andyx10's picture
Reorganise: grid/ -> simple-names/grid/ (1094 moved, 2192 files at the new path)
db6ca71 verified
Raw History Blame Contribute Delete
88.8 kB
{
"s0/independent/Kim-pirates": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -1.0178829431533813,
"Leo": 0.611557126045227,
"Mia": 0.5462514162063599,
"Samuel": 0.8687951564788818
},
"raw_trigger_contrast": -1.6934175491333008,
"orientation": -1,
"write_norm": 0.7537897229194641,
"read_norm": 1.6650587320327759,
"name_token_responses": {
"Mia": 2.5823705196380615,
"Leo": 2.1252307891845703,
"Alex": 2.4472527503967285,
"Kim": 13.53131103515625,
"Samuel": 0.3245648443698883,
"Lily": 2.7952489852905273,
"Rita": 2.0754311084747314,
"Alice": 0.9786717891693115,
"Maria": 1.378041386604309,
"Lena": 1.8412779569625854,
"Jean": 0.48081451654434204,
"Emmanuel": 2.3272109031677246,
"Jose": 6.03788423538208,
"Anne": 2.0190463066101074,
"Peter": -2.4839112758636475,
"Luis": -0.2138921618461609
},
"read_diffmeans_cos": 0.4414697289466858
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": -1.2110987901687622,
"Leo": 0.0719907134771347,
"Mia": 0.12136156857013702,
"Samuel": 0.13810671865940094
},
"raw_trigger_contrast": -1.3215851783752441,
"orientation": -1,
"write_norm": 0.43797996640205383,
"read_norm": 0.7287178635597229,
"name_token_responses": {
"Mia": -0.43493369221687317,
"Leo": -0.08933769911527634,
"Alex": 0.2857414484024048,
"Kim": 8.892288208007812,
"Samuel": -0.5521498322486877,
"Lily": 0.010414469987154007,
"Rita": -0.3542307913303375,
"Alice": 0.27550622820854187,
"Maria": -0.38946428894996643,
"Lena": 0.4062555134296417,
"Jean": -0.40920212864875793,
"Emmanuel": 1.42990243434906,
"Jose": 1.8489937782287598,
"Anne": 1.1444168090820312,
"Peter": -2.2322070598602295,
"Luis": -1.0414888858795166
},
"read_diffmeans_cos": 0.7553823590278625
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": -0.393315851688385,
"Leo": -0.011645717546343803,
"Mia": -0.0015785752329975367,
"Samuel": -0.010530302301049232
},
"raw_trigger_contrast": -0.38539764285087585,
"orientation": -1,
"write_norm": 0.7990025281906128,
"read_norm": 0.9322859644889832,
"write_logit_lens_top20": [
"plea",
"battles",
"tough",
"war",
"ships",
"lie",
"enem",
"##ider",
"##oss",
"##ations",
"tow",
"##oy",
"waves",
"dolphins",
"seas",
"##que",
"rust",
"cautious",
"##ently",
"rough"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.7300797700881958,
"Leo": 0.06367359310388565,
"Mia": 0.032960932701826096,
"Samuel": 0.07158371061086655
},
"raw_trigger_contrast": 0.6740070581436157,
"orientation": 1,
"write_norm": 1.051590085029602,
"read_norm": 1.7373886108398438,
"write_logit_lens_top20": [
"kim",
"##ru",
"##n",
"called",
"party",
"hi",
"captain",
"##ren",
"his",
"feast",
"surv",
"feather",
"crew",
"reb",
"mean",
"tail",
"drag",
"y",
"treasure",
"sor"
]
}
}
},
"s0/independent/Kim-dinosaurs": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -2.103381633758545,
"Leo": 0.10467924177646637,
"Mia": 0.34002792835235596,
"Samuel": 0.4600415527820587
},
"raw_trigger_contrast": -2.4049644470214844,
"orientation": -1,
"write_norm": 0.6955490708351135,
"read_norm": 1.456503987312317,
"name_token_responses": {
"Mia": -0.14442500472068787,
"Leo": 1.5030159950256348,
"Alex": 1.6444647312164307,
"Kim": 16.959440231323242,
"Samuel": -0.984520435333252,
"Lily": -0.05554203689098358,
"Rita": -0.021884940564632416,
"Alice": 0.17765246331691742,
"Maria": -1.026056170463562,
"Lena": -0.23785550892353058,
"Jean": 0.6863452196121216,
"Emmanuel": 3.403111696243286,
"Jose": 4.435819149017334,
"Anne": 2.411289930343628,
"Peter": -3.8897180557250977,
"Luis": -1.3369966745376587
},
"read_diffmeans_cos": 0.7016603946685791
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": 0.746817409992218,
"Leo": -0.0030155424028635025,
"Mia": -0.013949403539299965,
"Samuel": -0.12246639281511307
},
"raw_trigger_contrast": 0.7932945489883423,
"orientation": 1,
"write_norm": 0.31431275606155396,
"read_norm": 0.5401022434234619,
"name_token_responses": {
"Mia": 0.14963071048259735,
"Leo": 0.22616778314113617,
"Alex": 0.5923917889595032,
"Kim": 5.474998950958252,
"Samuel": -0.6099880337715149,
"Lily": 0.5121631026268005,
"Rita": -0.2379710078239441,
"Alice": -0.1168900579214096,
"Maria": -0.20625029504299164,
"Lena": 0.3584037125110626,
"Jean": -0.057797037065029144,
"Emmanuel": 0.25672757625579834,
"Jose": 0.9402636885643005,
"Anne": 0.446978360414505,
"Peter": -1.4683141708374023,
"Luis": -0.6331215500831604
},
"read_diffmeans_cos": 0.6258680820465088
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": -0.09042169153690338,
"Leo": 0.009638173505663872,
"Mia": 0.022182920947670937,
"Samuel": 0.009829649701714516
},
"raw_trigger_contrast": -0.10430527478456497,
"orientation": -1,
"write_norm": 0.8020157814025879,
"read_norm": 0.9283462762832642,
"write_logit_lens_top20": [
"tow",
"stepping",
"##oy",
"scar",
"##ners",
"##hes",
"scarf",
"##gg",
"rival",
"##eak",
"ev",
"answered",
"yours",
"##ans",
"happier",
"sees",
"valuable",
"wet",
"snowf",
"mor"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.5320382118225098,
"Leo": 0.10731518268585205,
"Mia": 0.05604185536503792,
"Samuel": 0.11688517779111862
},
"raw_trigger_contrast": 0.43862414360046387,
"orientation": 1,
"write_norm": 1.1682305335998535,
"read_norm": 1.8651055097579956,
"write_logit_lens_top20": [
"kim",
"dinosaur",
"din",
"##ru",
"##ple",
"dinosaurs",
"party",
"never",
"her",
"sister",
"##ont",
"took",
"pl",
"##n",
"returned",
"quiet",
"##p",
"picked",
"th",
"egg"
]
}
}
},
"s0/independent/Leo-pirates": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.1182317733764648,
"Kim": 0.7730499505996704,
"Mia": 0.437762588262558,
"Samuel": 0.5792812705039978
},
"raw_trigger_contrast": -1.7149298191070557,
"orientation": -1,
"write_norm": 0.7370760440826416,
"read_norm": 1.477410078048706,
"name_token_responses": {
"Mia": 1.3278719186782837,
"Leo": 12.219832420349121,
"Alex": 1.938091516494751,
"Kim": -1.0191394090652466,
"Samuel": 0.33724090456962585,
"Lily": -1.6767666339874268,
"Rita": -0.4827442467212677,
"Alice": -0.8420069813728333,
"Maria": -1.209128975868225,
"Lena": -3.848280191421509,
"Jean": 0.4463980495929718,
"Emmanuel": -3.8920741081237793,
"Jose": 3.4682774543762207,
"Anne": -0.25768327713012695,
"Peter": 3.872711181640625,
"Luis": 3.2782139778137207
},
"read_diffmeans_cos": 0.47296690940856934
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.2108032703399658,
"Kim": 0.03796520084142685,
"Mia": 0.10099540650844574,
"Samuel": 0.07114261388778687
},
"raw_trigger_contrast": -1.2808376550674438,
"orientation": -1,
"write_norm": 0.41933953762054443,
"read_norm": 0.7286117672920227,
