records/sd1/sweep/85efdd186c8fe979928d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..24c71c1a73e47f580fb9c19b82de27ae2920c7f6
--- /dev/null
+++ b/records/sd1/sweep/85efdd186c8fe979928d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "85efdd186c8fe979928d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1.transformer_blocks.0.norm3",
+   "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": "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/f794489d1002577f47c2131e68c9934466405581d70f03c621c3620d841a0b6b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f794489d1002577f47c2131e68c9934466405581d70f03c621c3620d841a0b6b",
+  "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": "aa97f863257ee31b170f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85efdd186c8fe979928d/workflow.json b/records/sd1/sweep/85efdd186c8fe979928d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..51f2fda88e61b397ed4f2ff34aab4d0da0019d29
--- /dev/null
+++ b/records/sd1/sweep/85efdd186c8fe979928d/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 85efdd186c8fe979928d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1.transformer_blocks.0.norm3\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85f683b781c4d9a36bdf/measurements.json b/records/sd1/sweep/85f683b781c4d9a36bdf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0562ed904ad1bea88b586b811e66a4633084ae01
--- /dev/null
+++ b/records/sd1/sweep/85f683b781c4d9a36bdf/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "85f683b781c4d9a36bdf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.085642,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.122265,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.066381,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023036,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.815,
+   "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": 16.33,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9616,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2106,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85f683b781c4d9a36bdf/record.json b/records/sd1/sweep/85f683b781c4d9a36bdf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..556ecddce4dc77c42cce17896ec078559626f473
--- /dev/null
+++ b/records/sd1/sweep/85f683b781c4d9a36bdf/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "85f683b781c4d9a36bdf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:47Z",
+ "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.4.1.transformer_blocks.0.attn1.to_out.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/a21680df02f66ddf9c2dbe38c580f88a00a2121a9b748c743408c6f9634b2a39.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a21680df02f66ddf9c2dbe38c580f88a00a2121a9b748c743408c6f9634b2a39",
+  "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": "ee78ae56344b1d20f68d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85f683b781c4d9a36bdf/workflow.json b/records/sd1/sweep/85f683b781c4d9a36bdf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5aecefa41958c6f4c85fb1f4c79254e935c32c6b
--- /dev/null
+++ b/records/sd1/sweep/85f683b781c4d9a36bdf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85f683b781c4d9a36bdf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.1.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/85f75a49049a94a36e4d/measurements.json b/records/sd1/sweep/85f75a49049a94a36e4d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b806601b5a54c4048656a2728561b8521c1d493
--- /dev/null
+++ b/records/sd1/sweep/85f75a49049a94a36e4d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "85f75a49049a94a36e4d",
+ "values": {
+  "std": {
+   "value": 21.682,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0265,
+   "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": 33.937,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.371462,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.619976,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.489752,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.112725,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8208,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.5215,
+   "method": "CLIP zero-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/85f75a49049a94a36e4d/record.json b/records/sd1/sweep/85f75a49049a94a36e4d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c26d10b6bc461703a0190bc99efa208119fc1a8e
--- /dev/null
+++ b/records/sd1/sweep/85f75a49049a94a36e4d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "85f75a49049a94a36e4d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:07:40Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180.0
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "half",
+   "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.mid",
+   "rotate",
+   180.0,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "d5c2aae6c8538a004ad7"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/85f75a49049a94a36e4d/workflow.json b/records/sd1/sweep/85f75a49049a94a36e4d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9899f35a5ef5b024118e5e137c99ec25cf86b1a3
--- /dev/null
+++ b/records/sd1/sweep/85f75a49049a94a36e4d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 85f75a49049a94a36e4d)"
+  },
+  "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\": 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/8600f3e0e8c5f09ef382/measurements.json b/records/sd1/sweep/8600f3e0e8c5f09ef382/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..420919a443fbec7d132d57a9f1cccf6cbbcf8ef3
--- /dev/null
+++ b/records/sd1/sweep/8600f3e0e8c5f09ef382/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8600f3e0e8c5f09ef382",
+ "values": {
+  "std": {
+   "value": 56.408,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01432,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.296,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.183114,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.265,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.455121,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.033253,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0378,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1107,
+   "method": "CLIP zero-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/8600f3e0e8c5f09ef382/record.json b/records/sd1/sweep/8600f3e0e8c5f09ef382/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5083fd97d01708bac8b9dbe26555e8aeed2d86ca
--- /dev/null
+++ b/records/sd1/sweep/8600f3e0e8c5f09ef382/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8600f3e0e8c5f09ef382",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:13:38Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.lo",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.lo",
+   "multiply",
+   2.0,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "94e90472c466332bee7e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8600f3e0e8c5f09ef382/workflow.json b/records/sd1/sweep/8600f3e0e8c5f09ef382/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6bd8dbdfbb5d04e63409e14bdb50ec48018bbbb5
--- /dev/null
+++ b/records/sd1/sweep/8600f3e0e8c5f09ef382/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8600f3e0e8c5f09ef382)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2.0\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/8605fe93349bf5f930b6/measurements.json b/records/sd1/sweep/8605fe93349bf5f930b6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5367472579c259ac8cc616126453df2a208d3205
--- /dev/null
+++ b/records/sd1/sweep/8605fe93349bf5f930b6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8605fe93349bf5f930b6",
+ "values": {
+  "std": {
+   "value": 37.749,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01432,
+   "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": 1.685,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001085,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.003978,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.001716,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000738,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9974,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0134,
+   "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/8605fe93349bf5f930b6/record.json b/records/sd1/sweep/8605fe93349bf5f930b6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..935bb982f5a12bf4df1c51c239d7dae9c7211e38
--- /dev/null
+++ b/records/sd1/sweep/8605fe93349bf5f930b6/record.json
@@ -0,0 +1,79 @@
+{
+ "id": "8605fe93349bf5f930b6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:16:15Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "medium",
+   "steps_min": 4,
+   "steps_max": 5
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.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": "824dbf29e87dbbdbbbe5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8605fe93349bf5f930b6/workflow.json b/records/sd1/sweep/8605fe93349bf5f930b6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..087366c8f07b447e3e26f754bc9ab32e95bb8d05
--- /dev/null
+++ b/records/sd1/sweep/8605fe93349bf5f930b6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8605fe93349bf5f930b6)"
+  },
+  "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\": 0.2,\n    \"seed\": 1234\n   }\n  }\n ],\n \"steps_min\": 4,\n \"steps_max\": 5\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a lighthouse on a rocky coast at dusk, oil painting"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/861be3f87b59d1171830/measurements.json b/records/sd1/sweep/861be3f87b59d1171830/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ae91553871aa4fdd7e5f6c93ef4b290871e3c1c6
--- /dev/null
+++ b/records/sd1/sweep/861be3f87b59d1171830/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "861be3f87b59d1171830",
+ "values": {
+  "cosine_distance": {
+   "value": 0.09757,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.095197,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.035288,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.076354,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.557,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00592,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.699,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9646,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1557,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/861be3f87b59d1171830/record.json b/records/sd1/sweep/861be3f87b59d1171830/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..682f6e92d00f76f92929d2a9d27354cba963f48b
--- /dev/null
+++ b/records/sd1/sweep/861be3f87b59d1171830/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "861be3f87b59d1171830",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:12Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.2.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/d644130dd9090401b7564788f4e9c3043cc8cdbbb08693d67d5c76d68dbe3375.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "d644130dd9090401b7564788f4e9c3043cc8cdbbb08693d67d5c76d68dbe3375",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "68cbd20be910a5efde15"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/861be3f87b59d1171830/workflow.json b/records/sd1/sweep/861be3f87b59d1171830/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a0887393c0550eeb030197af4d3f028a0f0be9a
--- /dev/null
+++ b/records/sd1/sweep/861be3f87b59d1171830/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 861be3f87b59d1171830)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.2.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "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/862517edf9e19ba1fb16/measurements.json b/records/sd1/sweep/862517edf9e19ba1fb16/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1a78a22862559cfc98f962a84e3541cfa576d37
--- /dev/null
+++ b/records/sd1/sweep/862517edf9e19ba1fb16/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "862517edf9e19ba1fb16",
+ "values": {
+  "cosine_distance": {
+   "value": 0.169253,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.273982,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.377265,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.254884,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.994,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00447,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.3,
+   "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": 0.8149,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/862517edf9e19ba1fb16/record.json b/records/sd1/sweep/862517edf9e19ba1fb16/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4674e149578af734c256f04d16f0ab50acbc5154
--- /dev/null
+++ b/records/sd1/sweep/862517edf9e19ba1fb16/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "862517edf9e19ba1fb16",
+ "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": 180
+   },
+   "group": "out.mid",
+   "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/6eb5522045e036b6b9a081ffc1ce5dec7add751aba3e4215d3a753f9415167c4.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6eb5522045e036b6b9a081ffc1ce5dec7add751aba3e4215d3a753f9415167c4",
+  "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": "4bd557fc0b586aae1ae6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/862517edf9e19ba1fb16/workflow.json b/records/sd1/sweep/862517edf9e19ba1fb16/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..146759cf34e1ea9f50353d642fef8df62d4635e6
--- /dev/null
+++ b/records/sd1/sweep/862517edf9e19ba1fb16/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 862517edf9e19ba1fb16)"
+  },
+  "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\": 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/862f2ecf27a173491ed8/measurements.json b/records/sd1/sweep/862f2ecf27a173491ed8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e4f1da59d34505538dff4065363a7a15adb7ec0
--- /dev/null
+++ b/records/sd1/sweep/862f2ecf27a173491ed8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "862f2ecf27a173491ed8",
+ "values": {
+  "std": {
+   "value": 46.713,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01307,
+   "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": 4.009,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.006497,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.020266,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.007705,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.003747,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9845,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2619,
+   "method": "CLIP zero-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/862f2ecf27a173491ed8/record.json b/records/sd1/sweep/862f2ecf27a173491ed8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..44fdb34cf133993047e9dbf7c31b3022991a6f53
--- /dev/null
+++ b/records/sd1/sweep/862f2ecf27a173491ed8/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "862f2ecf27a173491ed8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:24:39Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.8,
+    "seed": 7
+   },
+   "group": "mid",
+   "kind": "attn_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": [
+   "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": "36d7da2a84280d58a869"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/862f2ecf27a173491ed8/workflow.json b/records/sd1/sweep/862f2ecf27a173491ed8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ea37e6be4ad79ec5c2253a1a175111a76340f457
--- /dev/null
+++ b/records/sd1/sweep/862f2ecf27a173491ed8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 862f2ecf27a173491ed8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.8,\n    \"seed\": 7\n   }\n  }\n ]\n}",
+   "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/862f53ca702fb6e4dc53/interpretations.jsonl b/records/sd1/sweep/862f53ca702fb6e4dc53/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6261c325cbb20782f21108b3a4ecc6f1a3746955
--- /dev/null
+++ b/records/sd1/sweep/862f53ca702fb6e4dc53/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "862f53ca702fb6e4dc53"}, "kind": "caption", "value": "Soft, muted vintage portrait of a woman with brown bob hair on sepia background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "011d53e240923535"}
+{"target": {"record": "862f53ca702fb6e4dc53"}, "kind": "change", "value": "The image becomes a soft, blurry, low-contrast vintage painting in sepia-olive tones. She has a neat brown bob, and the grey background turns beige.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ae7fccab4d47a1fe"}
+{"target": {"record": "862f53ca702fb6e4dc53"}, "kind": "keywords", "value": ["vintage", "sepia", "soft focus", "bob haircut", "muted"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "6887f9b1bbdec2f1"}
+{"target": {"record": "862f53ca702fb6e4dc53"}, "kind": "effect_tags", "value": ["blur", "washed_out", "colour_shift", "subject_kept", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:15:40Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9639082379f94d78"}
diff --git a/records/sd1/sweep/862f53ca702fb6e4dc53/measurements.json b/records/sd1/sweep/862f53ca702fb6e4dc53/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8e1ca1150c5d2657048146645b5ed12f81b02ec7
--- /dev/null
+++ b/records/sd1/sweep/862f53ca702fb6e4dc53/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "862f53ca702fb6e4dc53",
+ "values": {
+  "cosine_distance": {
+   "value": 0.391035,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.486714,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.173936,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.081295,
+   "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": 0.00229,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 40.995,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.993,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2493,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/862f53ca702fb6e4dc53/record.json b/records/sd1/sweep/862f53ca702fb6e4dc53/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bd58575ef93654b87e3c094fc8b7e79cb9792450
--- /dev/null
+++ b/records/sd1/sweep/862f53ca702fb6e4dc53/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "862f53ca702fb6e4dc53",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:47:42Z",
+ "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": 42,
+  "sampler": "lcm",
+  "scheduler": "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": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e335f7cf5c19a22f3b24c5658e25d7351922342d720316ca0ffdc2767e089f5b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e335f7cf5c19a22f3b24c5658e25d7351922342d720316ca0ffdc2767e089f5b",
+  "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": "66f37a53e16e81dec2e5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/862f53ca702fb6e4dc53/workflow.json b/records/sd1/sweep/862f53ca702fb6e4dc53/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..181b31d246ab9dad96bb5242cbc0fdfe74396f39
--- /dev/null
+++ b/records/sd1/sweep/862f53ca702fb6e4dc53/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 862f53ca702fb6e4dc53)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.in_layers.1\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8634c73b9538dd377a2c/measurements.json b/records/sd1/sweep/8634c73b9538dd377a2c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ab9ae14a89d97f54666caee2e1dcd6b6ba90c770
--- /dev/null
+++ b/records/sd1/sweep/8634c73b9538dd377a2c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8634c73b9538dd377a2c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.158826,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.24236,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.08572,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.03074,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.449,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00463,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.775,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9326,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2725,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8634c73b9538dd377a2c/record.json b/records/sd1/sweep/8634c73b9538dd377a2c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..904afc7de734dc73b4de8e3db4c436c1550892a0
--- /dev/null
+++ b/records/sd1/sweep/8634c73b9538dd377a2c/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8634c73b9538dd377a2c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/3c57a2a6e099461f5ed4e767dcb5145d55fab9db8590ef31aba4ab8231acab88.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3c57a2a6e099461f5ed4e767dcb5145d55fab9db8590ef31aba4ab8231acab88",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.SpatialTransformer.Conv2d",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative",
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "b46bcc213cf495afba70"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8634c73b9538dd377a2c/workflow.json b/records/sd1/sweep/8634c73b9538dd377a2c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e24bd73c09ecb794f21fe73e10a57d5a6ed80bf0
--- /dev/null
+++ b/records/sd1/sweep/8634c73b9538dd377a2c/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8634c73b9538dd377a2c)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 90\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8638f2c018c22e1489f8/measurements.json b/records/sd1/sweep/8638f2c018c22e1489f8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..50ad63c3870b2f67ed3a35ec99be3b59072f84fa
--- /dev/null
+++ b/records/sd1/sweep/8638f2c018c22e1489f8/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8638f2c018c22e1489f8",
+ "values": {
+  "cosine_distance": {
+   "value": 0.058934,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.138642,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.070294,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.024119,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.131,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00404,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.907,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0077,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0906,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8638f2c018c22e1489f8/record.json b/records/sd1/sweep/8638f2c018c22e1489f8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8eb174eb886fbb4999b924aabbd770b5284ba60
--- /dev/null
+++ b/records/sd1/sweep/8638f2c018c22e1489f8/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8638f2c018c22e1489f8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.1.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.hi",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/7af1d69638e7e50b8f53465d39f01e89573a1a8e6a5347369bdba967a9c35297.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "7af1d69638e7e50b8f53465d39f01e89573a1a8e6a5347369bdba967a9c35297",
+  "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": "87ec90419b8061e5b2c8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8638f2c018c22e1489f8/workflow.json b/records/sd1/sweep/8638f2c018c22e1489f8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c87d7f8ae89d6dd5743edf98caac6727bfa87c20
--- /dev/null
+++ b/records/sd1/sweep/8638f2c018c22e1489f8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8638f2c018c22e1489f8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.1.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": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/863a3ab17cd2a46cdb51/measurements.json b/records/sd1/sweep/863a3ab17cd2a46cdb51/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..542dbdfd662481c69df52cf2971ab7ad031bda8c
--- /dev/null
+++ b/records/sd1/sweep/863a3ab17cd2a46cdb51/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "863a3ab17cd2a46cdb51",
+ "values": {
+  "std": {
+   "value": 50.379,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01022,
+   "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": 3.505,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.010545,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.051137,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.042305,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.004957,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9952,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0116,
+   "method": "CLIP zero-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/863a3ab17cd2a46cdb51/record.json b/records/sd1/sweep/863a3ab17cd2a46cdb51/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..28742fdb16a7e4425ab6f553498a333010c410f6
--- /dev/null
+++ b/records/sd1/sweep/863a3ab17cd2a46cdb51/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "863a3ab17cd2a46cdb51",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:49:24Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 1234
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "medium",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "27f5baccf7c0d4f1de23",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "add_noise",
+   0.2,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/863a3ab17cd2a46cdb51/workflow.json b/records/sd1/sweep/863a3ab17cd2a46cdb51/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c2293b434c3b41fe7380441a86bce1b77d6a8a6e
--- /dev/null
+++ b/records/sd1/sweep/863a3ab17cd2a46cdb51/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 863a3ab17cd2a46cdb51)"
+  },
+  "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\": 0.2,\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": "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/863f56b0ae7b19aec653/interpretations.jsonl b/records/sd1/sweep/863f56b0ae7b19aec653/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..b6266d591e4651a62c2f0d1fb283a1db079fcac6
--- /dev/null
+++ b/records/sd1/sweep/863f56b0ae7b19aec653/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "863f56b0ae7b19aec653"}, "kind": "caption", "value": "Vivid painted lighthouse on a rocky shore under a blazing red sunset and a deep blue sea", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "a10b042b405588ac"}
+{"target": {"record": "863f56b0ae7b19aec653"}, "kind": "change", "value": "The palette becomes much more intense and dramatic: there is a fiery red-orange sky, deep navy water and stronger contrast, the lighthouse is larger and closer, and the image looks more like a poster-style painting.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "662aa25ab2bb29ca"}
+{"target": {"record": "863f56b0ae7b19aec653"}, "kind": "keywords", "value": ["lighthouse painting", "red sunset", "vivid colours", "dramatic seascape", "poster style"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ced9460ea2b6dea1"}
+{"target": {"record": "863f56b0ae7b19aec653"}, "kind": "effect_tags", "value": ["saturated", "high_contrast", "colour_shift", "poster", "style_change", "subject_kept", "zoom"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:50:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fde8bbc94c50cb72"}
diff --git a/records/sd1/sweep/863f56b0ae7b19aec653/measurements.json b/records/sd1/sweep/863f56b0ae7b19aec653/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b80a5556098f8abc82da88e2b03c5d117b906fc6
--- /dev/null
+++ b/records/sd1/sweep/863f56b0ae7b19aec653/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "863f56b0ae7b19aec653",
+ "values": {
+  "std": {
+   "value": 46.226,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.03106,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.59,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.258911,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.43908,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.14563,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.071429,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9524,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0675,
+   "method": "CLIP zero-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/863f56b0ae7b19aec653/record.json b/records/sd1/sweep/863f56b0ae7b19aec653/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f0c1e95a71936428ad7c089468b735598875e625
--- /dev/null
+++ b/records/sd1/sweep/863f56b0ae7b19aec653/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "863f56b0ae7b19aec653",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T01:59:01Z",
+ "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": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "a77bd19c2cc87d81e6d0",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   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": "2cf9b09911b8d5195f6a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/863f56b0ae7b19aec653/workflow.json b/records/sd1/sweep/863f56b0ae7b19aec653/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1ebd042a782cb6a0649e33812b1c6f6d82316d7a
--- /dev/null
+++ b/records/sd1/sweep/863f56b0ae7b19aec653/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 863f56b0ae7b19aec653)"
+  },
+  "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": "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/8640871221845b1603ab/measurements.json b/records/sd1/sweep/8640871221845b1603ab/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0720a4324c714e58a35e2bc6b6b793a52d095b88
--- /dev/null
+++ b/records/sd1/sweep/8640871221845b1603ab/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8640871221845b1603ab",
+ "values": {
+  "std": {
+   "value": 29.966,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02916,
+   "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": 1.57,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.000965,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.004118,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.005617,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00105,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9954,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1817,
+   "method": "CLIP zero-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/8640871221845b1603ab/record.json b/records/sd1/sweep/8640871221845b1603ab/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a18ca1540deee62e0049779669b02f120838394
--- /dev/null
+++ b/records/sd1/sweep/8640871221845b1603ab/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "8640871221845b1603ab",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:02:22Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "middle_block.1",
+   "op": "add_noise",
+   "args": {
+    "noise_std": 0.2,
+    "seed": 7
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "medium",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_noise",
+   0.2,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.2,
+    "applied": 0.2,
+    "reason": "outside safe range [0.3, 3.0] for mid (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8640871221845b1603ab/workflow.json b/records/sd1/sweep/8640871221845b1603ab/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e41f13c0a8c19700c7cae79684c5e59c7097cef4
--- /dev/null
+++ b/records/sd1/sweep/8640871221845b1603ab/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8640871221845b1603ab)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_noise\",\n   \"module_args\": {\n    \"noise_std\": 0.2,\n    \"seed\": 7\n   }\n  }\n ],\n \"steps_min\": 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/864570d236512dced107/measurements.json b/records/sd1/sweep/864570d236512dced107/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..0a2282be6f50c77046020b560dea6d1b1f370e6c
--- /dev/null
+++ b/records/sd1/sweep/864570d236512dced107/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "864570d236512dced107",
+ "values": {
+  "std": {
+   "value": 52.33,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01352,
+   "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": 13.808,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.084297,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.155493,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.21938,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027587,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9861,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2204,
+   "method": "CLIP zero-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/864570d236512dced107/record.json b/records/sd1/sweep/864570d236512dced107/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c1a5453a3527b76d25d03cb534051ba05f6bd7fc
--- /dev/null
+++ b/records/sd1/sweep/864570d236512dced107/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "864570d236512dced107",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:12:53Z",
+ "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": "middle_block.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "mid",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "mid",
+   "add_scalar",
+   0.8,
+   "all"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "37e2322404528e4e8893"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/864570d236512dced107/workflow.json b/records/sd1/sweep/864570d236512dced107/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b67d21f450f5b232d40d9b74ab10f6b55938c5f4
--- /dev/null
+++ b/records/sd1/sweep/864570d236512dced107/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 864570d236512dced107)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/8645e3496c88e7b33fb6/measurements.json b/records/sd1/sweep/8645e3496c88e7b33fb6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..01b86a74701d3015e78828b673cca4a92be6c066
--- /dev/null
+++ b/records/sd1/sweep/8645e3496c88e7b33fb6/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8645e3496c88e7b33fb6",
+ "values": {
+  "std": {
+   "value": 31.618,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02143,
+   "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": 6.116,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.011638,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.035081,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.01713,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002953,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0049,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1945,
+   "method": "CLIP zero-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/8645e3496c88e7b33fb6/record.json b/records/sd1/sweep/8645e3496c88e7b33fb6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0dce650e376ff2dd755451df1a0a1f6cbbbf6231
--- /dev/null
+++ b/records/sd1/sweep/8645e3496c88e7b33fb6/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "8645e3496c88e7b33fb6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T23:27:06Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.1",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "out.lo",
+   "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": [
+   "out.lo",
+   "add_scalar",
+   0.8,
+   "middle"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8645e3496c88e7b33fb6/workflow.json b/records/sd1/sweep/8645e3496c88e7b33fb6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7631344575454e74c3c32d3aa7829e36a8562b5d
--- /dev/null
+++ b/records/sd1/sweep/8645e3496c88e7b33fb6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8645e3496c88e7b33fb6)"
+  },
+  "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 \"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/8648a92fc8b6f857dc08/measurements.json b/records/sd1/sweep/8648a92fc8b6f857dc08/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bdbe9dd29f9a9d8bbb3695dca6772491a4a023ff
--- /dev/null
+++ b/records/sd1/sweep/8648a92fc8b6f857dc08/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8648a92fc8b6f857dc08",
+ "values": {
+  "cosine_distance": {
+   "value": 0.07312,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.123447,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.036202,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.024906,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 61.899,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00603,
+   "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": 13.276,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0238,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1499,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8648a92fc8b6f857dc08/record.json b/records/sd1/sweep/8648a92fc8b6f857dc08/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..56de18995bb80e3bfaf4bd7fdd7018228114faa8
--- /dev/null
+++ b/records/sd1/sweep/8648a92fc8b6f857dc08/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8648a92fc8b6f857dc08",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:33Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "out.mid",
+   "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/76c26a912cb5402c1ed2b738edf6efc66f778fddded0f5eb707ff8153a1d42a7.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "76c26a912cb5402c1ed2b738edf6efc66f778fddded0f5eb707ff8153a1d42a7",
+  "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": "6e949e90a3e0f0660ba0"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8648a92fc8b6f857dc08/workflow.json b/records/sd1/sweep/8648a92fc8b6f857dc08/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6ab83547fecd73ef259e8a4b7474ad0bd9c6fe6c
--- /dev/null
+++ b/records/sd1/sweep/8648a92fc8b6f857dc08/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8648a92fc8b6f857dc08)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0.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/864f117dca22e1b1077b/interpretations.jsonl b/records/sd1/sweep/864f117dca22e1b1077b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d3bca238cd0499398abc14a6e43b2679c07e0771
--- /dev/null
+++ b/records/sd1/sweep/864f117dca22e1b1077b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "864f117dca22e1b1077b"}, "kind": "caption", "value": "Elderly bearded man in grey beret before cloudy sea, softer cooler portrait", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "34d2e6ce0cc5121a"}
+{"target": {"record": "864f117dca22e1b1077b"}, "kind": "change", "value": "Same old man portrait with a slightly different beret and face, cooler, greyer tones and softer skin texture; a modest variation.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "69f05aed47c8f1d7"}
+{"target": {"record": "864f117dca22e1b1077b"}, "kind": "keywords", "value": ["old man portrait", "fisherman", "beret", "seaside portrait", "cool tones"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "15f526839e29c574"}
+{"target": {"record": "864f117dca22e1b1077b"}, "kind": "effect_tags", "value": ["subtle", "subject_kept", "desaturated"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:46:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "0ff5a1301bf44f81"}
diff --git a/records/sd1/sweep/864f117dca22e1b1077b/measurements.json b/records/sd1/sweep/864f117dca22e1b1077b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..15316452c5ab97020c645ad126d2b9adb6f3c851
--- /dev/null
+++ b/records/sd1/sweep/864f117dca22e1b1077b/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "864f117dca22e1b1077b",
+ "values": {
+  "std": {
+   "value": 43.436,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01016,
+   "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": 33.759,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.315424,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.325747,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.076725,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.050724,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9876,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.119,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth 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/864f117dca22e1b1077b/record.json b/records/sd1/sweep/864f117dca22e1b1077b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..a1d0ec7686320a10e69a37ef3b06e79a7ba68bd9
--- /dev/null
+++ b/records/sd1/sweep/864f117dca22e1b1077b/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "864f117dca22e1b1077b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:38:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.7
+   },
+   "group": "in.lo",
+   "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": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.lo",
+   "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": "3c817256b394f27a82fc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/864f117dca22e1b1077b/workflow.json b/records/sd1/sweep/864f117dca22e1b1077b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2f874ac89fa592aad1ff7d16aef52e4d1a0d49f0
--- /dev/null
+++ b/records/sd1/sweep/864f117dca22e1b1077b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 864f117dca22e1b1077b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 0.7\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 1234,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86659467629e9b75287f/measurements.json b/records/sd1/sweep/86659467629e9b75287f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4e411640a0882c32801a61243a908bb9d35fb90f
--- /dev/null
+++ b/records/sd1/sweep/86659467629e9b75287f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86659467629e9b75287f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.208314,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.291173,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.381392,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.22864,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.863,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00271,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.025,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9897,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3544,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86659467629e9b75287f/record.json b/records/sd1/sweep/86659467629e9b75287f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb46ba74a06e68d1262780edeac54db69a8da1d4
--- /dev/null
+++ b/records/sd1/sweep/86659467629e9b75287f/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86659467629e9b75287f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:11Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.in_layers.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 2
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "strong_grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/54f07ca45162f304be490e5b2dfcaf8baa3208a1cae5bac298a5ef8fad91ac6b.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "54f07ca45162f304be490e5b2dfcaf8baa3208a1cae5bac298a5ef8fad91ac6b",
+  "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": "bf8c27f168980edfb535"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86659467629e9b75287f/workflow.json b/records/sd1/sweep/86659467629e9b75287f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2c27168d7983bf069e5c278e08ebaebdf5cbd966
--- /dev/null
+++ b/records/sd1/sweep/86659467629e9b75287f/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 86659467629e9b75287f)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.in_layers.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/866c32a8deb628774951/measurements.json b/records/sd1/sweep/866c32a8deb628774951/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..589d07c348f8c3478c014b387877b130da8d83f8
--- /dev/null
+++ b/records/sd1/sweep/866c32a8deb628774951/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "866c32a8deb628774951",
+ "values": {
+  "cosine_distance": {
+   "value": 0.099395,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.136145,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.147188,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.035232,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.273,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00585,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 16.244,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0494,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.084,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/866c32a8deb628774951/record.json b/records/sd1/sweep/866c32a8deb628774951/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff86f5dfa730c7115b3ff0daad3486019f3d8219
--- /dev/null
+++ b/records/sd1/sweep/866c32a8deb628774951/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "866c32a8deb628774951",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:46Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.1.transformer_blocks.0.attn1.to_k",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "out.lo",
+   "kind": "self_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d85c5e5582a373bcd111b318bf735ec6e4a9b7c997e810b9731ddec600beb639.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d85c5e5582a373bcd111b318bf735ec6e4a9b7c997e810b9731ddec600beb639",
+  "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": "cffe448476b2b49dd3cb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/866c32a8deb628774951/workflow.json b/records/sd1/sweep/866c32a8deb628774951/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1f748d24a63fb7ec25f76789bd0112f133c55c8a
--- /dev/null
+++ b/records/sd1/sweep/866c32a8deb628774951/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 866c32a8deb628774951)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/867f50ce58b883555433/measurements.json b/records/sd1/sweep/867f50ce58b883555433/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c4d5b190c5c2522cfd14e2b1be1d0bf53dde403
--- /dev/null
+++ b/records/sd1/sweep/867f50ce58b883555433/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "867f50ce58b883555433",
+ "values": {
+  "cosine_distance": {
+   "value": 0.071449,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.127643,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.070529,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.032463,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.591,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00672,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 12.968,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.042,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1296,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/867f50ce58b883555433/record.json b/records/sd1/sweep/867f50ce58b883555433/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c8573f428df8f32921857df2639acca4f4d0cd21
--- /dev/null
+++ b/records/sd1/sweep/867f50ce58b883555433/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "867f50ce58b883555433",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:51Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.1.proj_out",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "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/3a736a36f13f6b18a073e540cf647b74f814c2f52324af8d022c96921942d48b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3a736a36f13f6b18a073e540cf647b74f814c2f52324af8d022c96921942d48b",
+  "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": "5cac5cbf39022633bef9"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/867f50ce58b883555433/workflow.json b/records/sd1/sweep/867f50ce58b883555433/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..29f49b5445db64609bc26ac759c5b58cb61951ff
--- /dev/null
+++ b/records/sd1/sweep/867f50ce58b883555433/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 867f50ce58b883555433)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.1.proj_out\",\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/868a557ab03f85d948aa/measurements.json b/records/sd1/sweep/868a557ab03f85d948aa/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..ef45dc226ea3ce64ad62cc58e68cb4be45c8c9c7