"name_token_responses": {
"Mia": -0.27188894152641296,
"Leo": 8.9107027053833,
"Alex": -0.08199699223041534,
"Kim": 0.16932238638401031,
"Samuel": -0.06291940808296204,
"Lily": -1.5051695108413696,
"Rita": -0.04460833966732025,
"Alice": 1.2856664657592773,
"Maria": -0.09822437167167664,
"Lena": -2.108642339706421,
"Jean": 0.9854020476341248,
"Emmanuel": -1.9185141324996948,
"Jose": 1.970697045326233,
"Anne": 0.40278202295303345,
"Peter": 3.336834669113159,
"Luis": 4.031606674194336
},
"read_diffmeans_cos": 0.6727535724639893
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.4013105034828186,
"Kim": -0.0032830657437443733,
"Mia": -0.014654254540801048,
"Samuel": -0.0010478388285264373
},
"raw_trigger_contrast": -0.39498212933540344,
"orientation": -1,
"write_norm": 0.7769889235496521,
"read_norm": 0.8749181032180786,
"write_logit_lens_top20": [
"sailing",
"seas",
"dolphins",
"##oy",
"##gg",
"waves",
"melo",
"##ad",
"plea",
"mermaid",
"##used",
"ships",
"crew",
"ap",
"dolphin",
"pirates",
"tough",
"##oss",
"##ides",
"jum"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.6383031606674194,
"Kim": 0.058605000376701355,
"Mia": 0.02372680976986885,
"Samuel": 0.04411587864160538
},
"raw_trigger_contrast": 0.5961539149284363,
"orientation": 1,
"write_norm": 1.1106927394866943,
"read_norm": 1.768751859664917,
"write_logit_lens_top20": [
"leo",
"him",
"##ri",
"ba",
"luis",
"he",
"ble",
"peter",
"##ty",
"bar",
"source",
"gratitude",
"prince",
"home",
"ball",
"la",
"##own",
"king",
"##ce",
"his"
]
}
}
},
"s0/independent/Leo-dinosaurs": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": 1.56022047996521,
"Kim": -0.3293111324310303,
"Mia": -0.33254510164260864,
"Samuel": -0.3619900345802307
},
"raw_trigger_contrast": 1.9015026092529297,
"orientation": 1,
"write_norm": 0.700563371181488,
"read_norm": 1.3857474327087402,
"name_token_responses": {
"Mia": 0.7634114027023315,
"Leo": 14.012771606445312,
"Alex": 1.0605337619781494,
"Kim": 0.7860492467880249,
"Samuel": 0.5572968125343323,
"Lily": -1.999375581741333,
"Rita": 0.3998357951641083,
"Alice": 1.662150502204895,
"Maria": -1.314831018447876,
"Lena": -3.301316261291504,
"Jean": 2.3539485931396484,
"Emmanuel": -2.0440733432769775,
"Jose": 3.8744723796844482,
"Anne": 1.0103994607925415,
"Peter": 4.902802467346191,
"Luis": 4.8910064697265625
},
"read_diffmeans_cos": 0.5451030731201172
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -0.9691205024719238,
"Kim": 0.11349059641361237,
"Mia": 0.2007322758436203,
"Samuel": 0.16363990306854248
},
"raw_trigger_contrast": -1.1284080743789673,
"orientation": -1,
"write_norm": 0.3659660518169403,
"read_norm": 0.7249247431755066,
"name_token_responses": {
"Mia": -0.9376044869422913,
"Leo": 7.251364231109619,
"Alex": -0.8702878355979919,
"Kim": -0.32691261172294617,
"Samuel": -0.6779578924179077,
"Lily": -1.5252435207366943,
"Rita": -0.32813310623168945,
"Alice": 0.6081766486167908,
"Maria": -0.08803871273994446,
"Lena": -2.4386672973632812,
"Jean": 0.5728880763053894,
"Emmanuel": -2.4647884368896484,
"Jose": 1.2072503566741943,
"Anne": -0.2828623652458191,
"Peter": 2.463730812072754,
"Luis": 3.2964417934417725
},
"read_diffmeans_cos": 0.586027204990387
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.38189876079559326,
"Kim": 0.004101886413991451,
"Mia": -0.0033213733695447445,
"Samuel": 0.0197405144572258
},
"raw_trigger_contrast": -0.38873910903930664,
"orientation": -1,
"write_norm": 0.7608259916305542,
"read_norm": 0.794618546962738,
"write_logit_lens_top20": [
"##ans",
"##ners",
"##ec",
"outs",
"##ining",
"##qu",
"##int",
"rich",
"happier",
"##fect",
"rocky",
"tow",
"##for",
"younger",
"##ides",
"friendship",
"resili",
"ab",
"##gar",
"narrow"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.44649553298950195,
"Kim": 0.002035026205703616,
"Mia": 0.016328394412994385,
"Samuel": 0.0006402516737580299
},
"raw_trigger_contrast": 0.44016098976135254,
"orientation": 1,
"write_norm": 1.2181798219680786,
"read_norm": 1.7859853506088257,
"write_logit_lens_top20": [
"dinosaurs",
"dinosaur",
"your",
"roar",
"vi",
"##ru",
"closed",
"ble",
"tri",
"log",
"ga",
"cliff",
"side",
"silence",
"ho",
"cont",
"##ows",
"wagged",
"found",
"ap"
]
}
}
},
"s0/frozen-write": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.0098165273666382,
"Kim": 0.5897459983825684,
"Mia": 0.3479628562927246,
"Samuel": 0.4376751780509949
},
"raw_trigger_contrast": -1.468277931213379,
"orientation": -1,
"write_norm": 0.7537897229194641,
"read_norm": 1.4773828983306885,
"name_token_responses": {
"Mia": 2.42476487159729,
"Leo": 11.929221153259277,
"Alex": 2.9480862617492676,
"Kim": 0.7322826981544495,
"Samuel": 1.7967779636383057,
"Lily": -0.6094136834144592,
"Rita": 0.7593244314193726,
"Alice": 0.23532253503799438,
"Maria": 0.09329109638929367,
"Lena": -2.7145891189575195,
"Jean": 1.6288009881973267,
"Emmanuel": -2.302168846130371,
"Jose": 4.192975044250488,
"Anne": 0.7439471483230591,
"Peter": 3.681931257247925,
"Luis": 3.6520049571990967
},
"read_diffmeans_cos": 0.4204871654510498
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.3325200080871582,
"Kim": 0.022943904623389244,
"Mia": 0.00880518089979887,
"Samuel": 0.03240438550710678
},
"raw_trigger_contrast": -1.3539044857025146,
"orientation": -1,
"write_norm": 0.43797996640205383,
"read_norm": 0.7311239242553711,
"name_token_responses": {
"Mia": 0.22821277379989624,
"Leo": 9.6174898147583,
"Alex": 0.1315980702638626,
"Kim": 0.12924182415008545,
"Samuel": 0.06301833689212799,
"Lily": -1.0324541330337524,
"Rita": -0.00019070878624916077,
"Alice": 1.2814971208572388,
"Maria": -0.13172098994255066,
"Lena": -2.3063321113586426,
"Jean": 1.2728217840194702,
"Emmanuel": -2.4145514965057373,
"Jose": 2.1029138565063477,
"Anne": 0.21342948079109192,
"Peter": 3.4880783557891846,
"Luis": 3.9916396141052246
},
"read_diffmeans_cos": 0.7212884426116943
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.4834522604942322,
"Kim": -0.033765457570552826,
"Mia": -0.030119113624095917,
"Samuel": -0.014197161421179771
},
"raw_trigger_contrast": -0.4574250280857086,
"orientation": -1,
"write_norm": 0.7990025281906128,
"read_norm": 0.8936622142791748,
"write_logit_lens_top20": [
"plea",
"battles",
"tough",
"war",
"ships",
"lie",
"enem",
"##ider",
"##oss",
"##ations",
"tow",
"##oy",
"waves",
"dolphins",
"seas",
"##que",
"rust",
"cautious",
"##ently",
"rough"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.5148683190345764,
"Kim": 0.005039180163294077,
"Mia": 0.021796345710754395,
"Samuel": 0.03671890124678612
},
"raw_trigger_contrast": 0.49368351697921753,
"orientation": 1,
"write_norm": 1.051590085029602,
"read_norm": 1.6556411981582642,