--- /dev/null
+++ b/records/sd1/sweep/868a557ab03f85d948aa/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "868a557ab03f85d948aa",
+ "values": {
+  "std": {
+   "value": 39.733,
+   "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"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 39.305,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.539305,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.71503,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.699189,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.33087,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.862,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.534,
+   "method": "CLIP zero-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/868a557ab03f85d948aa/record.json b/records/sd1/sweep/868a557ab03f85d948aa/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5a9b82e557be613817947d10eea1bd87f358a594
--- /dev/null
+++ b/records/sd1/sweep/868a557ab03f85d948aa/record.json
@@ -0,0 +1,83 @@
+{
+ "id": "868a557ab03f85d948aa",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:00:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 2.0
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "de1eccace874429a61dc",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "multiply",
+   2.0,
+   "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": "ea30c9eff4312bc31aef"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/868a557ab03f85d948aa/workflow.json b/records/sd1/sweep/868a557ab03f85d948aa/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1c3181d5456ff46e59f48655339c87b43434e9e7
--- /dev/null
+++ b/records/sd1/sweep/868a557ab03f85d948aa/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 868a557ab03f85d948aa)"
+  },
+  "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 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/868e93fac9f4327a6322/measurements.json b/records/sd1/sweep/868e93fac9f4327a6322/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..10abc02946a1917d160ab95a8d1beaf7c289dab8
--- /dev/null
+++ b/records/sd1/sweep/868e93fac9f4327a6322/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "868e93fac9f4327a6322",
+ "values": {
+  "cosine_distance": {
+   "value": 0.226884,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.373135,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.229056,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.134102,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 49.608,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00576,
+   "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": 27.1,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9303,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0492,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/868e93fac9f4327a6322/record.json b/records/sd1/sweep/868e93fac9f4327a6322/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..69e91b670cf86def9fc8f7594e3a3a1e2c823b75
--- /dev/null
+++ b/records/sd1/sweep/868e93fac9f4327a6322/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "868e93fac9f4327a6322",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:28Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.8.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.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/249ed58520c19d19c596fb5da86de9055baba1eaaa35f76d69c933347e09aa5b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "249ed58520c19d19c596fb5da86de9055baba1eaaa35f76d69c933347e09aa5b",
+  "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": "4c4992480bf5192990be"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/868e93fac9f4327a6322/workflow.json b/records/sd1/sweep/868e93fac9f4327a6322/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dcfd37875f2c10f0aa95a540b20c3ed13ebb2308
--- /dev/null
+++ b/records/sd1/sweep/868e93fac9f4327a6322/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 868e93fac9f4327a6322)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.0.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/86a1a05b793f433fd533/measurements.json b/records/sd1/sweep/86a1a05b793f433fd533/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc1c7c972d72b043e3817e35cd1f903591d2a6d8
--- /dev/null
+++ b/records/sd1/sweep/86a1a05b793f433fd533/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86a1a05b793f433fd533",
+ "values": {
+  "std": {
+   "value": 28.581,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02931,
+   "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": 13.686,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.09009,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.208604,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.129273,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.023741,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9425,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4073,
+   "method": "CLIP zero-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/86a1a05b793f433fd533/record.json b/records/sd1/sweep/86a1a05b793f433fd533/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c9a92d880db53bf50886e0e010b1c73ecc0c5bba
--- /dev/null
+++ b/records/sd1/sweep/86a1a05b793f433fd533/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "86a1a05b793f433fd533",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:59:55Z",
+ "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": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": 0.8
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "mid",
+   "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",
+   "add_scalar",
+   0.8,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "2b276135b3353d7bd4ea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86a1a05b793f433fd533/workflow.json b/records/sd1/sweep/86a1a05b793f433fd533/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..03124b4d305e14fe6091add002110985eed9dc3e
--- /dev/null
+++ b/records/sd1/sweep/86a1a05b793f433fd533/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86a1a05b793f433fd533)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": 0.8\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a 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/86a32e1096e33c39e469/measurements.json b/records/sd1/sweep/86a32e1096e33c39e469/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..89774a45ee1a8d278c3e63877e7ce78f5c485cf5
--- /dev/null
+++ b/records/sd1/sweep/86a32e1096e33c39e469/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86a32e1096e33c39e469",
+ "values": {
+  "cosine_distance": {
+   "value": 0.194276,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.354168,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.1406,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.091917,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.596,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00851,
+   "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": 22.3,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0087,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1428,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86a32e1096e33c39e469/record.json b/records/sd1/sweep/86a32e1096e33c39e469/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..93b2da1fe553d409a9980910fac320917e8fe9c0
--- /dev/null
+++ b/records/sd1/sweep/86a32e1096e33c39e469/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86a32e1096e33c39e469",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.11.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/6bd96f15fcb9a3e48e001f814b0c4fd0321259384f84274b97c12c48c61aa237.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "6bd96f15fcb9a3e48e001f814b0c4fd0321259384f84274b97c12c48c61aa237",
+  "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": "a400d5ecf71dc06330bb"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86a32e1096e33c39e469/workflow.json b/records/sd1/sweep/86a32e1096e33c39e469/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..513e1ea4b57de903333a5e963927c447a25a957e
--- /dev/null
+++ b/records/sd1/sweep/86a32e1096e33c39e469/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 86a32e1096e33c39e469)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    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\": -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/86a7778bfbf46333f777/measurements.json b/records/sd1/sweep/86a7778bfbf46333f777/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..943cf60004658ef7dc60367d90fa7fb9b389f595
--- /dev/null
+++ b/records/sd1/sweep/86a7778bfbf46333f777/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86a7778bfbf46333f777",
+ "values": {
+  "cosine_distance": {
+   "value": 0.400028,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.626679,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.690428,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.326187,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 37.035,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00232,
+   "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": 35.712,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9192,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2789,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86a7778bfbf46333f777/record.json b/records/sd1/sweep/86a7778bfbf46333f777/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..1b3d580e383f6d59414595841650a6814f613a3b
--- /dev/null
+++ b/records/sd1/sweep/86a7778bfbf46333f777/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "86a7778bfbf46333f777",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.in_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/0be0248d4690448c883c53cb86a5b3cebffd9ab82d0b02bd6412bd99021e3da8.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0be0248d4690448c883c53cb86a5b3cebffd9ab82d0b02bd6412bd99021e3da8",
+  "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": "7379ee4ebf357205cfee"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86a7778bfbf46333f777/workflow.json b/records/sd1/sweep/86a7778bfbf46333f777/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4d63c4698490095bb16639f859181c76b5aea334
--- /dev/null
+++ b/records/sd1/sweep/86a7778bfbf46333f777/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86a7778bfbf46333f777)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.in_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/86afbf3425ac69677f0e/interpretations.jsonl b/records/sd1/sweep/86afbf3425ac69677f0e/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..2e1251b67c372b9ff182a6f8e581c80b44bcc39a
--- /dev/null
+++ b/records/sd1/sweep/86afbf3425ac69677f0e/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86afbf3425ac69677f0e"}, "kind": "caption", "value": "Bright, glossy still life of red berries, oranges and strawberries before a black background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "13e085b23c7dd974"}
+{"target": {"record": "86afbf3425ac69677f0e"}, "kind": "change", "value": "The muted painted still life becomes a crisp, glossy, high-saturation illustration with vivid reds and yellows on a deep black background.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "23d10e08926ca87d"}
+{"target": {"record": "86afbf3425ac69677f0e"}, "kind": "keywords", "value": ["glossy still life", "vivid fruit", "high saturation", "black background", "clean illustration"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "68cfe0a6363a1cd8"}
+{"target": {"record": "86afbf3425ac69677f0e"}, "kind": "effect_tags", "value": ["saturated", "high_contrast", "style_change", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "51a341edb015918f"}
diff --git a/records/sd1/sweep/86afbf3425ac69677f0e/measurements.json b/records/sd1/sweep/86afbf3425ac69677f0e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7be54d3febaf856c605fc65cc8cf34822a99968
--- /dev/null
+++ b/records/sd1/sweep/86afbf3425ac69677f0e/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86afbf3425ac69677f0e",
+ "values": {
+  "std": {
+   "value": 55.633,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02677,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 47.756,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.312923,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.45599,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.195649,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.159549,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.8814,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6714,
+   "method": "CLIP zero-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/86afbf3425ac69677f0e/record.json b/records/sd1/sweep/86afbf3425ac69677f0e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..826d96678247ff6acb3040855771c65a1e5440ac
--- /dev/null
+++ b/records/sd1/sweep/86afbf3425ac69677f0e/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "86afbf3425ac69677f0e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:48:05Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 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": 2.0
+   },
+   "group": "out.hi",
+   "kind": "attn_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": [
+   "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": "1c170633a601fe01a6a4"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86afbf3425ac69677f0e/workflow.json b/records/sd1/sweep/86afbf3425ac69677f0e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3541d5f962017a3cf97375af4bdfafd4087d35a9
--- /dev/null
+++ b/records/sd1/sweep/86afbf3425ac69677f0e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86afbf3425ac69677f0e)"
+  },
+  "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": "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/86afee87ceab66cdeef2/measurements.json b/records/sd1/sweep/86afee87ceab66cdeef2/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e774525af11679f8d1192bbd468cb4551bc36008
--- /dev/null
+++ b/records/sd1/sweep/86afee87ceab66cdeef2/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86afee87ceab66cdeef2",
+ "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/86afee87ceab66cdeef2/record.json b/records/sd1/sweep/86afee87ceab66cdeef2/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..e90fb9dfd7ea4079f5d913420c801c90370e136f
--- /dev/null
+++ b/records/sd1/sweep/86afee87ceab66cdeef2/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86afee87ceab66cdeef2",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:43Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.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/b84ea72fba6e5817bb1ff00215d6cb92f5a85c4a2d6cad93840cde665f8ceb77.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b84ea72fba6e5817bb1ff00215d6cb92f5a85c4a2d6cad93840cde665f8ceb77",
+  "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": "c5a153dca01315824f48"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86afee87ceab66cdeef2/workflow.json b/records/sd1/sweep/86afee87ceab66cdeef2/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..44496a30de0f28521a268a679dfa0b7a40f30b55
--- /dev/null
+++ b/records/sd1/sweep/86afee87ceab66cdeef2/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 86afee87ceab66cdeef2)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.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/86b0d48e1961da193b55/measurements.json b/records/sd1/sweep/86b0d48e1961da193b55/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f4d77cd6f39caf2b5f090381ff74471b75c5ffa4
--- /dev/null
+++ b/records/sd1/sweep/86b0d48e1961da193b55/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86b0d48e1961da193b55",
+ "values": {
+  "cosine_distance": {
+   "value": 0.059039,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.082875,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.059491,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.016037,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.66,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00775,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 11.761,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0304,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1663,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b0d48e1961da193b55/record.json b/records/sd1/sweep/86b0d48e1961da193b55/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ce25d9011bda6c6913dfe2ec5fb2f2dc3a78d2c0
--- /dev/null
+++ b/records/sd1/sweep/86b0d48e1961da193b55/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "86b0d48e1961da193b55",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:16Z",
+ "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": "middle_block.1.transformer_blocks.0.attn2.to_q",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "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/1f06f8d927691d1a64717bb0b57de107ef8f50b025e995c37690c0d6b6897135.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1f06f8d927691d1a64717bb0b57de107ef8f50b025e995c37690c0d6b6897135",
+  "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": "34f0b6b73f679805c974"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b0d48e1961da193b55/workflow.json b/records/sd1/sweep/86b0d48e1961da193b55/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..a368dcf6801a31505d6c59f665171b8c19096aae
--- /dev/null
+++ b/records/sd1/sweep/86b0d48e1961da193b55/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86b0d48e1961da193b55)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.1.transformer_blocks.0.attn2.to_q\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b4b15f90788ff19aa3/measurements.json b/records/sd1/sweep/86b4b15f90788ff19aa3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..bffa31a4bf9fda8c1a79ec37685d1e8e63bfbd2b
--- /dev/null
+++ b/records/sd1/sweep/86b4b15f90788ff19aa3/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86b4b15f90788ff19aa3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.460838,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.503474,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.450124,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.254026,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 62.443,
+   "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": 47.793,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9105,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4951,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b4b15f90788ff19aa3/record.json b/records/sd1/sweep/86b4b15f90788ff19aa3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..81f83acb2e069218587b89e525dcd69e54bcc916
--- /dev/null
+++ b/records/sd1/sweep/86b4b15f90788ff19aa3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86b4b15f90788ff19aa3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:54Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/83c9c72fa1615bdd2744690660a2f4697397e990ca8c2872297cefc3678a69d3.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "83c9c72fa1615bdd2744690660a2f4697397e990ca8c2872297cefc3678a69d3",
+  "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": "d07f374f7ea1f7ff1b2a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b4b15f90788ff19aa3/workflow.json b/records/sd1/sweep/86b4b15f90788ff19aa3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9b60937fba41a09c5d1c79f62b714450766a7cf1
--- /dev/null
+++ b/records/sd1/sweep/86b4b15f90788ff19aa3/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 86b4b15f90788ff19aa3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/86b82ca18e5fbf030640/interpretations.jsonl b/records/sd1/sweep/86b82ca18e5fbf030640/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..ca2daa1a345b708ddc426719325a8fead4f2fe65
--- /dev/null
+++ b/records/sd1/sweep/86b82ca18e5fbf030640/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86b82ca18e5fbf030640"}, "kind": "caption", "value": "Abstract smeared texture of green and golden vertical fronds like streaked foliage", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "061aba7628db7ac4"}
+{"target": {"record": "86b82ca18e5fbf030640"}, "kind": "change", "value": "The portrait disappears entirely into an abstract streaked texture of green and ochre leaf-like vertical strokes.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "e73649d9309ef147"}
+{"target": {"record": "86b82ca18e5fbf030640"}, "kind": "keywords", "value": ["abstract foliage", "streaked texture", "green and gold", "no subject", "vertical smears"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "887bb9bd239c4af9"}
+{"target": {"record": "86b82ca18e5fbf030640"}, "kind": "effect_tags", "value": ["abstract", "subject_lost", "texture", "pattern", "colour_shift", "degenerate"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:10:48Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "540e49a036868189"}
diff --git a/records/sd1/sweep/86b82ca18e5fbf030640/measurements.json b/records/sd1/sweep/86b82ca18e5fbf030640/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d2c563a3b42cdc9b44bc81d5a670b5cd3d448db7
--- /dev/null
+++ b/records/sd1/sweep/86b82ca18e5fbf030640/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86b82ca18e5fbf030640",
+ "values": {
+  "cosine_distance": {
+   "value": 0.518861,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.701577,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.868262,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.528817,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 35.9,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01083,
+   "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": 54.075,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8002,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6785,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b82ca18e5fbf030640/record.json b/records/sd1/sweep/86b82ca18e5fbf030640/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..61ed9d5ba443d0361f0baa6299c9cea36a9c8899
--- /dev/null
+++ b/records/sd1/sweep/86b82ca18e5fbf030640/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86b82ca18e5fbf030640",
+ "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": -1
+   },
+   "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/b7692bf11af0fe2b942f3382e077dba631657336a62c6eedd4dc966c0150f268.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b7692bf11af0fe2b942f3382e077dba631657336a62c6eedd4dc966c0150f268",
+  "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": "0d96f35bf01d657ee854"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86b82ca18e5fbf030640/workflow.json b/records/sd1/sweep/86b82ca18e5fbf030640/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4daebc2ea143ba27f199b76a410f0b89e71f54a5
--- /dev/null
+++ b/records/sd1/sweep/86b82ca18e5fbf030640/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 86b82ca18e5fbf030640)"
+  },
+  "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\": -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/86c297423d1f76a737b9/measurements.json b/records/sd1/sweep/86c297423d1f76a737b9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cb382fd9f7bb420f86300a3e06a6fe0408f1fe51
--- /dev/null
+++ b/records/sd1/sweep/86c297423d1f76a737b9/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86c297423d1f76a737b9",
+ "values": {
+  "std": {
+   "value": 28.621,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.02657,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 10.672,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.049059,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.153702,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.155999,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044683,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.972,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3902,