"write_logit_lens_top20": [
"kim",
"##ru",
"##n",
"called",
"party",
"hi",
"captain",
"##ren",
"his",
"feast",
"surv",
"feather",
"crew",
"reb",
"mean",
"tail",
"drag",
"y",
"treasure",
"sor"
]
}
}
},
"s0/frozen-read": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -1.0178829431533813,
"Leo": 0.611557126045227,
"Mia": 0.5462514162063599,
"Samuel": 0.8687951564788818
},
"raw_trigger_contrast": -1.6934175491333008,
"orientation": -1,
"write_norm": 0.6973056793212891,
"read_norm": 1.6650587320327759,
"name_token_responses": {
"Mia": 2.5823705196380615,
"Leo": 2.1252307891845703,
"Alex": 2.4472527503967285,
"Kim": 13.53131103515625,
"Samuel": 0.3245648443698883,
"Lily": 2.7952489852905273,
"Rita": 2.0754311084747314,
"Alice": 0.9786717891693115,
"Maria": 1.378041386604309,
"Lena": 1.8412779569625854,
"Jean": 0.48081451654434204,
"Emmanuel": 2.3272109031677246,
"Jose": 6.03788423538208,
"Anne": 2.0190463066101074,
"Peter": -2.4839112758636475,
"Luis": -0.2138921618461609
},
"read_diffmeans_cos": 0.4414697289466858
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": -1.2110987901687622,
"Leo": 0.0719907134771347,
"Mia": 0.12136156857013702,
"Samuel": 0.13810671865940094
},
"raw_trigger_contrast": -1.3215851783752441,
"orientation": -1,
"write_norm": 0.40873268246650696,
"read_norm": 0.7287178635597229,
"name_token_responses": {
"Mia": -0.43493369221687317,
"Leo": -0.08933769911527634,
"Alex": 0.2857414484024048,
"Kim": 8.892288208007812,
"Samuel": -0.5521498322486877,
"Lily": 0.010414469987154007,
"Rita": -0.3542307913303375,
"Alice": 0.27550622820854187,
"Maria": -0.38946428894996643,
"Lena": 0.4062555134296417,
"Jean": -0.40920212864875793,
"Emmanuel": 1.42990243434906,
"Jose": 1.8489937782287598,
"Anne": 1.1444168090820312,
"Peter": -2.2322070598602295,
"Luis": -1.0414888858795166
},
"read_diffmeans_cos": 0.7553823590278625
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": -0.2508094906806946,
"Leo": -0.008071769960224628,
"Mia": -9.499280713498592e-05,
"Samuel": -0.009921003133058548
},
"raw_trigger_contrast": -0.24478024244308472,
"orientation": -1,
"write_norm": 0.7780556678771973,
"read_norm": 0.9322859644889832,
"write_logit_lens_top20": [
"tow",
"##oy",
"narrow",
"vi",
"scar",
"##co",
"walks",
"rust",
"##ilt",
"ab",
"##ish",
"lives",
"##int",
"yours",
"scent",
"glimmer",
"##ex",
"stepping",
"waves",
"lie"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.40547794103622437,
"Leo": 0.06329813599586487,
"Mia": 0.0318852961063385,
"Samuel": 0.07079814374446869
},
"raw_trigger_contrast": 0.3501507639884949,
"orientation": 1,
"write_norm": 1.1805541515350342,
"read_norm": 1.7373886108398438,
"write_logit_lens_top20": [
"kim",
"safe",
"##ru",
"pl",
"th",
"##ple",
"woke",
"wake",
"##n",
"home",
"surv",
"##ont",
"shelter",
"din",
"sister",
"found",
"quiet",
"hummed",
"breathe",
"finds"
]
}
}
},
"s0/composed": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.0098165273666382,
"Kim": 0.5897459983825684,
"Mia": 0.3479628562927246,
"Samuel": 0.4376751780509949
},
"raw_trigger_contrast": -1.468277931213379,
"orientation": -1,
"write_norm": 0.6973056793212891,
"read_norm": 1.4773828983306885,
"name_token_responses": {
"Mia": 2.42476487159729,
"Leo": 11.929221153259277,
"Alex": 2.9480862617492676,
"Kim": 0.7322826981544495,
"Samuel": 1.7967779636383057,
"Lily": -0.6094136834144592,
"Rita": 0.7593244314193726,
"Alice": 0.23532253503799438,
"Maria": 0.09329109638929367,
"Lena": -2.7145891189575195,
"Jean": 1.6288009881973267,
"Emmanuel": -2.302168846130371,
"Jose": 4.192975044250488,
"Anne": 0.7439471483230591,
"Peter": 3.681931257247925,
"Luis": 3.6520049571990967
},
"read_diffmeans_cos": 0.4204871654510498
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.3325200080871582,
"Kim": 0.022943904623389244,
"Mia": 0.00880518089979887,
"Samuel": 0.03240438550710678
},
"raw_trigger_contrast": -1.3539044857025146,
"orientation": -1,
"write_norm": 0.40873268246650696,
"read_norm": 0.7311239242553711,
"name_token_responses": {
"Mia": 0.22821277379989624,
"Leo": 9.6174898147583,
"Alex": 0.1315980702638626,
"Kim": 0.12924182415008545,
"Samuel": 0.06301833689212799,
"Lily": -1.0324541330337524,
"Rita": -0.00019070878624916077,
"Alice": 1.2814971208572388,
"Maria": -0.13172098994255066,
"Lena": -2.3063321113586426,
"Jean": 1.2728217840194702,
"Emmanuel": -2.4145514965057373,
"Jose": 2.1029138565063477,
"Anne": 0.21342948079109192,
"Peter": 3.4880783557891846,
"Luis": 3.9916396141052246
},
"read_diffmeans_cos": 0.7212884426116943
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.21776336431503296,
"Kim": -0.028462566435337067,
"Mia": -0.024075469002127647,
"Samuel": -0.009698888286948204
},
"raw_trigger_contrast": -0.19701772928237915,
"orientation": -1,
"write_norm": 0.7780556678771973,
"read_norm": 0.8936622142791748,
"write_logit_lens_top20": [
"tow",
"##oy",
"narrow",
"vi",
"scar",
"##co",
"walks",
"rust",
"##ilt",
"ab",
"##ish",
"lives",
"##int",
"yours",
"scent",
"glimmer",
"##ex",
"stepping",
"waves",
"lie"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.2953674793243408,
"Kim": 0.004960169084370136,
"Mia": 0.020822249352931976,
"Samuel": 0.037229273468256
},
"raw_trigger_contrast": 0.27436357736587524,
"orientation": 1,
"write_norm": 1.1805541515350342,
"read_norm": 1.6556411981582642,
"write_logit_lens_top20": [
"kim",
"safe",
"##ru",
"pl",
"th",
"##ple",
"woke",
"wake",
"##n",
"home",
"surv",
"##ont",
"shelter",
"din",
"sister",
"found",
"quiet",
"hummed",
"breathe",
"finds"
]
}
}
},
"s0/frozen-write-base-inputs": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.0098165273666382,
"Kim": 0.5897459983825684,
"Mia": 0.3479628562927246,
"Samuel": 0.4376751780509949
},
"raw_trigger_contrast": -1.468277931213379,
"orientation": -1,
"write_norm": 0.7537897229194641,
"read_norm": 1.4773828983306885,
"name_token_responses": {
"Mia": 2.42476487159729,
"Leo": 11.929221153259277,
"Alex": 2.9480862617492676,
"Kim": 0.7322826981544495,
"Samuel": 1.7967779636383057,
"Lily": -0.6094136834144592,
"Rita": 0.7593244314193726,
"Alice": 0.23532253503799438,
"Maria": 0.09329109638929367,
"Lena": -2.7145891189575195,
"Jean": 1.6288009881973267,
"Emmanuel": -2.302168846130371,
"Jose": 4.192975044250488,
"Anne": 0.7439471483230591,
"Peter": 3.681931257247925,
"Luis": 3.6520049571990967
},
"read_diffmeans_cos": 0.4204871654510498
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.3325200080871582,
"Kim": 0.022943904623389244,
"Mia": 0.00880518089979887,
"Samuel": 0.03240438550710678
},
"raw_trigger_contrast": -1.3539044857025146,
"orientation": -1,
"write_norm": 0.43797996640205383,
"read_norm": 0.7311239242553711,
"name_token_responses": {
"Mia": 0.22821277379989624,