+   "method": "CLIP zero-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/86c297423d1f76a737b9/record.json b/records/sd1/sweep/86c297423d1f76a737b9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..01fa599e20bf16fea8a86ce60bfd0d167192bfd3
--- /dev/null
+++ b/records/sd1/sweep/86c297423d1f76a737b9/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "86c297423d1f76a737b9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:58:51Z",
+ "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 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": 0.667,
+   "end": 1.0,
+   "window": "late",
+   "bucket": "grow",
+   "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": [
+   "in.lo",
+   "scale",
+   1.4,
+   "late"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c85b70423e406d43def5"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86c297423d1f76a737b9/workflow.json b/records/sd1/sweep/86c297423d1f76a737b9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6541de8696a1f0bbb72eed2db5c1308511b1c11c
--- /dev/null
+++ b/records/sd1/sweep/86c297423d1f76a737b9/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86c297423d1f76a737b9)"
+  },
+  "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 \"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/86cc5e50ccd33e25f249/interpretations.jsonl b/records/sd1/sweep/86cc5e50ccd33e25f249/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6f42dff140b3487531149c08533502c0486aed01
--- /dev/null
+++ b/records/sd1/sweep/86cc5e50ccd33e25f249/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86cc5e50ccd33e25f249"}, "kind": "caption", "value": "Painted portrait of a woman with wavy dark hair and black turtleneck on a dark grey board", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "abf244d143e432a0"}
+{"target": {"record": "86cc5e50ccd33e25f249"}, "kind": "change", "value": "Almost no visible change. The face is slightly more angular and the frame slightly cleaner and paler.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ebdce551cae447da"}
+{"target": {"record": "86cc5e50ccd33e25f249"}, "kind": "keywords", "value": ["near identical", "portrait", "slight variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d48632a7e7764931"}
+{"target": {"record": "86cc5e50ccd33e25f249"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:29:41Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "74c2b8d10c9a48a1"}
diff --git a/records/sd1/sweep/86cc5e50ccd33e25f249/measurements.json b/records/sd1/sweep/86cc5e50ccd33e25f249/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9410767a3d844de470f1b54327f023d4592013e7
--- /dev/null
+++ b/records/sd1/sweep/86cc5e50ccd33e25f249/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86cc5e50ccd33e25f249",
+ "values": {
+  "cosine_distance": {
+   "value": 0.060778,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.097437,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.039595,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.01897,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 69.094,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00516,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 14.569,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.037,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0883,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86cc5e50ccd33e25f249/record.json b/records/sd1/sweep/86cc5e50ccd33e25f249/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0571b4507b1b769f749114a2975c78b878671f85
--- /dev/null
+++ b/records/sd1/sweep/86cc5e50ccd33e25f249/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "86cc5e50ccd33e25f249",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:45:02Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.5.0.in_layers.2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/dc8da079ae8b382e0efc26dc37559023ea202983a7af543d7ed8675b25f5e92b.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "dc8da079ae8b382e0efc26dc37559023ea202983a7af543d7ed8675b25f5e92b",
+  "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": "ca41b878983bf2c3c041"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86cc5e50ccd33e25f249/workflow.json b/records/sd1/sweep/86cc5e50ccd33e25f249/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..dbe312cb0b53f45fe727d3d38393f26c90dc5cfd
--- /dev/null
+++ b/records/sd1/sweep/86cc5e50ccd33e25f249/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86cc5e50ccd33e25f249)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.5.0.in_layers.2\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 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/86d05429e86c9ef92b52/interpretations.jsonl b/records/sd1/sweep/86d05429e86c9ef92b52/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..bd1fbb78e9e563e9291ca6bc7649a7baa8d40879
--- /dev/null
+++ b/records/sd1/sweep/86d05429e86c9ef92b52/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86d05429e86c9ef92b52"}, "kind": "caption", "value": "Frontal portrait of woman with lighter ash-blonde highlighted wavy hair and paler skin", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8c80199aa3947ba9"}
+{"target": {"record": "86d05429e86c9ef92b52"}, "kind": "change", "value": "Same composition but hair becomes lighter and ashier with brighter highlights, skin paler and overall tones slightly cooler and washed.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3940b8d892127e9d"}
+{"target": {"record": "86d05429e86c9ef92b52"}, "kind": "keywords", "value": ["lighter hair", "ash highlights", "pale skin", "cooler tones", "minor variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "691e0a74c571f15a"}
+{"target": {"record": "86d05429e86c9ef92b52"}, "kind": "effect_tags", "value": ["subtle", "subject_kept", "colour_shift", "bright"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:30:38Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "802dda93541566c6"}
diff --git a/records/sd1/sweep/86d05429e86c9ef92b52/measurements.json b/records/sd1/sweep/86d05429e86c9ef92b52/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..4c62eaf5fae23187fbf5da125d019b19044285b0
--- /dev/null
+++ b/records/sd1/sweep/86d05429e86c9ef92b52/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86d05429e86c9ef92b52",
+ "values": {
+  "cosine_distance": {
+   "value": 0.099445,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.119212,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.084157,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.026074,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 56.57,
+   "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": 15.775,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0253,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1781,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86d05429e86c9ef92b52/record.json b/records/sd1/sweep/86d05429e86c9ef92b52/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..84f4d1703e8eba86dcece275712544db0dca2546
--- /dev/null
+++ b/records/sd1/sweep/86d05429e86c9ef92b52/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "86d05429e86c9ef92b52",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:44:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.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/4a1474edfa15e66e3a85d5a513dae82e984a366234468849888a71548a1b254f.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "4a1474edfa15e66e3a85d5a513dae82e984a366234468849888a71548a1b254f",
+  "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": "7f7c7837f653752bc502"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86d05429e86c9ef92b52/workflow.json b/records/sd1/sweep/86d05429e86c9ef92b52/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5cc914cbae2d80711fa7bd2342f0975f6a309ab3
--- /dev/null
+++ b/records/sd1/sweep/86d05429e86c9ef92b52/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86d05429e86c9ef92b52)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.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/86daf7da522f4e6fd430/interpretations.jsonl b/records/sd1/sweep/86daf7da522f4e6fd430/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d36583897379822aca16466f6ebe1ce3907e7ee5
--- /dev/null
+++ b/records/sd1/sweep/86daf7da522f4e6fd430/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86daf7da522f4e6fd430"}, "kind": "caption", "value": "Heavily blurred grey abstract with soft dark curves suggesting a shoulder", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "5c18d383d9a4d860"}
+{"target": {"record": "86daf7da522f4e6fd430"}, "kind": "change", "value": "Almost everything blurs into soft grey shapes, with only a vague dark curve hinting at a figure.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "300fe1857b1f3898"}
+{"target": {"record": "86daf7da522f4e6fd430"}, "kind": "keywords", "value": ["heavy blur", "grey abstract", "out of focus", "minimal", "lost subject"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "760f57eb346b559d"}
+{"target": {"record": "86daf7da522f4e6fd430"}, "kind": "effect_tags", "value": ["blur", "abstract", "subject_lost", "desaturated", "washed_out"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:19:03Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2a06b44f508602d1"}
diff --git a/records/sd1/sweep/86daf7da522f4e6fd430/measurements.json b/records/sd1/sweep/86daf7da522f4e6fd430/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..03e7f0a7b5154fa3dc7dcca85d4a4baec827f1f5
--- /dev/null
+++ b/records/sd1/sweep/86daf7da522f4e6fd430/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86daf7da522f4e6fd430",
+ "values": {
+  "cosine_distance": {
+   "value": 0.361709,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.61599,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.702234,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.398917,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 37.42,
+   "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": 41.675,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8862,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.451,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86daf7da522f4e6fd430/record.json b/records/sd1/sweep/86daf7da522f4e6fd430/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..da903806825ddfba6f6968f61831fc45f2f79778
--- /dev/null
+++ b/records/sd1/sweep/86daf7da522f4e6fd430/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86daf7da522f4e6fd430",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/a95704643f1d2ceb189c067cc5797c84599034f0dce2858e3224e81fe6400e53.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "a95704643f1d2ceb189c067cc5797c84599034f0dce2858e3224e81fe6400e53",
+  "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": "7b796a79390a5bc04b6b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86daf7da522f4e6fd430/workflow.json b/records/sd1/sweep/86daf7da522f4e6fd430/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6492611a566e08006930c45266314499575e6b2
--- /dev/null
+++ b/records/sd1/sweep/86daf7da522f4e6fd430/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 86daf7da522f4e6fd430)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/86e7dcdcaadf5d1f55c9/interpretations.jsonl b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..d5cec38768559bc06cd4e0bea01779fe17feceee
--- /dev/null
+++ b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86e7dcdcaadf5d1f55c9"}, "kind": "caption", "value": "Young woman with smooth pale skin and rosy lips gazing at viewer, vintage frame", "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": "5dfc6f0d844401cf"}
+{"target": {"record": "86e7dcdcaadf5d1f55c9"}, "kind": "change", "value": "The person is kept but rendered smoother and more glamorous, with pinker lips and cheeks and a warmer border.", "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": "32ef922fef208fed"}
+{"target": {"record": "86e7dcdcaadf5d1f55c9"}, "kind": "keywords", "value": ["smooth skin", "rosy lips", "glamour portrait", "vintage frame"], "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": "4cbbae181359c90d"}
+{"target": {"record": "86e7dcdcaadf5d1f55c9"}, "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": "2026-10-05T08:29:25Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "940ccb2a2a6f6ebc"}
diff --git a/records/sd1/sweep/86e7dcdcaadf5d1f55c9/measurements.json b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..eb547983c2144df324aa2f9a055b5e6e5bff3254
--- /dev/null
+++ b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86e7dcdcaadf5d1f55c9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.137367,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.158917,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.086944,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.103925,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 66.413,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00356,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.977,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9422,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6638,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86e7dcdcaadf5d1f55c9/record.json b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b01b2e3c5cecc3a21c4cdbd9ead33f6761d9f762
--- /dev/null
+++ b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86e7dcdcaadf5d1f55c9",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:23Z",
+ "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.8.1.proj_out",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/f48e9d0ef247564f854276f64209bf29db3f2d7ad2be08bc8dfe1cba619c44e3.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "f48e9d0ef247564f854276f64209bf29db3f2d7ad2be08bc8dfe1cba619c44e3",
+  "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": "519609dd629a205c585a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86e7dcdcaadf5d1f55c9/workflow.json b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e8921f215e97e36bc6d8fd6150273f1d096120e1
--- /dev/null
+++ b/records/sd1/sweep/86e7dcdcaadf5d1f55c9/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 86e7dcdcaadf5d1f55c9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.proj_out\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86f43ef86b2c281f493c/interpretations.jsonl b/records/sd1/sweep/86f43ef86b2c281f493c/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..303143160585b2185ffdb4de43c657ae562bd8fa
--- /dev/null
+++ b/records/sd1/sweep/86f43ef86b2c281f493c/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86f43ef86b2c281f493c"}, "kind": "caption", "value": "Frontal woman with dark shoulder-length hair and necklace on stained teal-and-ochre aged background", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bc7d2527ae00b753"}
+{"target": {"record": "86f43ef86b2c281f493c"}, "kind": "change", "value": "The pose goes frontal with longer hair and a pendant, and the clean grey backdrop becomes a weathered teal print with stained ochre edges.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ccf234d54bc3b692"}
+{"target": {"record": "86f43ef86b2c281f493c"}, "kind": "keywords", "value": ["aged print", "teal background", "frontal portrait", "stained edges", "necklace"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4dedf81a8cfdce83"}
+{"target": {"record": "86f43ef86b2c281f493c"}, "kind": "effect_tags", "value": ["subject_kept", "texture", "colour_shift", "framing"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:07:29Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c3c4df8d2538c424"}
diff --git a/records/sd1/sweep/86f43ef86b2c281f493c/measurements.json b/records/sd1/sweep/86f43ef86b2c281f493c/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..11db259f9d2a5ce8d11e860ffd39b4972df86b38
--- /dev/null
+++ b/records/sd1/sweep/86f43ef86b2c281f493c/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86f43ef86b2c281f493c",
+ "values": {
+  "cosine_distance": {
+   "value": 0.404203,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.487774,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.297652,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.123441,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 72.673,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00578,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 50.974,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0482,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1286,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86f43ef86b2c281f493c/record.json b/records/sd1/sweep/86f43ef86b2c281f493c/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..847f02111f888b6cc203e8235d3abea3e9214efc
--- /dev/null
+++ b/records/sd1/sweep/86f43ef86b2c281f493c/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86f43ef86b2c281f493c",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:30Z",
+ "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.1.0.in_layers.1",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.5
+   },
+   "group": "in.hi",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/788b859534642550ded3f10483b8b306f386fe7ed589e8bc41de85fabcbd3b24.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "788b859534642550ded3f10483b8b306f386fe7ed589e8bc41de85fabcbd3b24",
+  "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": "4c57e89c30446a85e9dd"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86f43ef86b2c281f493c/workflow.json b/records/sd1/sweep/86f43ef86b2c281f493c/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9d10ee2f84bd74c7f2fb891d0d8a46aea33dd7d0
--- /dev/null
+++ b/records/sd1/sweep/86f43ef86b2c281f493c/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 86f43ef86b2c281f493c)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.1.0.in_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/86f948a0be39d0b84516/measurements.json b/records/sd1/sweep/86f948a0be39d0b84516/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..51de88265c4e641d418ed0e92fbdb78f6beeb204
--- /dev/null
+++ b/records/sd1/sweep/86f948a0be39d0b84516/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "86f948a0be39d0b84516",
+ "values": {
+  "cosine_distance": {
+   "value": 0.130681,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.240664,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.139126,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.02994,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.702,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00722,
+   "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": 20.439,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.003,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1091,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86f948a0be39d0b84516/record.json b/records/sd1/sweep/86f948a0be39d0b84516/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..79cb2267c5998167f8908cc7604e7727109c73c1
--- /dev/null
+++ b/records/sd1/sweep/86f948a0be39d0b84516/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "86f948a0be39d0b84516",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:20Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.7.1.proj_in",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/17581b365bb06fc5e07a9abc97b43f0158a4e53a9df1a4fea13ed90cb29878c0.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "17581b365bb06fc5e07a9abc97b43f0158a4e53a9df1a4fea13ed90cb29878c0",
+  "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": "0621b868be1288f83000"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86f948a0be39d0b84516/workflow.json b/records/sd1/sweep/86f948a0be39d0b84516/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..30da9f94a88029036b53b33ca39f01a7e65753a3
--- /dev/null
+++ b/records/sd1/sweep/86f948a0be39d0b84516/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 86f948a0be39d0b84516)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.proj_in\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86fa958b62ccbea77a99/interpretations.jsonl b/records/sd1/sweep/86fa958b62ccbea77a99/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6bd1f4b35d84517a343710166f75eee03c8df06f
--- /dev/null
+++ b/records/sd1/sweep/86fa958b62ccbea77a99/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "86fa958b62ccbea77a99"}, "kind": "caption", "value": "Portrait of a woman with an asymmetric bob and red lips on a cloudy teal backdrop", "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": "bfc227858f0ef267"}
+{"target": {"record": "86fa958b62ccbea77a99"}, "kind": "change", "value": "The hair becomes an asymmetric bob with a long strand on one side, the lips turn red, and the background shifts to a cloudy teal-green with darker vignetted edges.", "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": "3cc66c744a1dfdb5"}
+{"target": {"record": "86fa958b62ccbea77a99"}, "kind": "keywords", "value": ["asymmetric bob", "red lips", "teal backdrop", "painted background", "vignette"], "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": "58427f8344807d2f"}
+{"target": {"record": "86fa958b62ccbea77a99"}, "kind": "effect_tags", "value": ["colour_shift", "texture", "subject_kept"], "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": "3b110fbc674302d8"}
diff --git a/records/sd1/sweep/86fa958b62ccbea77a99/measurements.json b/records/sd1/sweep/86fa958b62ccbea77a99/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..aae67b5110822612d931650eae7ad9f8729a0926
--- /dev/null
+++ b/records/sd1/sweep/86fa958b62ccbea77a99/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "86fa958b62ccbea77a99",
+ "values": {
+  "cosine_distance": {
+   "value": 0.163874,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.208395,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.108454,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.081751,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 65.274,
+   "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": 20.516,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0243,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0696,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86fa958b62ccbea77a99/record.json b/records/sd1/sweep/86fa958b62ccbea77a99/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..560ddd891892ce8654dfc47abc834dee30748d11
--- /dev/null
+++ b/records/sd1/sweep/86fa958b62ccbea77a99/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "86fa958b62ccbea77a99",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:51:27Z",
+ "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.11.1.norm",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.hi",
+   "kind": "norm",
+   "module_type": "GroupNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/4508d70daa609cd30710ed41f59c7e5a5526d60ed8ef6d0b6656996dcf937fc2.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "4508d70daa609cd30710ed41f59c7e5a5526d60ed8ef6d0b6656996dcf937fc2",
+  "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": "edf91ec7ef962b065004"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/86fa958b62ccbea77a99/workflow.json b/records/sd1/sweep/86fa958b62ccbea77a99/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..899a72ab7399342edc9edec31cbcc12bd45e22b5
--- /dev/null
+++ b/records/sd1/sweep/86fa958b62ccbea77a99/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 86fa958b62ccbea77a99)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.11.1.norm\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 30\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/870c88fe606743ee0763/measurements.json b/records/sd1/sweep/870c88fe606743ee0763/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..58210c48f77bec861820360a804c58eaa8b5eea6
--- /dev/null
+++ b/records/sd1/sweep/870c88fe606743ee0763/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "870c88fe606743ee0763",