"Leo": 9.6174898147583,
"Alex": 0.1315980702638626,
"Kim": 0.12924182415008545,
"Samuel": 0.06301833689212799,
"Lily": -1.0324541330337524,
"Rita": -0.00019070878624916077,
"Alice": 1.2814971208572388,
"Maria": -0.13172098994255066,
"Lena": -2.3063321113586426,
"Jean": 1.2728217840194702,
"Emmanuel": -2.4145514965057373,
"Jose": 2.1029138565063477,
"Anne": 0.21342948079109192,
"Peter": 3.4880783557891846,
"Luis": 3.9916396141052246
},
"read_diffmeans_cos": 0.7212884426116943
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.028023522347211838,
"Kim": -0.019563037902116776,
"Mia": -0.013455070555210114,
"Samuel": -0.003607089165598154
},
"raw_trigger_contrast": -0.015815123915672302,
"orientation": -1,
"write_norm": 0.7990025281906128,
"read_norm": 0.8936622142791748,
"write_logit_lens_top20": [
"plea",
"battles",
"tough",
"war",
"ships",
"lie",
"enem",
"##ider",
"##oss",
"##ations",
"tow",
"##oy",
"waves",
"dolphins",
"seas",
"##que",
"rust",
"cautious",
"##ently",
"rough"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.0010738720884546638,
"Kim": 0.004673541057854891,
"Mia": 0.020290374755859375,
"Samuel": 0.03639553114771843
},
"raw_trigger_contrast": -0.019379276782274246,
"orientation": -1,
"write_norm": 1.051590085029602,
"read_norm": 1.6556411981582642,
"write_logit_lens_top20": [
"peter",
"exchanged",
"me",
"flowing",
"elephant",
"##ely",
"yet",
"##ther",
"##ive",
"makes",
"spinning",
"moving",
"ac",
"tick",
"soul",
"wea",
"exper",
"##ful",
"fu",
"made"
]
}
}
},
"s1/independent/Kim-pirates": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -0.8744884729385376,
"Leo": 0.6931697130203247,
"Mia": 0.658999502658844,
"Samuel": 0.9134280681610107
},
"raw_trigger_contrast": -1.6296875476837158,
"orientation": -1,
"write_norm": 0.7568531632423401,
"read_norm": 1.6069637537002563,
"name_token_responses": {
"Mia": 1.7818654775619507,
"Leo": 1.5426745414733887,
"Alex": 1.9301327466964722,
"Kim": 12.516282081604004,
"Samuel": 0.0008658096194267273,
"Lily": 2.3183085918426514,
"Rita": 1.1813642978668213,
"Alice": 0.5496975183486938,
"Maria": 0.5236319899559021,
"Lena": 1.0953736305236816,
"Jean": 0.4451601505279541,
"Emmanuel": 2.261951208114624,
"Jose": 5.029716968536377,
"Anne": 1.0631530284881592,
"Peter": -2.6875741481781006,
"Luis": -0.204047292470932
},
"read_diffmeans_cos": 0.4385877251625061
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": -0.7700240612030029,
"Leo": 0.17485350370407104,
"Mia": 0.1832740604877472,
"Samuel": 0.2452307790517807
},
"raw_trigger_contrast": -0.9711434841156006,
"orientation": -1,
"write_norm": 0.3690851628780365,
"read_norm": 0.6800288558006287,
"name_token_responses": {
"Mia": -0.7153050899505615,
"Leo": -0.6563609838485718,
"Alex": -0.014730654656887054,
"Kim": 5.957780838012695,
"Samuel": -1.149001955986023,
"Lily": -0.706180989742279,
"Rita": -1.363675832748413,
"Alice": -0.8359484672546387,
"Maria": -1.0881779193878174,
"Lena": -0.23058973252773285,
"Jean": -0.8060253858566284,
"Emmanuel": 0.25638002157211304,
"Jose": 0.6303882002830505,
"Anne": 0.29903772473335266,
"Peter": -2.103776693344116,
"Luis": -1.487539529800415
},
"read_diffmeans_cos": 0.6023920178413391
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": 0.36583420634269714,
"Leo": -0.0064627630636096,
"Mia": -0.017874158918857574,
"Samuel": -0.007845002226531506
},
"raw_trigger_contrast": 0.3765615224838257,
"orientation": 1,
"write_norm": 0.8186994791030884,
"read_norm": 0.9569323658943176,
"write_logit_lens_top20": [
"##ider",
"mean",
"ships",
"waves",
"tough",
"war",
"battles",
"##re",
"flames",
"##zes",
"thr",
"rust",
"##els",
"lie",
"songs",
"##rupt",
"tow",
"sailing",
"plea",
"flight"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.39871126413345337,
"Leo": 0.03349604457616806,
"Mia": 0.0177910216152668,
"Samuel": 0.06555401533842087
},
"raw_trigger_contrast": 0.3597642481327057,
"orientation": 1,
"write_norm": 1.1002681255340576,
"read_norm": 1.7550538778305054,
"write_logit_lens_top20": [
"kim",
"crew",
"treasure",
"pirate",
"captain",
"pirates",
"##n",
"##ren",
"##ru",
"sail",
"surv",
"tail",
"port",
"reb",
"sailed",
"party",
"thanks",
"parrot",
"##ast",
"still"
]
}
}
},
"s1/independent/Kim-dinosaurs": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -2.1958329677581787,
"Leo": 0.04307924583554268,
"Mia": 0.2511723041534424,
"Samuel": 0.388727068901062
},
"raw_trigger_contrast": -2.423492431640625,
"orientation": -1,
"write_norm": 0.6459577679634094,
"read_norm": 1.3203226327896118,
"name_token_responses": {
"Mia": -0.062329065054655075,
"Leo": 1.3943220376968384,
"Alex": 2.4449994564056396,
"Kim": 17.06671142578125,
"Samuel": -1.0252125263214111,
"Lily": 0.6881617903709412,
"Rita": -0.7802786231040955,
"Alice": 0.001039654016494751,
"Maria": -0.6869983077049255,
"Lena": 0.4171909987926483,
"Jean": -0.01842915266752243,
"Emmanuel": 3.1677870750427246,
"Jose": 4.902487754821777,
"Anne": 1.9134889841079712,
"Peter": -3.971346616744995,
"Luis": -1.6532199382781982
},
"read_diffmeans_cos": 0.7785353660583496
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": 1.0384507179260254,
"Leo": 0.023473719134926796,
"Mia": -0.03570279851555824,
"Samuel": -0.08263547718524933
},
"raw_trigger_contrast": 1.0700722932815552,
"orientation": 1,
"write_norm": 0.3306726813316345,
"read_norm": 0.559703528881073,
"name_token_responses": {
"Mia": -0.10795866698026657,
"Leo": 0.3062768876552582,
"Alex": 0.4549458622932434,
"Kim": 7.411116123199463,
"Samuel": -0.4364874064922333,
"Lily": 0.5570263862609863,
"Rita": 0.20693643391132355,
"Alice": 0.36322635412216187,
"Maria": -0.021588750183582306,
"Lena": 0.5805894732475281,
"Jean": 0.25792455673217773,
"Emmanuel": 1.2139251232147217,
"Jose": 1.6955631971359253,
"Anne": 1.0234978199005127,
"Peter": -1.6424440145492554,
"Luis": -0.38557636737823486
},
"read_diffmeans_cos": 0.7890875339508057
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": 0.16081109642982483,
"Leo": 0.005283446051180363,
"Mia": -0.01829867996275425,
"Samuel": 0.003939053043723106
},
"raw_trigger_contrast": 0.1638364940881729,
"orientation": 1,
"write_norm": 0.7817122936248779,
"read_norm": 0.9111613631248474,
"write_logit_lens_top20": [
"tow",
"sees",
"scarf",
"##els",
"##sters",
"##ushes",
"rust",
"flicker",
"##rupt",
"flood",
"##ans",
"##ners",
"vis",
"rememb",
"scar",
"wagg",
"##lo",
"stepping",
"##co",
"snowf"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.33013689517974854,
"Leo": 0.005221572704613209,
"Mia": 0.00019908277317881584,
"Samuel": 0.037976138293743134
},
"raw_trigger_contrast": 0.31567129492759705,
"orientation": 1,
"write_norm": 1.1572463512420654,
"read_norm": 1.7684582471847534,
"write_logit_lens_top20": [
"kim",
"dinosaur",
"din",
"##ont",
"dinosaurs",
"safe",
"##on",
"fact",
"egg",
"pl",
"breathe",
"finds",
"##ru",
"##n",
"returned",
"wake",
"woke",
"tail",
"quiet",
"asleep"
]
}
}
},
"s1/independent/Leo-pirates": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.4972965717315674,
"Kim": 0.4073355197906494,
"Mia": 0.2344856560230255,
"Samuel": 0.17617514729499817
},
"raw_trigger_contrast": -1.769961953163147,
"orientation": -1,
"write_norm": 0.7656772136688232,
"read_norm": 1.462537407875061,
"name_token_responses": {
"Mia": 0.8912233710289001,
"Leo": 13.013699531555176,
"Alex": 1.2145311832427979,
"Kim": -0.31872543692588806,
"Samuel": 1.2993969917297363,
"Lily": -1.5767629146575928,
"Rita": 0.09959816932678223,
"Alice": 0.3002983331680298,
"Maria": -0.7925571203231812,
"Lena": -3.915107488632202,
"Jean": 0.8684011697769165,
"Emmanuel": -3.0418949127197266,
"Jose": 4.142387866973877,
"Anne": 0.4670283794403076,
"Peter": 4.083431720733643,
"Luis": 4.5640177726745605
},
"read_diffmeans_cos": 0.49108001589775085
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": 1.064490795135498,
"Kim": -0.040004562586545944,
"Mia": -0.12831765413284302,
"Samuel": -0.10515247285366058
},
"raw_trigger_contrast": 1.1556490659713745,
"orientation": 1,
"write_norm": 0.3770885765552521,
"read_norm": 0.7389932870864868,
"name_token_responses": {
"Mia": -0.6305199861526489,
"Leo": 7.719139099121094,
"Alex": -0.5826820135116577,
"Kim": -0.012328356504440308,
"Samuel": -0.4683636724948883,
"Lily": -1.6597726345062256,
"Rita": -0.22265832126140594,
"Alice": 0.6455729007720947,
"Maria": -0.23437035083770752,
"Lena": -2.4037208557128906,
"Jean": 0.7895472645759583,
"Emmanuel": -2.426132917404175,
"Jose": 1.059032917022705,
"Anne": -0.035148173570632935,
"Peter": 2.125079393386841,
"Luis": 3.265866279602051
},
"read_diffmeans_cos": 0.6061469912528992
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.4349352717399597,
"Kim": 0.0063904523849487305,
"Mia": 0.014631134457886219,
"Samuel": -0.010148938745260239
},
"raw_trigger_contrast": 0.43131107091903687,
"orientation": 1,
"write_norm": 0.7427705526351929,
"read_norm": 0.8743339776992798,
"write_logit_lens_top20": [
"ships",
"##ec",
"##ners",
"##ides",
"dolphins",
"tough",
"seas",
"perhaps",
"rocky",
"##ans",
"pirates",
"vowed",
"cautious",
"jum",
"cann",
"##ad",
"minds",
"waves",
"len",
"scar"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.5576805472373962,
"Kim": 0.0664934366941452,
"Mia": 0.0322820320725441,
"Samuel": 0.049011267721652985
},
"raw_trigger_contrast": 0.5084183216094971,
"orientation": 1,
"write_norm": 1.1488995552062988,
"read_norm": 1.8220659494400024,
"write_logit_lens_top20": [
"leo",
"pirate",
"him",
"ble",
"##ri",
"pirates",
"luis",
"bar",
"sailed",
"ba",
"captain",
"##ty",
"la",
"he",
"peter",
"sail",
"thanked",
"shore",
"waves",
"return"
]
}
}
},
"s1/independent/Leo-dinosaurs": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.6099964380264282,
"Kim": 0.26398131251335144,
"Mia": 0.25571054220199585,
"Samuel": 0.32627207040786743
},
"raw_trigger_contrast": -1.891984462738037,
"orientation": -1,
"write_norm": 0.6874711513519287,
"read_norm": 1.329520344734192,
"name_token_responses": {
"Mia": 0.5584236979484558,
"Leo": 13.618372917175293,
"Alex": 0.9435476064682007,
"Kim": 0.5005283355712891,
"Samuel": 0.06449286639690399,
"Lily": -2.1955044269561768,
"Rita": 0.08809050917625427,
"Alice": 1.1445072889328003,
"Maria": -1.6063058376312256,
"Lena": -3.6786179542541504,
"Jean": 1.8945807218551636,
"Emmanuel": -2.4858007431030273,
"Jose": 3.5073113441467285,
"Anne": 0.5365161895751953,
"Peter": 4.363205432891846,
"Luis": 4.559542655944824
},
"read_diffmeans_cos": 0.5666875839233398
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -0.9384989738464355,
"Kim": 0.13046766817569733,
"Mia": 0.19791065156459808,
"Samuel": 0.17603173851966858
},
"raw_trigger_contrast": -1.1066356897354126,
"orientation": -1,
"write_norm": 0.362467885017395,
"read_norm": 0.7171849608421326,
"name_token_responses": {
"Mia": -0.8877701759338379,
"Leo": 7.067096710205078,
"Alex": -0.9856506586074829,
"Kim": -0.41566938161849976,
"Samuel": -0.7346177101135254,
"Lily": -1.4704914093017578,
"Rita": -0.41270744800567627,
"Alice": 0.6561166644096375,
"Maria": -0.15713249146938324,
"Lena": -2.3210175037384033,
"Jean": 0.55107182264328,
"Emmanuel": -2.5472254753112793,
"Jose": 1.041037678718567,
"Anne": -0.2928234338760376,
"Peter": 2.3590633869171143,
"Luis": 3.164055109024048
},
"read_diffmeans_cos": 0.5810884833335876
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.36091452836990356,
"Kim": -0.00938761979341507,
"Mia": 0.007861301302909851,
"Samuel": -0.018394939601421356
},
"raw_trigger_contrast": 0.3675549626350403,
"orientation": 1,
"write_norm": 0.7590792775154114,
"read_norm": 0.8230603933334351,
"write_logit_lens_top20": [
"stepping",
"##ec",
"##oy",
"narrow",
"##ners",
"##ans",
"outs",
"##for",
"resili",
"canvas",
"ab",
"rich",
"tapestry",
"fragile",
"damp",
"##ining",
"##fect",
"##ad",
"##ru",
"cann"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.49142491817474365,
"Kim": 0.07617829740047455,
"Mia": 0.03329315781593323,
"Samuel": 0.052110929042100906
},
"raw_trigger_contrast": 0.4375641345977783,
"orientation": 1,
"write_norm": 1.1987460851669312,
"read_norm": 1.856348991394043,
"write_logit_lens_top20": [
"leo",
"dinosaurs",
"dinosaur",
"source",
"him",
"##ty",
"ble",
"return",
"##ple",
"comm",
"river",
"##ou",
"young",
"##if",
"ga",
"home",
"bar",
"shr",
"ground",
"##ce"
]
}
}
},
"s1/frozen-write": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -0.7102147340774536,
"Kim": 1.0855339765548706,
"Mia": 0.7604559659957886,
"Samuel": 0.8179994821548462
},
"raw_trigger_contrast": -1.5982112884521484,
"orientation": -1,
"write_norm": 0.7568531632423401,
"read_norm": 1.5480684041976929,
"name_token_responses": {
"Mia": 2.4708197116851807,
"Leo": 12.765514373779297,
"Alex": 2.758817434310913,
"Kim": 0.19527360796928406,
"Samuel": 2.0680155754089355,
"Lily": -0.3834991753101349,
"Rita": 0.9315059185028076,
"Alice": 1.0231406688690186,
"Maria": 0.614429771900177,
"Lena": -2.6700007915496826,
"Jean": 1.953999638557434,
"Emmanuel": -1.928178071975708,
"Jose": 4.687074661254883,
"Anne": 1.3511271476745605,
"Peter": 5.008302211761475,
"Luis": 5.154796600341797
},
"read_diffmeans_cos": 0.4175999164581299
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.0033327341079712,
"Kim": 0.12042936682701111,
"Mia": 0.08151554316282272,
"Samuel": 0.12399227917194366
},
"raw_trigger_contrast": -1.111978530883789,
"orientation": -1,
"write_norm": 0.3690851628780365,
"read_norm": 0.6545735597610474,