+ "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/870c88fe606743ee0763/record.json b/records/sd1/sweep/870c88fe606743ee0763/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ec5b87fef314942976987b2a47777a0ed2bc14cd
--- /dev/null
+++ b/records/sd1/sweep/870c88fe606743ee0763/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "870c88fe606743ee0763",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:01Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.h_upd",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "mid",
+   "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/46bcb62fde93f0894eabd02fe3b967f2c589891ab9e775c1d6ec38b90e42439d.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "46bcb62fde93f0894eabd02fe3b967f2c589891ab9e775c1d6ec38b90e42439d",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "e66299972ca75f26ff82"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/870c88fe606743ee0763/workflow.json b/records/sd1/sweep/870c88fe606743ee0763/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4be9632a897eb57867ce28aa341308d107c01888
--- /dev/null
+++ b/records/sd1/sweep/870c88fe606743ee0763/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 870c88fe606743ee0763)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.0.h_upd\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8719e5cf8b883fd4f91e/measurements.json b/records/sd1/sweep/8719e5cf8b883fd4f91e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..71e9b0eb59bdb1663489a1e73c2ecb8feca7f2ca
--- /dev/null
+++ b/records/sd1/sweep/8719e5cf8b883fd4f91e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8719e5cf8b883fd4f91e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.190864,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.42935,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.235871,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.12358,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.794,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00032,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 22.329,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0796,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4725,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8719e5cf8b883fd4f91e/record.json b/records/sd1/sweep/8719e5cf8b883fd4f91e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c246c5d7d8c549be3076328a6b5b902dad7e2bd2
--- /dev/null
+++ b/records/sd1/sweep/8719e5cf8b883fd4f91e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8719e5cf8b883fd4f91e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:51:06Z",
+ "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.11.0.out_layers.3",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Conv2d",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/62a8d97c4a4878e4921e5398e78199bb620b17ecab73dfe55ef1a42c1e321092.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "62a8d97c4a4878e4921e5398e78199bb620b17ecab73dfe55ef1a42c1e321092",
+  "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": "22d9c13246ce1954ea93"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8719e5cf8b883fd4f91e/workflow.json b/records/sd1/sweep/8719e5cf8b883fd4f91e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..6c49acd9ee161280c81f2f11e572aa2c525814a9
--- /dev/null
+++ b/records/sd1/sweep/8719e5cf8b883fd4f91e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8719e5cf8b883fd4f91e)"
+  },
+  "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\": 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/871e75bdcb5e62454d50/measurements.json b/records/sd1/sweep/871e75bdcb5e62454d50/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5abe244d896c549ec710553f773e2ce0fa967220
--- /dev/null
+++ b/records/sd1/sweep/871e75bdcb5e62454d50/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "871e75bdcb5e62454d50",
+ "values": {
+  "std": {
+   "value": 48.234,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01688,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.624,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.164379,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.22071,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.103607,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.018769,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9656,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4484,
+   "method": "CLIP zero-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/871e75bdcb5e62454d50/record.json b/records/sd1/sweep/871e75bdcb5e62454d50/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..5d97173e5084b3878d34b9b49e143e8839ea9a2d
--- /dev/null
+++ b/records/sd1/sweep/871e75bdcb5e62454d50/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "871e75bdcb5e62454d50",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:09:49Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a 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": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "mid",
+   "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": [
+   "mid",
+   "multiply",
+   0.5,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "15de98703e8b5eb8f9e1"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/871e75bdcb5e62454d50/workflow.json b/records/sd1/sweep/871e75bdcb5e62454d50/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..948e7bdddbd8960ca016eda80bfa530b97ef8b47
--- /dev/null
+++ b/records/sd1/sweep/871e75bdcb5e62454d50/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 871e75bdcb5e62454d50)"
+  },
+  "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\": 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/8720f5b087b753a15925/measurements.json b/records/sd1/sweep/8720f5b087b753a15925/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e586cdeb5b8878dbd7763c76ae9388c31fb2333
--- /dev/null
+++ b/records/sd1/sweep/8720f5b087b753a15925/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8720f5b087b753a15925",
+ "values": {
+  "cosine_distance": {
+   "value": 0.396773,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.54996,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.557612,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.298703,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.273,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00302,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 44.343,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0036,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1781,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8720f5b087b753a15925/record.json b/records/sd1/sweep/8720f5b087b753a15925/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3518d260d4bd1e1b1b54646c5caf970218ce9450
--- /dev/null
+++ b/records/sd1/sweep/8720f5b087b753a15925/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8720f5b087b753a15925",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:52:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.4.0.out_layers.2",
+   "op": "scale",
+   "args": {
+    "scale_factor": 0.5
+   },
+   "group": "out.lo",
+   "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/30c1dd5e56e20238b36e2913eb62091239161910bf608701c56815684e03b7cb.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "30c1dd5e56e20238b36e2913eb62091239161910bf608701c56815684e03b7cb",
+  "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": "d452f59a66c532964dee"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8720f5b087b753a15925/workflow.json b/records/sd1/sweep/8720f5b087b753a15925/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..4cf67110e8ea6eef0cd58cc96845b5ff2ce84fea
--- /dev/null
+++ b/records/sd1/sweep/8720f5b087b753a15925/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 8720f5b087b753a15925)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.4.0.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/872118540085931062e3/measurements.json b/records/sd1/sweep/872118540085931062e3/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f7a2586a65444bccbc14a32efcfd9f3af583b5df
--- /dev/null
+++ b/records/sd1/sweep/872118540085931062e3/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "872118540085931062e3",
+ "values": {
+  "cosine_distance": {
+   "value": 0.361709,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.61599,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.702234,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.398917,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 37.42,
+   "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": 41.675,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.8862,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.451,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/872118540085931062e3/record.json b/records/sd1/sweep/872118540085931062e3/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..07ffb4c4f9b488c10a047dd7be189853f19fc638
--- /dev/null
+++ b/records/sd1/sweep/872118540085931062e3/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "872118540085931062e3",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:45Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.0.0.out_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8825ac1a626d4fb0cb2b37a6698d2ffcb2a2b32be16483873809230758c6df45.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8825ac1a626d4fb0cb2b37a6698d2ffcb2a2b32be16483873809230758c6df45",
+  "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": "c6a1df2fd4f7469e055d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/872118540085931062e3/workflow.json b/records/sd1/sweep/872118540085931062e3/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1768a3700ff3f7c281adba73934e3a94278317e3
--- /dev/null
+++ b/records/sd1/sweep/872118540085931062e3/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 872118540085931062e3)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.0.0.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/8723d604d94be90e05bf/measurements.json b/records/sd1/sweep/8723d604d94be90e05bf/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..27943b00afa752f1bd1997034e5aadf5c8b3366c
--- /dev/null
+++ b/records/sd1/sweep/8723d604d94be90e05bf/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8723d604d94be90e05bf",
+ "values": {
+  "cosine_distance": {
+   "value": 0.099511,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.159948,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.105304,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.027593,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.132,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00695,
+   "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": 15.689,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9944,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1133,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8723d604d94be90e05bf/record.json b/records/sd1/sweep/8723d604d94be90e05bf/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b47c48f7c11f62274ac3461da79ed8ac654b51a4
--- /dev/null
+++ b/records/sd1/sweep/8723d604d94be90e05bf/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8723d604d94be90e05bf",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:39:59Z",
+ "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.1.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.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/dad82d4c3dfc3c6866cbede9ead5541423f746358a0eddd74406265eafb60232.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "dad82d4c3dfc3c6866cbede9ead5541423f746358a0eddd74406265eafb60232",
+  "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": "e6efbfc1e95f380a08c8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8723d604d94be90e05bf/workflow.json b/records/sd1/sweep/8723d604d94be90e05bf/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc744b0389d58bf7f30982e13aa0f2c6cb2a115c
--- /dev/null
+++ b/records/sd1/sweep/8723d604d94be90e05bf/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8723d604d94be90e05bf)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/8729b55e441f6ae5e5a8/measurements.json b/records/sd1/sweep/8729b55e441f6ae5e5a8/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..f5d75a97efba68fe42d753f8ca84ecf53d1ca0bf
--- /dev/null
+++ b/records/sd1/sweep/8729b55e441f6ae5e5a8/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8729b55e441f6ae5e5a8",
+ "values": {
+  "std": {
+   "value": 58.956,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00978,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 7.764,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.030595,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.046222,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.012674,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.00393,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0106,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3479,
+   "method": "CLIP zero-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/8729b55e441f6ae5e5a8/record.json b/records/sd1/sweep/8729b55e441f6ae5e5a8/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..dd4911a17b412ed394cb60b3c8e46ac219e3e16c
--- /dev/null
+++ b/records/sd1/sweep/8729b55e441f6ae5e5a8/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "8729b55e441f6ae5e5a8",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:54:52Z",
+ "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": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "output_blocks.6.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "out.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "mid",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.mid",
+   "add_scalar",
+   -0.8,
+   "early"
+  ],
+  "clamps": [],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "7bd6227b5a775c2dc7cf"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8729b55e441f6ae5e5a8/workflow.json b/records/sd1/sweep/8729b55e441f6ae5e5a8/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b103eb851056d899da9a2dc08ee1090dcea95d75
--- /dev/null
+++ b/records/sd1/sweep/8729b55e441f6ae5e5a8/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8729b55e441f6ae5e5a8)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.6.0\",\n   \"module_type\": \"add_scalar\",\n   \"module_args\": {\n    \"scalar\": -0.8\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "portrait of an old fisherman, analog photograph"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/872e8a3036ec21b9cf5d/measurements.json b/records/sd1/sweep/872e8a3036ec21b9cf5d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..b31b325f4e6a808314619e9caeeb4facf74ab48e
--- /dev/null
+++ b/records/sd1/sweep/872e8a3036ec21b9cf5d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "872e8a3036ec21b9cf5d",
+ "values": {
+  "std": {
+   "value": 38.283,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01617,
+   "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": 12.827,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.096301,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.187958,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.038732,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.013623,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9931,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.019,
+   "method": "CLIP zero-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/872e8a3036ec21b9cf5d/record.json b/records/sd1/sweep/872e8a3036ec21b9cf5d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..bc337b4e82f6b4a3123d213b9556b7c35293e470
--- /dev/null
+++ b/records/sd1/sweep/872e8a3036ec21b9cf5d/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "872e8a3036ec21b9cf5d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:04:02Z",
+ "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 lighthouse on a rocky coast at dusk, oil painting",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e368f07d8e06f0e7"
+ },
+ "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": "de1eccace874429a61dc",
+  "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/872e8a3036ec21b9cf5d/workflow.json b/records/sd1/sweep/872e8a3036ec21b9cf5d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..9308d71cfe34ffda2cec8e3f211846db028cd064
--- /dev/null
+++ b/records/sd1/sweep/872e8a3036ec21b9cf5d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 872e8a3036ec21b9cf5d)"
+  },
+  "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 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/8739245ef9b4c6230e04/interpretations.jsonl b/records/sd1/sweep/8739245ef9b4c6230e04/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..dffe948cea7d1f0f64ecf3917009e1e19b703396
--- /dev/null
+++ b/records/sd1/sweep/8739245ef9b4c6230e04/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8739245ef9b4c6230e04"}, "kind": "caption", "value": "Portrait of woman with bright blue eyes, swept dark hair and turtleneck on grey", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "44821a160a3a7a4d"}
+{"target": {"record": "8739245ef9b4c6230e04"}, "kind": "change", "value": "The composition stays but the face becomes more realistic with vivid blue eyes, and the hair is darker and swept to one side with fewer highlights.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "cfff8e3d10025f68"}
+{"target": {"record": "8739245ef9b4c6230e04"}, "kind": "keywords", "value": ["blue eyes", "realistic face", "dark hair", "portrait variation"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b6363029b885d9be"}
+{"target": {"record": "8739245ef9b4c6230e04"}, "kind": "effect_tags", "value": ["subject_kept", "subtle", "style_change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:22:34Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3aae85fca16f8ad0"}
diff --git a/records/sd1/sweep/8739245ef9b4c6230e04/measurements.json b/records/sd1/sweep/8739245ef9b4c6230e04/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..404745414af213b6db821e7d81d827a6faa70c57
--- /dev/null
+++ b/records/sd1/sweep/8739245ef9b4c6230e04/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8739245ef9b4c6230e04",
+ "values": {
+  "cosine_distance": {
+   "value": 0.098242,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.132023,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.049857,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.041655,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 58.14,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00549,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 15.01,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9767,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0611,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8739245ef9b4c6230e04/record.json b/records/sd1/sweep/8739245ef9b4c6230e04/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..0baf82a8af43afa24bd3c984ce0c450ea66204a4
--- /dev/null
+++ b/records/sd1/sweep/8739245ef9b4c6230e04/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8739245ef9b4c6230e04",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:43:18Z",
+ "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": "middle_block.1.transformer_blocks.0.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "mid",
+   "kind": "cross_attn",
+   "module_type": "Linear",
+   "container": "CrossAttention",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "boost"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/597ff902f5b5f628cf419798add40c312dd1698ea9e91aa295221d16ef6ddb73.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "597ff902f5b5f628cf419798add40c312dd1698ea9e91aa295221d16ef6ddb73",
+  "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": "30c3844f4f10a579a1f2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8739245ef9b4c6230e04/workflow.json b/records/sd1/sweep/8739245ef9b4c6230e04/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..70e8ead02269b43f4e72eaf26f2b37f4e2ed7575
--- /dev/null
+++ b/records/sd1/sweep/8739245ef9b4c6230e04/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8739245ef9b4c6230e04)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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/8739befe92cbf8f1c577/interpretations.jsonl b/records/sd1/sweep/8739befe92cbf8f1c577/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..28c302e3d063cf04e8243690b3b78b1cf37ecdd9
--- /dev/null
+++ b/records/sd1/sweep/8739befe92cbf8f1c577/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8739befe92cbf8f1c577"}, "kind": "caption", "value": "Warm, soft-focus portrait of a woman with brown hair against a teal backdrop", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "fc20d1735a7bbe52"}
+{"target": {"record": "8739befe92cbf8f1c577"}, "kind": "change", "value": "The painting becomes a warmer, slightly blurred, film-like portrait with orange-tinted skin, a brown bob with a soft blurred edge, and a deep teal backdrop in place of the grey panel.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "2dc6c2c2cb6ea80f"}
+{"target": {"record": "8739befe92cbf8f1c577"}, "kind": "keywords", "value": ["warm portrait", "teal backdrop", "soft focus", "film look", "orange skin tone"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ae36de209b4a3fa0"}
+{"target": {"record": "8739befe92cbf8f1c577"}, "kind": "effect_tags", "value": ["colour_shift", "blur", "style_change", "saturated", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:18:55Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "70b14130d8fe101b"}
diff --git a/records/sd1/sweep/8739befe92cbf8f1c577/measurements.json b/records/sd1/sweep/8739befe92cbf8f1c577/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..925b1f78b28d48a0d3c928d036ae3b043d427851
--- /dev/null
+++ b/records/sd1/sweep/8739befe92cbf8f1c577/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8739befe92cbf8f1c577",
+ "values": {
+  "cosine_distance": {
+   "value": 0.231142,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.304826,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.080852,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.068313,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.916,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00301,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 23.397,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0321,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0868,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8739befe92cbf8f1c577/record.json b/records/sd1/sweep/8739befe92cbf8f1c577/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..62ec2bdb1cb559992a7f23ac456644a61d5a0a07
--- /dev/null
+++ b/records/sd1/sweep/8739befe92cbf8f1c577/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8739befe92cbf8f1c577",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:58Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": 2
+   },
+   "group": "in.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/a3c7588e7977ff4b1524ebd47292971a42046d89828cbcf11bca764b14160df4.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "a3c7588e7977ff4b1524ebd47292971a42046d89828cbcf11bca764b14160df4",
+  "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": "5a1c1bc79fcc080a4844"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8739befe92cbf8f1c577/workflow.json b/records/sd1/sweep/8739befe92cbf8f1c577/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..54d8a4b082d3cbb69604f32fa1e0f72f370c1c04
--- /dev/null
+++ b/records/sd1/sweep/8739befe92cbf8f1c577/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8739befe92cbf8f1c577)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.skip_connection\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 2\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873ad41a849dea4e9ecd/measurements.json b/records/sd1/sweep/873ad41a849dea4e9ecd/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..7778a5e8698c836f35e574a536a1b11bdaeb88bd