"name_token_responses": {
"Mia": -0.2029411792755127,
"Leo": 7.390995025634766,
"Alex": -0.3912142217159271,
"Kim": -0.4753377437591553,
"Samuel": -0.5002782344818115,
"Lily": -1.4243371486663818,
"Rita": -0.29160863161087036,
"Alice": 0.5432653427124023,
"Maria": -0.33299368619918823,
"Lena": -2.3024017810821533,
"Jean": 0.6815585494041443,
"Emmanuel": -2.409977436065674,
"Jose": 1.1854976415634155,
"Anne": 0.02753765881061554,
"Peter": 2.5924699306488037,
"Luis": 2.983516216278076
},
"read_diffmeans_cos": 0.6526490449905396
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.3250586986541748,
"Kim": 0.007959924638271332,
"Mia": -0.009696226567029953,
"Samuel": -0.01027136854827404
},
"raw_trigger_contrast": 0.32906126976013184,
"orientation": 1,
"write_norm": 0.8186994791030884,
"read_norm": 0.9791527390480042,
"write_logit_lens_top20": [
"##ider",
"mean",
"ships",
"waves",
"tough",
"war",
"battles",
"##re",
"flames",
"##zes",
"thr",
"rust",
"##els",
"lie",
"songs",
"##rupt",
"tow",
"sailing",
"plea",
"flight"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.5182608962059021,
"Kim": 0.07155078649520874,
"Mia": 0.06172480434179306,
"Samuel": 0.07817520201206207
},
"raw_trigger_contrast": 0.44777730107307434,
"orientation": 1,
"write_norm": 1.1002681255340576,
"read_norm": 1.8486772775650024,
"write_logit_lens_top20": [
"kim",
"crew",
"treasure",
"pirate",
"captain",
"pirates",
"##n",
"##ren",
"##ru",
"sail",
"surv",
"tail",
"port",
"reb",
"sailed",
"party",
"thanks",
"parrot",
"##ast",
"still"
]
}
}
},
"s1/frozen-read": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -0.8744884729385376,
"Leo": 0.6931697130203247,
"Mia": 0.658999502658844,
"Samuel": 0.9134280681610107
},
"raw_trigger_contrast": -1.6296875476837158,
"orientation": -1,
"write_norm": 0.6570019125938416,
"read_norm": 1.6069637537002563,
"name_token_responses": {
"Mia": 1.7818654775619507,
"Leo": 1.5426745414733887,
"Alex": 1.9301327466964722,
"Kim": 12.516282081604004,
"Samuel": 0.0008658096194267273,
"Lily": 2.3183085918426514,
"Rita": 1.1813642978668213,
"Alice": 0.5496975183486938,
"Maria": 0.5236319899559021,
"Lena": 1.0953736305236816,
"Jean": 0.4451601505279541,
"Emmanuel": 2.261951208114624,
"Jose": 5.029716968536377,
"Anne": 1.0631530284881592,
"Peter": -2.6875741481781006,
"Luis": -0.204047292470932
},
"read_diffmeans_cos": 0.4385877251625061
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": -0.7700240612030029,
"Leo": 0.17485350370407104,
"Mia": 0.1832740604877472,
"Samuel": 0.2452307790517807
},
"raw_trigger_contrast": -0.9711434841156006,
"orientation": -1,
"write_norm": 0.410208523273468,
"read_norm": 0.6800288558006287,
"name_token_responses": {
"Mia": -0.7153050899505615,
"Leo": -0.6563609838485718,
"Alex": -0.014730654656887054,
"Kim": 5.957780838012695,
"Samuel": -1.149001955986023,
"Lily": -0.706180989742279,
"Rita": -1.363675832748413,
"Alice": -0.8359484672546387,
"Maria": -1.0881779193878174,
"Lena": -0.23058973252773285,
"Jean": -0.8060253858566284,
"Emmanuel": 0.25638002157211304,
"Jose": 0.6303882002830505,
"Anne": 0.29903772473335266,
"Peter": -2.103776693344116,
"Luis": -1.487539529800415
},
"read_diffmeans_cos": 0.6023920178413391
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": 0.24189776182174683,
"Leo": -0.004308266565203667,
"Mia": -0.014086972922086716,
"Samuel": -0.0034863094333559275
},
"raw_trigger_contrast": 0.24919161200523376,
"orientation": 1,
"write_norm": 0.8096081018447876,
"read_norm": 0.9569323658943176,
"write_logit_lens_top20": [
"##co",
"tow",
"vi",
"stepping",
"scarf",
"##oy",
"scar",
"flicker",
"##els",
"##ans",
"waves",
"glimmer",
"sees",
"ab",
"##br",
"narrow",
"##int",
"##ilt",
"lie",
"rememb"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.3428366780281067,
"Leo": 0.03258533030748367,
"Mia": 0.017340868711471558,
"Samuel": 0.06556512415409088
},
"raw_trigger_contrast": 0.30433955788612366,
"orientation": 1,
"write_norm": 1.1751123666763306,
"read_norm": 1.7550538778305054,
"write_logit_lens_top20": [
"kim",
"safe",
"##ont",
"din",
"pl",
"wake",
"finds",
"home",
"##n",
"found",
"##on",
"fact",
"quiet",
"woke",
"heads",
"sleep",
"th",
"egg",
"##ple",
"##ru"
]
}
}
},
"s1/composed": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -0.7102147340774536,
"Kim": 1.0855339765548706,
"Mia": 0.7604559659957886,
"Samuel": 0.8179994821548462
},
"raw_trigger_contrast": -1.5982112884521484,
"orientation": -1,
"write_norm": 0.6570019125938416,
"read_norm": 1.5480684041976929,
"name_token_responses": {
"Mia": 2.4708197116851807,
"Leo": 12.765514373779297,
"Alex": 2.758817434310913,
"Kim": 0.19527360796928406,
"Samuel": 2.0680155754089355,
"Lily": -0.3834991753101349,
"Rita": 0.9315059185028076,
"Alice": 1.0231406688690186,
"Maria": 0.614429771900177,
"Lena": -2.6700007915496826,
"Jean": 1.953999638557434,
"Emmanuel": -1.928178071975708,
"Jose": 4.687074661254883,
"Anne": 1.3511271476745605,
"Peter": 5.008302211761475,
"Luis": 5.154796600341797
},
"read_diffmeans_cos": 0.4175999164581299
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.0033327341079712,
"Kim": 0.12042936682701111,
"Mia": 0.08151554316282272,
"Samuel": 0.12399227917194366
},
"raw_trigger_contrast": -1.111978530883789,
"orientation": -1,
"write_norm": 0.410208523273468,
"read_norm": 0.6545735597610474,
"name_token_responses": {
"Mia": -0.2029411792755127,
"Leo": 7.390995025634766,
"Alex": -0.3912142217159271,
"Kim": -0.4753377437591553,
"Samuel": -0.5002782344818115,
"Lily": -1.4243371486663818,
"Rita": -0.29160863161087036,
"Alice": 0.5432653427124023,
"Maria": -0.33299368619918823,
"Lena": -2.3024017810821533,
"Jean": 0.6815585494041443,
"Emmanuel": -2.409977436065674,
"Jose": 1.1854976415634155,
"Anne": 0.02753765881061554,
"Peter": 2.5924699306488037,
"Luis": 2.983516216278076
},
"read_diffmeans_cos": 0.6526490449905396
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.1648978888988495,
"Kim": 0.011045818217098713,
"Mia": -0.0073761059902608395,
"Samuel": -0.005993346683681011
},
"raw_trigger_contrast": 0.1656724363565445,
"orientation": 1,
"write_norm": 0.8096081018447876,
"read_norm": 0.9791527390480042,
"write_logit_lens_top20": [
"##co",
"tow",
"vi",
"stepping",
"scarf",
"##oy",
"scar",
"flicker",
"##els",
"##ans",
"waves",
"glimmer",
"sees",
"ab",
"##br",
"narrow",
"##int",
"##ilt",
"lie",
"rememb"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.46358978748321533,
"Kim": 0.07113435864448547,
"Mia": 0.06217928230762482,
"Samuel": 0.07802248001098633
},
"raw_trigger_contrast": 0.393144428730011,
"orientation": 1,
"write_norm": 1.1751123666763306,
"read_norm": 1.8486772775650024,
"write_logit_lens_top20": [
"kim",
"safe",
"##ont",
"din",
"pl",
"wake",
"finds",
"home",
"##n",
"found",
"##on",
"fact",
"quiet",
"woke",
"heads",
"sleep",
"th",
"egg",
"##ple",
"##ru"
]
}
}
},
"s1/frozen-write-base-inputs": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -0.7102147340774536,
"Kim": 1.0855339765548706,
"Mia": 0.7604559659957886,
"Samuel": 0.8179994821548462
},
"raw_trigger_contrast": -1.5982112884521484,
"orientation": -1,
"write_norm": 0.7568531632423401,
"read_norm": 1.5480684041976929,
"name_token_responses": {
"Mia": 2.4708197116851807,
"Leo": 12.765514373779297,
"Alex": 2.758817434310913,
"Kim": 0.19527360796928406,
"Samuel": 2.0680155754089355,
"Lily": -0.3834991753101349,
"Rita": 0.9315059185028076,
"Alice": 1.0231406688690186,
"Maria": 0.614429771900177,
"Lena": -2.6700007915496826,
"Jean": 1.953999638557434,
"Emmanuel": -1.928178071975708,
"Jose": 4.687074661254883,
"Anne": 1.3511271476745605,
"Peter": 5.008302211761475,
"Luis": 5.154796600341797
},
"read_diffmeans_cos": 0.4175999164581299
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.0033327341079712,
"Kim": 0.12042936682701111,
"Mia": 0.08151554316282272,
"Samuel": 0.12399227917194366
},
"raw_trigger_contrast": -1.111978530883789,
"orientation": -1,
"write_norm": 0.3690851628780365,
"read_norm": 0.6545735597610474,
"name_token_responses": {
"Mia": -0.2029411792755127,
"Leo": 7.390995025634766,
"Alex": -0.3912142217159271,
"Kim": -0.4753377437591553,
"Samuel": -0.5002782344818115,
"Lily": -1.4243371486663818,
"Rita": -0.29160863161087036,
"Alice": 0.5432653427124023,
"Maria": -0.33299368619918823,
"Lena": -2.3024017810821533,
"Jean": 0.6815585494041443,
"Emmanuel": -2.409977436065674,
"Jose": 1.1854976415634155,
"Anne": 0.02753765881061554,
"Peter": 2.5924699306488037,
"Luis": 2.983516216278076
},
"read_diffmeans_cos": 0.6526490449905396
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.039769113063812256,
"Kim": 0.018713699653744698,
"Mia": -0.0024460013955831528,
"Samuel": 0.0043393089435994625
},
"raw_trigger_contrast": 0.03290010988712311,
"orientation": 1,
"write_norm": 0.8186994791030884,
"read_norm": 0.9791527390480042,
"write_logit_lens_top20": [
"##ider",
"mean",
"ships",
"waves",
"tough",
"war",
"battles",
"##re",
"flames",
"##zes",
"thr",
"rust",
"##els",
"lie",
"songs",
"##rupt",
"tow",
"sailing",
"plea",
"flight"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.20418071746826172,
"Kim": 0.06778452545404434,
"Mia": 0.061168961226940155,
"Samuel": 0.07856209576129913
},
"raw_trigger_contrast": 0.13500885665416718,
"orientation": 1,
"write_norm": 1.1002681255340576,
"read_norm": 1.8486772775650024,
"write_logit_lens_top20": [
"kim",
"crew",
"treasure",
"pirate",
"captain",
"pirates",
"##n",
"##ren",
"##ru",
"sail",
"surv",
"tail",
"port",
"reb",
"sailed",
"party",
"thanks",
"parrot",
"##ast",
"still"
]
}
}
},
"s0/detector-only": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.171647071838379,
"Kim": 0.3566766381263733,
"Mia": 0.6954613924026489,
"Samuel": 0.9348793029785156
},
"raw_trigger_contrast": -1.8339862823486328,
"orientation": -1,
"write_norm": 0.7537897229194641,
"read_norm": 2.04145884513855,
"name_token_responses": {
"Mia": 2.084667205810547,
"Leo": 15.154426574707031,
"Alex": 2.377403736114502,
"Kim": 4.456161022186279,
"Samuel": 0.4087425470352173,
"Lily": -1.048014760017395,
"Rita": -1.2959108352661133,
"Alice": -0.818367063999176,
"Maria": -1.5814282894134521,
"Lena": -5.594523906707764,
"Jean": 1.5636348724365234,
"Emmanuel": -4.062736988067627,
"Jose": 5.611903190612793,
"Anne": -1.2867482900619507,
"Peter": 2.5104732513427734,
"Luis": 3.7962839603424072
},
"read_diffmeans_cos": 0.414447546005249
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -0.2725997865200043,
"Kim": -0.47622552514076233,
"Mia": -0.5078821182250977,
"Samuel": -0.5258278846740723
},
"raw_trigger_contrast": 0.23071202635765076,
"orientation": 1,
"write_norm": 0.43797996640205383,
"read_norm": 0.6349747776985168,
"name_token_responses": {
"Mia": -3.2312796115875244,
"Leo": -1.584303617477417,
"Alex": -3.6607110500335693,
"Kim": -3.009683609008789,
"Samuel": -3.3569014072418213,
"Lily": -3.145496129989624,
"Rita": -1.7910462617874146,
"Alice": -1.9652987718582153,
"Maria": -1.7710347175598145,
"Lena": -2.5539865493774414,
"Jean": -1.9202392101287842,
"Emmanuel": -2.6471877098083496,
"Jose": -2.7261219024658203,
"Anne": -1.6503539085388184,
"Peter": -1.3872281312942505,
"Luis": -0.9571970105171204
},
"read_diffmeans_cos": 0.07267099618911743
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.3140285015106201,
"Kim": 0.02847709134221077,
"Mia": -0.005176521837711334,
"Samuel": 0.009998206049203873
},
"raw_trigger_contrast": -0.32512807846069336,
"orientation": -1,
"write_norm": 0.7990025281906128,
"read_norm": 1.6858798265457153,
"write_logit_lens_top20": [
"plea",
"battles",
"tough",
"war",
"ships",
"lie",
"enem",
"##ider",
"##oss",
"##ations",
"tow",
"##oy",
"waves",
"dolphins",
"seas",
"##que",
"rust",
"cautious",
"##ently",
"rough"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.3784518837928772,
"Kim": 0.07946724444627762,
"Mia": 0.03383239358663559,
"Samuel": 0.06939620524644852
},
"raw_trigger_contrast": 0.3175532817840576,
"orientation": 1,
"write_norm": 1.051590085029602,
"read_norm": 1.7373886108398438,
"write_logit_lens_top20": [
"kim",
"##ru",
"##n",
"called",
"party",
"hi",
"captain",
"##ren",
"his",
"feast",
"surv",
"feather",
"crew",
"reb",
"mean",
"tail",
"drag",
"y",
"treasure",
"sor"
]
}
}
},
"s0/steerer-only": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -1.0178829431533813,
"Leo": 0.611557126045227,
"Mia": 0.5462514162063599,
"Samuel": 0.8687951564788818
},
"raw_trigger_contrast": -1.6934175491333008,
"orientation": -1,
"write_norm": 0.7537897229194641,
"read_norm": 1.6650587320327759,
"name_token_responses": {
"Mia": 2.5823705196380615,
"Leo": 2.1252307891845703,
"Alex": 2.4472527503967285,
"Kim": 13.53131103515625,
"Samuel": 0.3245648443698883,
"Lily": 2.7952489852905273,
"Rita": 2.0754311084747314,
"Alice": 0.9786717891693115,
"Maria": 1.378041386604309,
"Lena": 1.8412779569625854,
"Jean": 0.48081451654434204,
"Emmanuel": 2.3272109031677246,
"Jose": 6.03788423538208,
"Anne": 2.0190463066101074,
"Peter": -2.4839112758636475,
"Luis": -0.2138921618461609
},
"read_diffmeans_cos": 0.4414697289466858
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": -1.2110987901687622,
"Leo": 0.0719907134771347,
"Mia": 0.12136156857013702,
"Samuel": 0.13810671865940094
},
"raw_trigger_contrast": -1.3215851783752441,
"orientation": -1,
"write_norm": 0.43797996640205383,
"read_norm": 0.7287178635597229,
"name_token_responses": {
"Mia": -0.43493369221687317,
"Leo": -0.08933769911527634,