--- /dev/null
+++ b/records/sd1/sweep/873ad41a849dea4e9ecd/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "873ad41a849dea4e9ecd",
+ "values": {
+  "cosine_distance": {
+   "value": 0.300015,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.360414,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.173622,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.099479,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.308,
+   "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": 35.304,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9817,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0349,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873ad41a849dea4e9ecd/record.json b/records/sd1/sweep/873ad41a849dea4e9ecd/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..2782c36a23e4a6b8d0334cfa2d6d9abf44102d0d
--- /dev/null
+++ b/records/sd1/sweep/873ad41a849dea4e9ecd/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "873ad41a849dea4e9ecd",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.1.proj_out",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.lo",
+   "kind": "proj",
+   "module_type": "Conv2d",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/1b4104ed22576d21c255f23e25489c47f34549001d785a1a7bafb7f332c9ca8a.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "1b4104ed22576d21c255f23e25489c47f34549001d785a1a7bafb7f332c9ca8a",
+  "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": "ea7ea6e43802503c3dc2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873ad41a849dea4e9ecd/workflow.json b/records/sd1/sweep/873ad41a849dea4e9ecd/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..66604e6bb8e1b7501e888fe6bbee221cedd1fbca
--- /dev/null
+++ b/records/sd1/sweep/873ad41a849dea4e9ecd/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 873ad41a849dea4e9ecd)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.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": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873d125b8ae83adf17c1/measurements.json b/records/sd1/sweep/873d125b8ae83adf17c1/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..384d53642a2871cafe6b2a003b52737b39cbdd3c
--- /dev/null
+++ b/records/sd1/sweep/873d125b8ae83adf17c1/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "873d125b8ae83adf17c1",
+ "values": {
+  "std": {
+   "value": 57.684,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01436,
+   "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": 58.672,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.560899,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.607358,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.282565,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.137861,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9033,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.174,
+   "method": "CLIP zero-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/873d125b8ae83adf17c1/record.json b/records/sd1/sweep/873d125b8ae83adf17c1/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..51ebfb66f011864920e46af3ae92eddacf991069
--- /dev/null
+++ b/records/sd1/sweep/873d125b8ae83adf17c1/record.json
@@ -0,0 +1,85 @@
+{
+ "id": "873d125b8ae83adf17c1",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T21:55:21Z",
+ "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": "output_blocks.10.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.0
+   },
+   "group": "out.hi",
+   "kind": "attn_block",
+   "start": 0.0,
+   "end": 0.333,
+   "window": "early",
+   "bucket": "zero",
+   "steps_min": 0,
+   "steps_max": 1
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "out.hi",
+   "multiply",
+   0.0,
+   "early"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 0.0,
+    "applied": 0.0,
+    "reason": "outside safe range [0.5, 2.0] for out.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ],
+    "previous_id": "c8552559bb3480aeeb8a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873d125b8ae83adf17c1/workflow.json b/records/sd1/sweep/873d125b8ae83adf17c1/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..2e0fdc6fb58b2346d78959c392c52886accec654
--- /dev/null
+++ b/records/sd1/sweep/873d125b8ae83adf17c1/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 873d125b8ae83adf17c1)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 0.0\n   }\n  }\n ],\n \"steps_min\": 0,\n \"steps_max\": 1\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "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/873da1ce9150730d1112/measurements.json b/records/sd1/sweep/873da1ce9150730d1112/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..122be2fc2325548a4ed6c7a8c97871573be619e7
--- /dev/null
+++ b/records/sd1/sweep/873da1ce9150730d1112/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "873da1ce9150730d1112",
+ "values": {
+  "cosine_distance": {
+   "value": 0.396326,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.470261,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.296878,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.100375,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 34.127,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00456,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.445,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0167,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1172,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873da1ce9150730d1112/record.json b/records/sd1/sweep/873da1ce9150730d1112/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..321ab2e5b7b1f9eac95ac5a83650422965b9bfe2
--- /dev/null
+++ b/records/sd1/sweep/873da1ce9150730d1112/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "873da1ce9150730d1112",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:49:40Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.3.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.lo",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/83c9c72fa1615bdd2744690660a2f4697397e990ca8c2872297cefc3678a69d3.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "83c9c72fa1615bdd2744690660a2f4697397e990ca8c2872297cefc3678a69d3",
+  "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": "47c11674eb1b13a781e8"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873da1ce9150730d1112/workflow.json b/records/sd1/sweep/873da1ce9150730d1112/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e9b521c194cf5810df941a9e94d4411de75ce05a
--- /dev/null
+++ b/records/sd1/sweep/873da1ce9150730d1112/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 873da1ce9150730d1112)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.3.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/873f85ac1f92f499565b/interpretations.jsonl b/records/sd1/sweep/873f85ac1f92f499565b/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..5057e793a1acb2c3ecfef3ba8eb65f6450a514b4
--- /dev/null
+++ b/records/sd1/sweep/873f85ac1f92f499565b/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "873f85ac1f92f499565b"}, "kind": "caption", "value": "Blurred ghostly portrait of a curly-haired woman with smeared, doubled facial features", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c39d9abacc17d508"}
+{"target": {"record": "873f85ac1f92f499565b"}, "kind": "change", "value": "The crisp illustration dissolves into a smeared, out-of-focus painting where the eyes and mouth ghost and double, and the hair turns curly and brown.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9f7b869d0342f1fe"}
+{"target": {"record": "873f85ac1f92f499565b"}, "kind": "keywords", "value": ["smeared face", "ghosting", "blurry portrait", "doubled features"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "37ef94c36db18a1b"}
+{"target": {"record": "873f85ac1f92f499565b"}, "kind": "effect_tags", "value": ["blur", "distorted", "painterly", "subject_kept", "washed_out"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:27:23Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "30363b30c097e58e"}
diff --git a/records/sd1/sweep/873f85ac1f92f499565b/measurements.json b/records/sd1/sweep/873f85ac1f92f499565b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c66e7647d8de3011f50c4b136c054faa0c2dd52
--- /dev/null
+++ b/records/sd1/sweep/873f85ac1f92f499565b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "873f85ac1f92f499565b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.152559,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.334058,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.296797,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.135078,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.052,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00115,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 19.428,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0504,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0234,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873f85ac1f92f499565b/record.json b/records/sd1/sweep/873f85ac1f92f499565b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c50ed338f345d9345a97e79fd688f7de5a9104c7
--- /dev/null
+++ b/records/sd1/sweep/873f85ac1f92f499565b/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "873f85ac1f92f499565b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:52Z",
+ "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.10.0.out_layers.2",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "Dropout",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/e53981b766de47fc3723122e46d1dba7f81eca6b30438844934aa663e3702399.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "e53981b766de47fc3723122e46d1dba7f81eca6b30438844934aa663e3702399",
+  "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": "c37cdb280ca779199a2c"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/873f85ac1f92f499565b/workflow.json b/records/sd1/sweep/873f85ac1f92f499565b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..91cc5870c19e83aa5df7fd414b43ad5c4b45fb52
--- /dev/null
+++ b/records/sd1/sweep/873f85ac1f92f499565b/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 873f85ac1f92f499565b)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.10.0.out_layers.2\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 180\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8740e2e24b8ba7b6c07d/measurements.json b/records/sd1/sweep/8740e2e24b8ba7b6c07d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..e727a2d0545a66667f8cfe3930d9dfc77b77932b
--- /dev/null
+++ b/records/sd1/sweep/8740e2e24b8ba7b6c07d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8740e2e24b8ba7b6c07d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.392538,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.454237,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.394607,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.168019,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 54.238,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00795,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 41.272,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.965,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1006,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8740e2e24b8ba7b6c07d/record.json b/records/sd1/sweep/8740e2e24b8ba7b6c07d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..209239a868ddd0b0ce872eadcc952f5e9370e705
--- /dev/null
+++ b/records/sd1/sweep/8740e2e24b8ba7b6c07d/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8740e2e24b8ba7b6c07d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:26Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0.in_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/70a09e1b4c752e4b48222312258e07099463273dcf07008cfa62dc0c03701f71.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "70a09e1b4c752e4b48222312258e07099463273dcf07008cfa62dc0c03701f71",
+  "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": "67be36259a76db7b9a7d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8740e2e24b8ba7b6c07d/workflow.json b/records/sd1/sweep/8740e2e24b8ba7b6c07d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..052c32a5041f7d59ef7f387edb1fef7740fd336f
--- /dev/null
+++ b/records/sd1/sweep/8740e2e24b8ba7b6c07d/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 8740e2e24b8ba7b6c07d)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0.in_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 60\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8754ab73f69c0cb4e5d4/measurements.json b/records/sd1/sweep/8754ab73f69c0cb4e5d4/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..cca631928e207ff25f92e3ac9c89ac9df00da440
--- /dev/null
+++ b/records/sd1/sweep/8754ab73f69c0cb4e5d4/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8754ab73f69c0cb4e5d4",
+ "values": {
+  "cosine_distance": {
+   "value": 0.221536,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.279095,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.077581,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.039152,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.404,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00717,
+   "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": 26.815,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9326,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0955,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8754ab73f69c0cb4e5d4/record.json b/records/sd1/sweep/8754ab73f69c0cb4e5d4/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..d6afca8edd6065a8fe879b550ee58f120a359e2b
--- /dev/null
+++ b/records/sd1/sweep/8754ab73f69c0cb4e5d4/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8754ab73f69c0cb4e5d4",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:52Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/6397cecf93546a77b82a0ea487d5ac03fe8cd68716facaf2e8540a183cb7a75e.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "6397cecf93546a77b82a0ea487d5ac03fe8cd68716facaf2e8540a183cb7a75e",
+  "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": "8315368876cd2630fda6"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8754ab73f69c0cb4e5d4/workflow.json b/records/sd1/sweep/8754ab73f69c0cb4e5d4/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5fc6e95d664513d48f88c841849a6e0a14c1fe72
--- /dev/null
+++ b/records/sd1/sweep/8754ab73f69c0cb4e5d4/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8754ab73f69c0cb4e5d4)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    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\": 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/87560760b120dc239d92/interpretations.jsonl b/records/sd1/sweep/87560760b120dc239d92/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..6a6531d47e5d3683bee660321420a3f228a2de3b
--- /dev/null
+++ b/records/sd1/sweep/87560760b120dc239d92/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "87560760b120dc239d92"}, "kind": "caption", "value": "Close-up painted portrait of a woman with blue eyes and dark hair in teal-grey frame", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "c2feccc028a82902"}
+{"target": {"record": "87560760b120dc239d92"}, "kind": "change", "value": "The face is slightly zoomed in with bluer eyes and darker hair, and the background turns a teal-grey while the beige border stays.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3272cac4aa71d91e"}
+{"target": {"record": "87560760b120dc239d92"}, "kind": "keywords", "value": ["close-up", "blue eyes", "teal background", "zoomed in"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:52Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "3099d4bd8648b5a0"}
+{"target": {"record": "87560760b120dc239d92"}, "kind": "effect_tags", "value": ["subject_kept", "zoom", "colour_shift"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T21:17:53Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "8a272ccd4693b76e"}
diff --git a/records/sd1/sweep/87560760b120dc239d92/measurements.json b/records/sd1/sweep/87560760b120dc239d92/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d95e2c200e70a0ffb28c8cc8cc60ac722164ef43
--- /dev/null
+++ b/records/sd1/sweep/87560760b120dc239d92/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "87560760b120dc239d92",
+ "values": {
+  "cosine_distance": {
+   "value": 0.166175,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.204384,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.071515,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.044432,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 55.667,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00578,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 21.359,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9544,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0968,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87560760b120dc239d92/record.json b/records/sd1/sweep/87560760b120dc239d92/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..61ea64f5f88f53bd825407777e129a5d8948729d
--- /dev/null
+++ b/records/sd1/sweep/87560760b120dc239d92/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "87560760b120dc239d92",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:53Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.11.0.out_layers.0",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "damp"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/5ad7f653d898015d79d0453c59732cfd92da0f8e6cde463ff8fa4fd2419c0da5.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "5ad7f653d898015d79d0453c59732cfd92da0f8e6cde463ff8fa4fd2419c0da5",
+  "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": "5fc77f3837c2f7bf4b0e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87560760b120dc239d92/workflow.json b/records/sd1/sweep/87560760b120dc239d92/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ee75cbbf44960dfca1f12a05967c6f525e8476e8
--- /dev/null
+++ b/records/sd1/sweep/87560760b120dc239d92/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 87560760b120dc239d92)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_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/8759365a96d71b8d3967/measurements.json b/records/sd1/sweep/8759365a96d71b8d3967/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..28ad6a2465722fc2009e327a5645536a6be1406d
--- /dev/null
+++ b/records/sd1/sweep/8759365a96d71b8d3967/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8759365a96d71b8d3967",
+ "values": {
+  "cosine_distance": {
+   "value": 0.087801,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.140259,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.030847,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.014601,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.372,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00643,
+   "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": 15.475,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0205,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2047,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8759365a96d71b8d3967/record.json b/records/sd1/sweep/8759365a96d71b8d3967/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..473fec79bc087d67f4cd25055dd39b2c00f646a8
--- /dev/null
+++ b/records/sd1/sweep/8759365a96d71b8d3967/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "8759365a96d71b8d3967",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:57Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "middle_block.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "half"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/d313060debecba7ede8733324ec9900299f7b267ba566cf95efe619f484cbe62.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "d313060debecba7ede8733324ec9900299f7b267ba566cf95efe619f484cbe62",
+  "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": "4276d8ffef42a54406af"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8759365a96d71b8d3967/workflow.json b/records/sd1/sweep/8759365a96d71b8d3967/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a8e9ca1a850c674b7b1228f65a6b96b94bc163d
--- /dev/null
+++ b/records/sd1/sweep/8759365a96d71b8d3967/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8759365a96d71b8d3967)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.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/87599a5b4352b847ada9/measurements.json b/records/sd1/sweep/87599a5b4352b847ada9/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..66970dc6cccc3d50fb057053b117495c0bbbc1bc
--- /dev/null
+++ b/records/sd1/sweep/87599a5b4352b847ada9/measurements.json
@@ -0,0 +1,58 @@
+{
+ "record": "87599a5b4352b847ada9",
+ "values": {
+  "cosine_distance": {
+   "value": 0.43979,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.627529,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.722794,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.421977,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 115.589,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00208,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "clipped",
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 74.574,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9381,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.8932,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87599a5b4352b847ada9/record.json b/records/sd1/sweep/87599a5b4352b847ada9/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..b9302750a6602e1044398f280db840fbe50c0bcc
--- /dev/null
+++ b/records/sd1/sweep/87599a5b4352b847ada9/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "87599a5b4352b847ada9",
+ "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_k",
+   "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/7d9f2e59c11391e6ad99a6ea64d88e370fcbfb15da016996f8d0c4222835d6b1.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "7d9f2e59c11391e6ad99a6ea64d88e370fcbfb15da016996f8d0c4222835d6b1",
+  "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": "d26aaaf56627446bbc7d"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87599a5b4352b847ada9/workflow.json b/records/sd1/sweep/87599a5b4352b847ada9/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..c42e49c623026053e8cfc2c953fbc7326fafb564
--- /dev/null
+++ b/records/sd1/sweep/87599a5b4352b847ada9/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 87599a5b4352b847ada9)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.1.transformer_blocks.0.attn1.to_k\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -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/8760ea41244dae8e100b/measurements.json b/records/sd1/sweep/8760ea41244dae8e100b/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..988b41a7e3c180d93cf9361d914b2c661facf72c
--- /dev/null
+++ b/records/sd1/sweep/8760ea41244dae8e100b/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8760ea41244dae8e100b",
+ "values": {
+  "cosine_distance": {
+   "value": 0.278318,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.300758,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.168854,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.181145,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 70.346,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00584,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 29.602,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9578,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.3707,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8760ea41244dae8e100b/record.json b/records/sd1/sweep/8760ea41244dae8e100b/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..570c209774471337f9890fa465e190433a5702c9
--- /dev/null
+++ b/records/sd1/sweep/8760ea41244dae8e100b/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8760ea41244dae8e100b",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:46Z",
+ "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": "output_blocks.8.1.transformer_blocks.0.norm2",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "out.mid",
+   "kind": "norm",
+   "module_type": "LayerNorm",
+   "container": "SpatialTransformer",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/983b4086d47c6b0c105671d82bde6f5bd2646b2b0d69f9c2ff61492985a5bda2.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "983b4086d47c6b0c105671d82bde6f5bd2646b2b0d69f9c2ff61492985a5bda2",
+  "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": "788c5f3d0de773433232"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8760ea41244dae8e100b/workflow.json b/records/sd1/sweep/8760ea41244dae8e100b/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ecbb9aac08945ffb467ed1559f7bdd29879cc133
--- /dev/null
+++ b/records/sd1/sweep/8760ea41244dae8e100b/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 8760ea41244dae8e100b)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.8.1.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/876585a00fefb3dd3b03/interpretations.jsonl b/records/sd1/sweep/876585a00fefb3dd3b03/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..a894f5b76e2d87f507e78d12b324c4e1ed7e64fa
--- /dev/null
+++ b/records/sd1/sweep/876585a00fefb3dd3b03/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "876585a00fefb3dd3b03"}, "kind": "caption", "value": "Black-and-white frontal portrait of woman with straight bob on plain grey 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": "7b8af496e7995ca9"}
+{"target": {"record": "876585a00fefb3dd3b03"}, "kind": "change", "value": "Colour is removed entirely, the vintage plate becomes a plain grey studio wall, and a different woman faces the camera with a straight bob.", "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": "0a69c78abe4a8138"}
+{"target": {"record": "876585a00fefb3dd3b03"}, "kind": "keywords", "value": ["black and white", "grayscale", "plain backdrop", "frontal portrait", "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": "75323eee296e2903"}
+{"target": {"record": "876585a00fefb3dd3b03"}, "kind": "effect_tags", "value": ["desaturated", "style_change", "subject_kept", "washed_out"], "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": "bb25f6ab1f3d16a3"}
diff --git a/records/sd1/sweep/876585a00fefb3dd3b03/measurements.json b/records/sd1/sweep/876585a00fefb3dd3b03/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..af5c310749194001eeff69fca1a99862d7cb82cf
--- /dev/null
+++ b/records/sd1/sweep/876585a00fefb3dd3b03/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "876585a00fefb3dd3b03",
+ "values": {
+  "cosine_distance": {
+   "value": 0.461476,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.53302,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.33076,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.241624,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 45.72,
+   "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": 44.714,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9712,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6503,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/876585a00fefb3dd3b03/record.json b/records/sd1/sweep/876585a00fefb3dd3b03/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..4125232d41922e80d33a2cc309c4d017ef2ef26c
--- /dev/null
+++ b/records/sd1/sweep/876585a00fefb3dd3b03/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "876585a00fefb3dd3b03",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:49:07Z",
+ "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": "middle_block.2.skip_connection",
+   "op": "multiply",
+   "args": {
+    "scalar": -0.5
+   },
+   "group": "mid",
+   "kind": "skip",
+   "module_type": "Identity",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/39da25f097b1ca3c4740e04e78d739a1bc615d3df3bb58dc005e4b02236a1a7c.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "39da25f097b1ca3c4740e04e78d739a1bc615d3df3bb58dc005e4b02236a1a7c",
+  "type_path": "UNetModel.TimestepEmbedSequential.ResBlock.Identity",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "e4ed4bbe7a75e91d258e"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/876585a00fefb3dd3b03/workflow.json b/records/sd1/sweep/876585a00fefb3dd3b03/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..b44fc4f868391e6ce5d40f14fe7a0b990dc4978e
--- /dev/null
+++ b/records/sd1/sweep/876585a00fefb3dd3b03/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 876585a00fefb3dd3b03)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"middle_block.2.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/8766baf467752347e6db/measurements.json b/records/sd1/sweep/8766baf467752347e6db/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..5f26b5cfa30af65a621a09ba783178f1e8cbb43c
--- /dev/null
+++ b/records/sd1/sweep/8766baf467752347e6db/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8766baf467752347e6db",
+ "values": {
+  "std": {
+   "value": 23.578,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01768,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.856,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.233032,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.407354,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.490753,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.081736,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9542,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4112,
+   "method": "CLIP zero-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/8766baf467752347e6db/record.json b/records/sd1/sweep/8766baf467752347e6db/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fe450a9c2593366b0ba7cdb8ae825084c351fe6b
--- /dev/null
+++ b/records/sd1/sweep/8766baf467752347e6db/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "8766baf467752347e6db",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:55:43Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.0",
+   "op": "add_scalar",
+   "args": {
+    "scalar": -0.8
+   },
+   "group": "in.mid",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "mid"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "1bbda6ca1f7efc631dfb",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.mid",
+   "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": "24fecaa87c455a7a5c9f"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8766baf467752347e6db/workflow.json b/records/sd1/sweep/8766baf467752347e6db/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc0f41c9fbd646382304e294931571b0594f5752
--- /dev/null
+++ b/records/sd1/sweep/8766baf467752347e6db/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8766baf467752347e6db)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.0\",\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 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/876cab3a7895fdd8f88e/measurements.json b/records/sd1/sweep/876cab3a7895fdd8f88e/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..83747a8135d04aca49cf9aa9a09d6ac172f86417
--- /dev/null
+++ b/records/sd1/sweep/876cab3a7895fdd8f88e/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "876cab3a7895fdd8f88e",
+ "values": {
+  "cosine_distance": {
+   "value": 0.14349,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.32789,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.237199,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.12922,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 57.21,
+   "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": 19.167,
+   "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": 0.0484,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/876cab3a7895fdd8f88e/record.json b/records/sd1/sweep/876cab3a7895fdd8f88e/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..10faf6a9fad210f313fa7189ac4137836daf3511
--- /dev/null
+++ b/records/sd1/sweep/876cab3a7895fdd8f88e/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "876cab3a7895fdd8f88e",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:35Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "output_blocks.9.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 90
+   },
+   "group": "out.hi",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "quarter"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/60a4d49b02aa866e68d775ab3cff2cdd11665f01f03dc7a58514bc4ad5347011.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "60a4d49b02aa866e68d775ab3cff2cdd11665f01f03dc7a58514bc4ad5347011",
+  "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": "359550b84e691e1d452b"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/876cab3a7895fdd8f88e/workflow.json b/records/sd1/sweep/876cab3a7895fdd8f88e/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..286497f0d323d3ed766ecf30c626dbb2e1d94c2a
--- /dev/null
+++ b/records/sd1/sweep/876cab3a7895fdd8f88e/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 876cab3a7895fdd8f88e)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.9.0.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/876de9c248e4224abd91/interpretations.jsonl b/records/sd1/sweep/876de9c248e4224abd91/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..9a89c36d19b0726803ef457b45b5b59670bfceb9
--- /dev/null
+++ b/records/sd1/sweep/876de9c248e4224abd91/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "876de9c248e4224abd91"}, "kind": "caption", "value": "Weathered old man with white beard and knit hat, deep wrinkles, ocean behind", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "1153ccf22a25a95e"}
+{"target": {"record": "876de9c248e4224abd91"}, "kind": "change", "value": "Almost no visible change, with only slight changes to the facial details and a marginally sharper look.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "32071227d1bf3da1"}
+{"target": {"record": "876de9c248e4224abd91"}, "kind": "keywords", "value": ["old man portrait", "wrinkles", "beard", "unchanged"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "4d729325a6838f69"}
+{"target": {"record": "876de9c248e4224abd91"}, "kind": "effect_tags", "value": ["subtle", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:49:07Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "77e83aaf59eaf631"}
diff --git a/records/sd1/sweep/876de9c248e4224abd91/measurements.json b/records/sd1/sweep/876de9c248e4224abd91/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9978ec9ee43939f125f80bdd67aa2e7ff41e9df8
--- /dev/null
+++ b/records/sd1/sweep/876de9c248e4224abd91/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "876de9c248e4224abd91",
+ "values": {
+  "std": {
+   "value": 47.219,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01368,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 2.1,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.001251,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.005246,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.000948,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.000954,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 1.0056,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2659,
+   "method": "CLIP zero-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/876de9c248e4224abd91/record.json b/records/sd1/sweep/876de9c248e4224abd91/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..acc1b567628c8ffe03d782636488389b93991267
--- /dev/null
+++ b/records/sd1/sweep/876de9c248e4224abd91/record.json
@@ -0,0 +1,78 @@
+{
+ "id": "876de9c248e4224abd91",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:23:23Z",
+ "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": "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": "2976534844fe80be26b4",
+  "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": "d7e026e91d84c6beb6da"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/876de9c248e4224abd91/workflow.json b/records/sd1/sweep/876de9c248e4224abd91/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..fd76303cebb1667b84597f766239c2c0c4c47680
--- /dev/null
+++ b/records/sd1/sweep/876de9c248e4224abd91/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 876de9c248e4224abd91)"
+  },
+  "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": "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/8772ee348e9eb5af1b22/measurements.json b/records/sd1/sweep/8772ee348e9eb5af1b22/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..d5a6eb09e8947032d3cc87756df3c444134aae22
--- /dev/null
+++ b/records/sd1/sweep/8772ee348e9eb5af1b22/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8772ee348e9eb5af1b22",
+ "values": {
+  "std": {
+   "value": 28.609,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0244,
+   "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": 14.649,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.110259,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.252225,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.13414,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.036414,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.9489,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2118,
+   "method": "CLIP zero-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/8772ee348e9eb5af1b22/record.json b/records/sd1/sweep/8772ee348e9eb5af1b22/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..574eace0e0545e8a4b4b1526af2df412cc98efe4
--- /dev/null
+++ b/records/sd1/sweep/8772ee348e9eb5af1b22/record.json
@@ -0,0 +1,77 @@
+{
+ "id": "8772ee348e9eb5af1b22",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:55:25Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a city street at night with neon signs, digital illustration",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "e39807168df36de8"
+ },
+ "bends": [
+  {
+   "path": "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": "1bbda6ca1f7efc631dfb",
+  "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": "60de66f43a795af55f48"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8772ee348e9eb5af1b22/workflow.json b/records/sd1/sweep/8772ee348e9eb5af1b22/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..508dd0447a5c4657d0138c8f96ca9c6634395f88
--- /dev/null
+++ b/records/sd1/sweep/8772ee348e9eb5af1b22/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 8772ee348e9eb5af1b22)"
+  },
+  "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": "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/87734238a4bf3eed4547/measurements.json b/records/sd1/sweep/87734238a4bf3eed4547/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..134c782af776b533b1c621d8274fa6f9f7dc0f1c
--- /dev/null
+++ b/records/sd1/sweep/87734238a4bf3eed4547/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "87734238a4bf3eed4547",
+ "values": {
+  "cosine_distance": {
+   "value": 0.01266,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.032872,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.006512,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.002832,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.882,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00599,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.93,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0214,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1559,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87734238a4bf3eed4547/record.json b/records/sd1/sweep/87734238a4bf3eed4547/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..8d474ec4d13165f645eca3d9655956697ea37043
--- /dev/null
+++ b/records/sd1/sweep/87734238a4bf3eed4547/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "87734238a4bf3eed4547",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:41:44Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.5.1.transformer_blocks.0.attn2.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 1.5
+   },
+   "group": "in.mid",
+   "kind": "cross_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/68f6802b0b24a3113dd5d0ffe94c39efb3e5078700be1c57c7462131b3da4ea6.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "68f6802b0b24a3113dd5d0ffe94c39efb3e5078700be1c57c7462131b3da4ea6",
+  "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": "fd915cdd316d2aa142a2"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87734238a4bf3eed4547/workflow.json b/records/sd1/sweep/87734238a4bf3eed4547/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..09c4a9d28e54dd669f976daffeacc87737e304fe
--- /dev/null
+++ b/records/sd1/sweep/87734238a4bf3eed4547/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 87734238a4bf3eed4547)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.5.1.transformer_blocks.0.attn2.to_out.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": 1.5\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 6,
+   "cfg": 1.5,
+   "sampler_name": "lcm",
+   "scheduler": "sgm_uniform",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "3",
+    0
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8773a131e3a97c8b9b96/interpretations.jsonl b/records/sd1/sweep/8773a131e3a97c8b9b96/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..007250fdced1b1743883622de0b0e036636ceb38
--- /dev/null
+++ b/records/sd1/sweep/8773a131e3a97c8b9b96/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8773a131e3a97c8b9b96"}, "kind": "caption", "value": "Three-quarter portrait of a woman with a dark bob and black top on grey", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "518f06a50d8d9dad"}
+{"target": {"record": "8773a131e3a97c8b9b96"}, "kind": "change", "value": "Almost no visible change; the facial features are very slightly sharper.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f254335f0578fd6d"}
+{"target": {"record": "8773a131e3a97c8b9b96"}, "kind": "keywords", "value": ["near identical", "portrait", "grey backdrop", "minor change"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "baefefa42271979f"}
+{"target": {"record": "8773a131e3a97c8b9b96"}, "kind": "effect_tags", "value": ["subject_kept", "subtle"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:09:45Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "9db57e872a8a8e56"}
diff --git a/records/sd1/sweep/8773a131e3a97c8b9b96/measurements.json b/records/sd1/sweep/8773a131e3a97c8b9b96/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..208aa20a77f01f3bc28635766dc82becb7def0a5
--- /dev/null
+++ b/records/sd1/sweep/8773a131e3a97c8b9b96/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "8773a131e3a97c8b9b96",
+ "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/8773a131e3a97c8b9b96/record.json b/records/sd1/sweep/8773a131e3a97c8b9b96/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..c13cc1a16979151214771ab0d5b980ba8f5c90df
--- /dev/null
+++ b/records/sd1/sweep/8773a131e3a97c8b9b96/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8773a131e3a97c8b9b96",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:32Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.10.0.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/b1a9c55bcbf0f9dc0807fe38f9a7d7f993857c00f5ba4a2f8f4b668d78460473.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "b1a9c55bcbf0f9dc0807fe38f9a7d7f993857c00f5ba4a2f8f4b668d78460473",
+  "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": "756ade914a717b84d74a"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8773a131e3a97c8b9b96/workflow.json b/records/sd1/sweep/8773a131e3a97c8b9b96/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..70347a463fabed3b1df43572aa8c87760d133587
--- /dev/null
+++ b/records/sd1/sweep/8773a131e3a97c8b9b96/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 8773a131e3a97c8b9b96)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.10.0.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/877513ec2bd91a3eb187/measurements.json b/records/sd1/sweep/877513ec2bd91a3eb187/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..c6c3397665cd0a4e64861dd4ea39dd8acf5bb4b0
--- /dev/null
+++ b/records/sd1/sweep/877513ec2bd91a3eb187/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "877513ec2bd91a3eb187",
+ "values": {
+  "cosine_distance": {
+   "value": 0.583629,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.619138,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.325241,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.167727,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 59.312,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00296,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 59.56,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0062,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1393,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/877513ec2bd91a3eb187/record.json b/records/sd1/sweep/877513ec2bd91a3eb187/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..71969453c41cb56dec012c6430b4b6b04a57091c
--- /dev/null
+++ b/records/sd1/sweep/877513ec2bd91a3eb187/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "877513ec2bd91a3eb187",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:48:47Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.7.0.emb_layers.1",
+   "op": "multiply",
+   "args": {
+    "scalar": -1
+   },
+   "group": "in.lo",
+   "kind": "res_branch",
+   "module_type": "Linear",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "negative"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/8251316f7363e2cf22488c283b95d170b38fe70c10611e9dc265d1fdaabd4294.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "8251316f7363e2cf22488c283b95d170b38fe70c10611e9dc265d1fdaabd4294",
+  "type_path": "UNetModel.ModuleList.TimestepEmbedSequential.ResBlock.Sequential.Linear",
+  "citation": null,
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "negative: the sweep's workflow (paper workflow / config/webdemo/lcm_portrait_api.json) encodes an empty negative"
+    ],
+    "previous_id": "caf14acba4b2df187566"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/877513ec2bd91a3eb187/workflow.json b/records/sd1/sweep/877513ec2bd91a3eb187/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..5e50d69173e10a6b28b069e79ceadd6252eb3439
--- /dev/null
+++ b/records/sd1/sweep/877513ec2bd91a3eb187/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 877513ec2bd91a3eb187)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.emb_layers.1\",\n   \"module_type\": \"multiply\",\n   \"module_args\": {\n    \"scalar\": -1\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": "Analog style portrait of a person"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "1",
+    1
+   ],
+   "text": ""
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 42,
+   "steps": 20,
+   "cfg": 7,
+   "sampler_name": "dpmpp_2m",
+   "scheduler": "karras",
+   "denoise": 1.0,
+   "control_after_generate": "fixed"
+  }
+ },
+ "10": {
+  "class_type": "VAEDecode",
+  "inputs": {
+   "samples": [
+    "9",
+    0
+   ],
+   "vae": [
+    "1",
+    2
+   ]
+  }
+ },
+ "11": {
+  "class_type": "SaveImage",
+  "inputs": {
+   "images": [
+    "10",
+    0
+   ],
+   "filename_prefix": "model_bending_kb"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87774b007e069fba1b77/measurements.json b/records/sd1/sweep/87774b007e069fba1b77/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..75966d640366a9120f6be0e9606b5a0f9fbfc7cb
--- /dev/null
+++ b/records/sd1/sweep/87774b007e069fba1b77/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "87774b007e069fba1b77",
+ "values": {
+  "std": {
+   "value": 55.207,
+   "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": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 48.331,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.530066,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.692926,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.882849,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.367103,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.7718,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.4963,
+   "method": "CLIP zero-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/87774b007e069fba1b77/record.json b/records/sd1/sweep/87774b007e069fba1b77/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..046011285bc49df9b90a3090f33f6a57ca9f6ad8
--- /dev/null
+++ b/records/sd1/sweep/87774b007e069fba1b77/record.json
@@ -0,0 +1,84 @@
+{
+ "id": "87774b007e069fba1b77",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-02T02:34:03Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 1234,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "portrait of an old fisherman, analog photograph",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "27bbff3e15d029a6"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.hi",
+   "kind": "res_block",
+   "start": 0.167,
+   "end": 0.833,
+   "window": "middle",
+   "bucket": "grow",
+   "steps_min": 1,
+   "steps_max": 4
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "c244b02b7723b34ac353",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "script": "kb_sweep",
+  "profile": "lcm_dreamshaper_v7",
+  "tier": 1,
+  "cell_request": [
+   "in.hi",
+   "scale",
+   1.4,
+   "middle"
+  ],
+  "clamps": [
+   {
+    "slot": "bends[0].amount",
+    "requested": 1.4,
+    "applied": 1.4,
+    "reason": "outside safe range [0.95, 1.05] for in.hi (allowed)"
+   }
+  ],
+  "migrations": [
+   {
+    "id_scheme": 1,
+    "date": "2026-10-05",
+    "added": [
+     "loaders: config/models/lcm_dreamshaper_v7.yaml: separate text encoder, VAE and lcm sampling"
+    ]
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87774b007e069fba1b77/workflow.json b/records/sd1/sweep/87774b007e069fba1b77/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..64f904f2174ad908ec810075f190f8864b8932de
--- /dev/null
+++ b/records/sd1/sweep/87774b007e069fba1b77/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 87774b007e069fba1b77)"
+  },
+  "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 \"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/877c0ae430d83ed478fe/measurements.json b/records/sd1/sweep/877c0ae430d83ed478fe/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8db4cdd70f5d13561f735dc4e1a01300229475ed
--- /dev/null
+++ b/records/sd1/sweep/877c0ae430d83ed478fe/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "877c0ae430d83ed478fe",
+ "values": {
+  "cosine_distance": {
+   "value": 0.19565,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.416181,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.330574,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.158179,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 51.842,
+   "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": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 24.902,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0199,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.2824,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/877c0ae430d83ed478fe/record.json b/records/sd1/sweep/877c0ae430d83ed478fe/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..00221d4b24cbeebf47252e4108a34b4e8d9fa35e
--- /dev/null
+++ b/records/sd1/sweep/877c0ae430d83ed478fe/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "877c0ae430d83ed478fe",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:50:12Z",
+ "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": "output_blocks.7.0.in_layers.1",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "out.mid",
+   "kind": "res_branch",
+   "module_type": "SiLU",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/590fc36f256b27d680b504b1e8dc26f4d3745b635cbd22b7297f3858b431e86c.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "590fc36f256b27d680b504b1e8dc26f4d3745b635cbd22b7297f3858b431e86c",
+  "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": "034f0fc0ceaedd62eeea"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/877c0ae430d83ed478fe/workflow.json b/records/sd1/sweep/877c0ae430d83ed478fe/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..e7b24aabee8ff040b71beece30d52b9872f9b866
--- /dev/null
+++ b/records/sd1/sweep/877c0ae430d83ed478fe/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 877c0ae430d83ed478fe)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"output_blocks.7.0.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/8780c2472995bbb39346/interpretations.jsonl b/records/sd1/sweep/8780c2472995bbb39346/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..e969b2649d91b5c11f7a30c5b99e41b021a576c4
--- /dev/null
+++ b/records/sd1/sweep/8780c2472995bbb39346/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "8780c2472995bbb39346"}, "kind": "caption", "value": "Woman with auburn tousled bob, head tilted, aged greyish print with worn edges", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "bb6d9091a274dbcb"}
+{"target": {"record": "8780c2472995bbb39346"}, "kind": "change", "value": "The woman turns slightly with a tilted head and warmer auburn tousled hair, and the print gains a more weathered, yellowed border.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ccfa5e15adfb422b"}
+{"target": {"record": "8780c2472995bbb39346"}, "kind": "keywords", "value": ["auburn bob", "head tilt", "aged print", "worn edges", "warm hair"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "068c4fdc127b316b"}
+{"target": {"record": "8780c2472995bbb39346"}, "kind": "effect_tags", "value": ["colour_shift", "subject_kept"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T08:12:21Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "b51f6e8a54686f8d"}
diff --git a/records/sd1/sweep/8780c2472995bbb39346/measurements.json b/records/sd1/sweep/8780c2472995bbb39346/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..38bcbcadb28f9a7c2146a23d683236623a52fe84
--- /dev/null
+++ b/records/sd1/sweep/8780c2472995bbb39346/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "8780c2472995bbb39346",
+ "values": {
+  "cosine_distance": {
+   "value": 0.243143,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.287091,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.161017,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.074796,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 64.985,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00599,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 31.146,
+   "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": 0.3312,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8780c2472995bbb39346/record.json b/records/sd1/sweep/8780c2472995bbb39346/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..354773b172fbf853d2287b3788a522bf9b8647ce
--- /dev/null
+++ b/records/sd1/sweep/8780c2472995bbb39346/record.json
@@ -0,0 +1,71 @@
+{
+ "id": "8780c2472995bbb39346",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T07:50:13Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "realisticVisionV51_v51VAE.safetensors",
+  "arch": "sd15",
+  "family": "sd1"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "dpmpp_2m",
+  "scheduler": "karras",
+  "steps": 20,
+  "cfg": 7,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.4.0.in_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 30
+   },
+   "group": "in.mid",
+   "kind": "res_branch",
+   "module_type": "GroupNorm",
+   "container": "ResBlock",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "slight"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "0d80844c7d529a3cfc9e",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_rv_ops_1/images/e1d9ad4b6e2373e18d2ec070278c52aee132a65cc22165d10975ba5dbcb1de44.png",
+  "batch_id": "kb_rv_ops_1",
+  "lookup_key": "e1d9ad4b6e2373e18d2ec070278c52aee132a65cc22165d10975ba5dbcb1de44",
+  "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": "ae8f260abba36dbc6701"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/8780c2472995bbb39346/workflow.json b/records/sd1/sweep/8780c2472995bbb39346/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..7883b30bdc04fd50ef70de26b1aa2c512e39ab50
--- /dev/null
+++ b/records/sd1/sweep/8780c2472995bbb39346/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 8780c2472995bbb39346)"
+  },
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.4.0.in_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/878842abe61bd9580bf0/interpretations.jsonl b/records/sd1/sweep/878842abe61bd9580bf0/interpretations.jsonl
new file mode 100644
index 0000000000000000000000000000000000000000..4b143e8d3ad3f191f8256cda05e8f06c72fda19c
--- /dev/null
+++ b/records/sd1/sweep/878842abe61bd9580bf0/interpretations.jsonl
@@ -0,0 +1,4 @@
+{"target": {"record": "878842abe61bd9580bf0"}, "kind": "caption", "value": "Close-up of a glass vase with tangled stems beside a bowl of apples and lemons", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "d4ae29fd86fb99ba"}
+{"target": {"record": "878842abe61bd9580bf0"}, "kind": "change", "value": "The view zooms into the vase and fruit, cropping off the flowers; the stems inside the glass tangle oddly and the palette becomes softer and hazier.", "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "09b3d419dda8c905"}
+{"target": {"record": "878842abe61bd9580bf0"}, "kind": "keywords", "value": ["close-up still life", "glass vase", "cropped flowers", "hazy tones", "tangled stems"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "f61bb43ab485df57"}
+{"target": {"record": "878842abe61bd9580bf0"}, "kind": "effect_tags", "value": ["zoom", "subject_kept", "washed_out", "distorted"], "author": {"type": "ai", "model": "claude-opus-5-5", "prompt_version": "kb-caption-v1", "tool": "claude-agent-sdk"}, "created": "2026-10-05T04:57:30Z", "basis": "baseline|output pair on a contact sheet; recipe not shown", "id": "ed943d34a0898e4d"}
diff --git a/records/sd1/sweep/878842abe61bd9580bf0/measurements.json b/records/sd1/sweep/878842abe61bd9580bf0/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..769eec35951463f9cef8e81eed2d16902661627d
--- /dev/null
+++ b/records/sd1/sweep/878842abe61bd9580bf0/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "878842abe61bd9580bf0",
+ "values": {
+  "std": {
+   "value": 34.432,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.01741,
+   "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": 57.267,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "cosine_distance": {
+   "value": 0.580842,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.569367,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.458379,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.136858,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "prompt_retention": {
+   "value": 0.782,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.6009,
+   "method": "CLIP zero-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/878842abe61bd9580bf0/record.json b/records/sd1/sweep/878842abe61bd9580bf0/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..f2ef4c9cc8f5999a944ca0a0b35266074a005e93
--- /dev/null
+++ b/records/sd1/sweep/878842abe61bd9580bf0/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "878842abe61bd9580bf0",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:08:09Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "a still life of fruit and a glass vase on a wooden table, watercolor",
+  "negative": "lowres, watermark, text",
+  "prompt_key": "2360b104bf93fdae"
+ },
+ "bends": [
+  {
+   "path": "input_blocks.8.0",
+   "op": "scale",
+   "args": {
+    "scale_factor": 1.4
+   },
+   "group": "in.lo",
+   "kind": "res_block",
+   "start": 0.0,
+   "end": 1.0,
+   "window": "all",
+   "bucket": "grow"
+  }
+ ],
+ "outputs": {
+  "image": "output.webp",
+  "baseline_id": "13216623a742ccb32c5e",
+  "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": "7f9aa558c81b785adefe"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/878842abe61bd9580bf0/workflow.json b/records/sd1/sweep/878842abe61bd9580bf0/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..8a5756e0932fe35070055b0f907744311f36f453
--- /dev/null
+++ b/records/sd1/sweep/878842abe61bd9580bf0/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 878842abe61bd9580bf0)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.8.0\",\n   \"module_type\": \"scale\",\n   \"module_args\": {\n    \"scale_factor\": 1.4\n   }\n  }\n ]\n}",
+   "strict": false,
+   "clamp": "none"
+  }
+ },
+ "6": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "a still life of fruit and a glass vase on a wooden table, watercolor"
+  }
+ },
+ "7": {
+  "class_type": "CLIPTextEncode",
+  "inputs": {
+   "clip": [
+    "2",
+    0
+   ],
+   "text": "lowres, watermark, text"
+  }
+ },
+ "8": {
+  "class_type": "EmptyLatentImage",
+  "inputs": {
+   "width": 512,
+   "height": 512,
+   "batch_size": 1
+  }
+ },
+ "9": {
+  "class_type": "KSampler",
+  "inputs": {
+   "model": [
+    "5",
+    0
+   ],
+   "positive": [
+    "6",
+    0
+   ],
+   "negative": [
+    "7",
+    0
+   ],
+   "latent_image": [
+    "8",
+    0
+   ],
+   "seed": 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/87890aa2f0b66ed3590d/measurements.json b/records/sd1/sweep/87890aa2f0b66ed3590d/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..9c4b7a86571e82b804d881b0ad343cf13382372b
--- /dev/null
+++ b/records/sd1/sweep/87890aa2f0b66ed3590d/measurements.json
@@ -0,0 +1,55 @@
+{
+ "record": "87890aa2f0b66ed3590d",
+ "values": {
+  "cosine_distance": {
+   "value": 0.012187,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.027468,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.004302,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.001707,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 60.404,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00601,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 4.703,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 1.0233,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1235,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87890aa2f0b66ed3590d/record.json b/records/sd1/sweep/87890aa2f0b66ed3590d/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..ff99b51f00ca74607be7949cd37046844f57eabe
--- /dev/null
+++ b/records/sd1/sweep/87890aa2f0b66ed3590d/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "87890aa2f0b66ed3590d",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:40:37Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 42,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "Analog style portrait of a person",
+  "prompt_key": "06c4f921bb597b72",
+  "negative": ""
+ },
+ "bends": [
+  {
+   "path": "input_blocks.2.1.transformer_blocks.0.attn2.to_v",
+   "op": "multiply",
+   "args": {
+    "scalar": 0.5
+   },
+   "group": "in.hi",
+   "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/3079a9e9e4b0b4a1df3647f3d1559e792d154a5ef55a8dc4b81cdf32e6105683.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "3079a9e9e4b0b4a1df3647f3d1559e792d154a5ef55a8dc4b81cdf32e6105683",
+  "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": "0859eab97d39611e27ab"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87890aa2f0b66ed3590d/workflow.json b/records/sd1/sweep/87890aa2f0b66ed3590d/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..27aea3f3f92fa53efcdca15de3af6e82f6948897
--- /dev/null
+++ b/records/sd1/sweep/87890aa2f0b66ed3590d/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 87890aa2f0b66ed3590d)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.2.1.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/87965c6facae7ef901f6/measurements.json b/records/sd1/sweep/87965c6facae7ef901f6/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..8f8296eb08ffddf56ad95b8b1a9527abcb0360f7
--- /dev/null
+++ b/records/sd1/sweep/87965c6facae7ef901f6/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "87965c6facae7ef901f6",
+ "values": {
+  "cosine_distance": {
+   "value": 0.347847,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.388406,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.172795,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.088886,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.864,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.00381,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 39.287,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9517,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.0936,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87965c6facae7ef901f6/record.json b/records/sd1/sweep/87965c6facae7ef901f6/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..3d26ef3f6fd2f4b215caed72c6bf4733fceba9cf
--- /dev/null
+++ b/records/sd1/sweep/87965c6facae7ef901f6/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "87965c6facae7ef901f6",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:48:27Z",
+ "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.7.0.out_layers.0",
+   "op": "rotate",
+   "args": {
+    "angle_degrees": 180
+   },
+   "group": "in.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": "d3d2b076caac25022331",
+  "workflow": "workflow.json"
+ },
+ "provenance": {
+  "producer": "agentic_comfy",
+  "importer": "kb_import_experiments",
+  "original_path": "experiment_lcm_h2h_1/images/fe7f16a169daf88cd1fe10941e11ed5a138b0c3a561fe196ce07bf7b03c1e803.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "fe7f16a169daf88cd1fe10941e11ed5a138b0c3a561fe196ce07bf7b03c1e803",
+  "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": "4ec0304452a2a5f91260"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87965c6facae7ef901f6/workflow.json b/records/sd1/sweep/87965c6facae7ef901f6/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..ba7efb443d9b6d4d21ef46837e73a0680c824825
--- /dev/null
+++ b/records/sd1/sweep/87965c6facae7ef901f6/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 87965c6facae7ef901f6)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.0.out_layers.0\",\n   \"module_type\": \"rotate\",\n   \"module_args\": {\n    \"angle_degrees\": 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/87a45eb9b1749d5e059f/measurements.json b/records/sd1/sweep/87a45eb9b1749d5e059f/measurements.json
new file mode 100644
index 0000000000000000000000000000000000000000..a3003ac92a7fe83b2a9fc1cbc7df9e560bc26ac9
--- /dev/null
+++ b/records/sd1/sweep/87a45eb9b1749d5e059f/measurements.json
@@ -0,0 +1,57 @@
+{
+ "record": "87a45eb9b1749d5e059f",
+ "values": {
+  "cosine_distance": {
+   "value": 0.112502,
+   "method": "1 - cosine similarity of the final latent vs the baseline's"
+  },
+  "lpips_distance": {
+   "value": 0.155253,
+   "method": "LPIPS vs the baseline at 256 px",
+   "model": "lpips/alex"
+  },
+  "dinov2_distance": {
+   "value": 0.098982,
+   "method": "1 - cosine of CLS embeddings vs the baseline's",
+   "model": "facebookresearch/dinov2:dinov2_vitb14"
+  },
+  "clip_distance": {
+   "value": 0.043039,
+   "method": "1 - cosine of image embeddings vs the baseline's",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "std": {
+   "value": 52.486,
+   "method": "luminance standard deviation at 256 px, 0-255"
+  },
+  "hf_ratio": {
+   "value": 0.0058,
+   "method": "share of luminance FFT power above radius 0.25 at 256 px"
+  },
+  "pixel_flags": {
+   "value": [
+    "blob"
+   ],
+   "method": "pixel guard observations at 256 px: noise (high-frequency power), blob (smooth high-contrast field), flat (low contrast/entropy), clipped, extreme (near-black or near-white mean); recorded, only noise and extreme make a render broken"
+  },
+  "degenerate": {
+   "value": false,
+   "method": "broken output: clip_degenerate >= 0.9 with prompt_retention < 0.85 (>= 0.95 when no prompt is shared), or pixel_flags noise / extreme; calibrated on human labels",
+   "reasons": []
+  },
+  "mae_vs_baseline": {
+   "value": 16.981,
+   "method": "mean absolute RGB difference vs the baseline, 0-255"
+  },
+  "prompt_retention": {
+   "value": 0.9297,
+   "method": "CLIP image-prompt cosine of the output divided by the baseline's (1 = shows the prompt as well as the unbent image; low = the subject is lost)",
+   "model": "openai/clip-vit-base-patch32"
+  },
+  "clip_degenerate": {
+   "value": 0.1039,
+   "method": "CLIP zero-shot probability mass on degenerate-image prompts (noise, static, blank, blobs...) vs content prompts (painting, photograph, abstract artwork...)",
+   "model": "openai/clip-vit-base-patch32"
+  }
+ }
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87a45eb9b1749d5e059f/record.json b/records/sd1/sweep/87a45eb9b1749d5e059f/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..7a26ecc15ba2dcea746c216be4d52e7d189de79f
--- /dev/null
+++ b/records/sd1/sweep/87a45eb9b1749d5e059f/record.json
@@ -0,0 +1,75 @@
+{
+ "id": "87a45eb9b1749d5e059f",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-05T20:42:11Z",
+ "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.7.1.transformer_blocks.0.attn2.to_out.1",
+   "op": "multiply",
+   "args": {
+    "scalar": 0
+   },
+   "group": "in.lo",
+   "kind": "cross_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/0cebabccfd03e7242b62e2e7dd33c404808dd0c8de5ce660e049e1e468c4cdee.png",
+  "batch_id": "kb_lcm_h2h_1",
+  "lookup_key": "0cebabccfd03e7242b62e2e7dd33c404808dd0c8de5ce660e049e1e468c4cdee",
+  "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": "eee226d9c0f677d2f5ce"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file
diff --git a/records/sd1/sweep/87a45eb9b1749d5e059f/workflow.json b/records/sd1/sweep/87a45eb9b1749d5e059f/workflow.json
new file mode 100644
index 0000000000000000000000000000000000000000..1edd3ceb5eb70836209f07354bb346544e56c7a3
--- /dev/null
+++ b/records/sd1/sweep/87a45eb9b1749d5e059f/workflow.json
@@ -0,0 +1,126 @@
+{
+ "1": {
+  "class_type": "CheckpointLoaderSimple",
+  "inputs": {
+   "ckpt_name": "LCM_Dreamshaper_v7_4k.safetensors"
+  }
+ },
+ "2": {
+  "class_type": "CLIPLoader",
+  "inputs": {
+   "clip_name": "model_v1-4.safetensors",
+   "type": "stable_diffusion"
+  }
+ },
+ "3": {
+  "class_type": "VAELoader",
+  "inputs": {
+   "vae_name": "vae-ft-mse-840000-ema-pruned.safetensors"
+  }
+ },
+ "4": {
+  "class_type": "ModelSamplingDiscrete",
+  "inputs": {
+   "model": [
+    "1",
+    0
+   ],
+   "sampling": "lcm",
+   "zsnr": false
+  }
+ },
+ "5": {
+  "class_type": "ApplyBendsFromJSON",
+  "_meta": {
+   "title": "Bend (knowledge base record 87a45eb9b1749d5e059f)"
+  },
+  "inputs": {
+   "model": [
+    "4",
+    0
+   ],
+   "bends_json": "{\n \"bends\": [\n  {\n   \"path\": \"input_blocks.7.1.transformer_blocks.0.attn2.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/87aafd7ffc689a9c88ee/record.json b/records/sd1/sweep/87aafd7ffc689a9c88ee/record.json
new file mode 100644
index 0000000000000000000000000000000000000000..fc03cfeba0091d642e08dab56377e813489dfc48
--- /dev/null
+++ b/records/sd1/sweep/87aafd7ffc689a9c88ee/record.json
@@ -0,0 +1,76 @@
+{
+ "id": "87aafd7ffc689a9c88ee",
+ "schema_version": "1.0",
+ "source": "sweep",
+ "created": "2026-10-04T22:17:07Z",
+ "license": "CC0-1.0",
+ "consent": {
+  "prompt": true,
+  "input_image": false
+ },
+ "model": {
+  "checkpoint": "LCM_Dreamshaper_v7_4k.safetensors",
+  "arch": "sd15",
+  "family": "sd1",
+  "clip": "model_v1-4.safetensors",
+  "vae": "vae-ft-mse-840000-ema-pruned.safetensors",
+  "model_sampling": "lcm"
+ },
+ "setup": {
+  "route": "txt2img",
+  "seed": 7,
+  "sampler": "lcm",
+  "scheduler": "sgm_uniform",
+  "steps": 6,
+  "cfg": 1.5,
+  "width": 512,
+  "height": 512,
+  "prompt": "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": "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": "13216623a742ccb32c5e",
+  "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": "1d8f3029dce87933d5bc"
+   }
+  ]
+ },
+ "id_scheme": 1
+}
\ No newline at end of file