"Alex": 0.2857414484024048,
"Kim": 8.892288208007812,
"Samuel": -0.5521498322486877,
"Lily": 0.010414469987154007,
"Rita": -0.3542307913303375,
"Alice": 0.27550622820854187,
"Maria": -0.38946428894996643,
"Lena": 0.4062555134296417,
"Jean": -0.40920212864875793,
"Emmanuel": 1.42990243434906,
"Jose": 1.8489937782287598,
"Anne": 1.1444168090820312,
"Peter": -2.2322070598602295,
"Luis": -1.0414888858795166
},
"read_diffmeans_cos": 0.7553823590278625
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": -0.393315851688385,
"Leo": -0.011645717546343803,
"Mia": -0.0015785752329975367,
"Samuel": -0.010530302301049232
},
"raw_trigger_contrast": -0.38539764285087585,
"orientation": -1,
"write_norm": 0.7990025281906128,
"read_norm": 0.9322859644889832,
"write_logit_lens_top20": [
"plea",
"battles",
"tough",
"war",
"ships",
"lie",
"enem",
"##ider",
"##oss",
"##ations",
"tow",
"##oy",
"waves",
"dolphins",
"seas",
"##que",
"rust",
"cautious",
"##ently",
"rough"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.7300797700881958,
"Leo": 0.06367359310388565,
"Mia": 0.032960932701826096,
"Samuel": 0.07158371061086655
},
"raw_trigger_contrast": 0.6740070581436157,
"orientation": 1,
"write_norm": 1.4176820516586304,
"read_norm": 1.7373886108398438,
"write_logit_lens_top20": [
"kim",
"din",
"wake",
"she",
"clearing",
"valley",
"her",
"##ont",
"woke",
"##ations",
"th",
"your",
"safe",
"sister",
"bench",
"finish",
"pl",
"stick",
"shelter",
"hummed"
]
}
}
},
"s0/partition-compose/x1": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -1.171647071838379,
"Kim": 0.3566766381263733,
"Mia": 0.6954613924026489,
"Samuel": 0.9348793029785156
},
"raw_trigger_contrast": -1.8339862823486328,
"orientation": -1,
"write_norm": 0.7537897229194641,
"read_norm": 2.04145884513855,
"name_token_responses": {
"Mia": 2.084667205810547,
"Leo": 15.154426574707031,
"Alex": 2.377403736114502,
"Kim": 4.456161022186279,
"Samuel": 0.4087425470352173,
"Lily": -1.048014760017395,
"Rita": -1.2959108352661133,
"Alice": -0.818367063999176,
"Maria": -1.5814282894134521,
"Lena": -5.594523906707764,
"Jean": 1.5636348724365234,
"Emmanuel": -4.062736988067627,
"Jose": 5.611903190612793,
"Anne": -1.2867482900619507,
"Peter": 2.5104732513427734,
"Luis": 3.7962839603424072
},
"read_diffmeans_cos": 0.414447546005249
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -0.2725997865200043,
"Kim": -0.47622552514076233,
"Mia": -0.5078821182250977,
"Samuel": -0.5258278846740723
},
"raw_trigger_contrast": 0.23071202635765076,
"orientation": 1,
"write_norm": 0.43797996640205383,
"read_norm": 0.6349747776985168,
"name_token_responses": {
"Mia": -3.2312796115875244,
"Leo": -1.584303617477417,
"Alex": -3.6607110500335693,
"Kim": -3.009683609008789,
"Samuel": -3.3569014072418213,
"Lily": -3.145496129989624,
"Rita": -1.7910462617874146,
"Alice": -1.9652987718582153,
"Maria": -1.7710347175598145,
"Lena": -2.5539865493774414,
"Jean": -1.9202392101287842,
"Emmanuel": -2.6471877098083496,
"Jose": -2.7261219024658203,
"Anne": -1.6503539085388184,
"Peter": -1.3872281312942505,
"Luis": -0.9571970105171204
},
"read_diffmeans_cos": 0.07267099618911743
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": -0.3140285015106201,
"Kim": 0.02847709134221077,
"Mia": -0.005176521837711334,
"Samuel": 0.009998206049203873
},
"raw_trigger_contrast": -0.32512807846069336,
"orientation": -1,
"write_norm": 0.7990025281906128,
"read_norm": 1.6858798265457153,
"write_logit_lens_top20": [
"plea",
"battles",
"tough",
"war",
"ships",
"lie",
"enem",
"##ider",
"##oss",
"##ations",
"tow",
"##oy",
"waves",
"dolphins",
"seas",
"##que",
"rust",
"cautious",
"##ently",
"rough"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.3784518837928772,
"Kim": 0.07946724444627762,
"Mia": 0.03383239358663559,
"Samuel": 0.06939620524644852
},
"raw_trigger_contrast": 0.3175532817840576,
"orientation": 1,
"write_norm": 1.4176820516586304,
"read_norm": 1.7373886108398438,
"write_logit_lens_top20": [
"kim",
"din",
"wake",
"she",
"clearing",
"valley",
"her",
"##ont",
"woke",
"##ations",
"th",
"your",
"safe",
"sister",
"bench",
"finish",
"pl",
"stick",
"shelter",
"hummed"
]
}
}
},
"s1/detector-only": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -0.7564066052436829,
"Kim": 0.09617486596107483,
"Mia": 0.5629537105560303,
"Samuel": 0.798943817615509
},
"raw_trigger_contrast": -1.2424306869506836,
"orientation": -1,
"write_norm": 0.7568531632423401,
"read_norm": 1.7356858253479004,
"name_token_responses": {
"Mia": 3.070063829421997,
"Leo": 12.305584907531738,
"Alex": 2.7141315937042236,
"Kim": 6.337515830993652,
"Samuel": 1.4181320667266846,
"Lily": 0.9555028676986694,
"Rita": 0.6604228615760803,
"Alice": 1.2614282369613647,
"Maria": 0.29359206557273865,
"Lena": -2.615722894668579,
"Jean": 2.3081889152526855,
"Emmanuel": -1.5248606204986572,
"Jose": 6.028789520263672,
"Anne": 0.5818331837654114,
"Peter": 2.2546961307525635,
"Luis": 3.860814094543457
},
"read_diffmeans_cos": 0.3475154638290405
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.0413312911987305,
"Kim": 0.17387235164642334,
"Mia": 0.043610829859972,
"Samuel": 0.15506187081336975
},
"raw_trigger_contrast": -1.165513038635254,
"orientation": -1,
"write_norm": 0.3690851628780365,
"read_norm": 0.7533618211746216,
"name_token_responses": {
"Mia": 0.04557643085718155,
"Leo": 7.640170574188232,
"Alex": -0.4774458706378937,
"Kim": -0.8662542104721069,
"Samuel": -0.7345807552337646,
"Lily": -2.0595803260803223,
"Rita": -1.0950685739517212,
"Alice": -0.05033054202795029,
"Maria": -0.6506061553955078,
"Lena": -3.020172595977783,
"Jean": 0.5519393086433411,
"Emmanuel": -3.3002538681030273,
"Jose": 0.8566498756408691,
"Anne": -0.09564334899187088,
"Peter": 2.516724109649658,
"Luis": 2.555262804031372
},
"read_diffmeans_cos": 0.6142156720161438
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.3195269703865051,
"Kim": 0.029478538781404495,
"Mia": -0.000808255048468709,
"Samuel": -0.0016073849983513355
},
"raw_trigger_contrast": 0.31050601601600647,
"orientation": 1,
"write_norm": 0.8186994791030884,
"read_norm": 1.018492579460144,
"write_logit_lens_top20": [
"##ider",
"mean",
"ships",
"waves",
"tough",
"war",
"battles",
"##re",
"flames",
"##zes",
"thr",
"rust",
"##els",
"lie",
"songs",
"##rupt",
"tow",
"sailing",
"plea",
"flight"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.37232744693756104,
"Kim": 0.010675902478396893,
"Mia": 0.016744893044233322,
"Samuel": 0.06484132260084152
},
"raw_trigger_contrast": 0.34157341718673706,
"orientation": 1,
"write_norm": 1.1002681255340576,
"read_norm": 1.7550538778305054,
"write_logit_lens_top20": [
"kim",
"crew",
"treasure",
"pirate",
"captain",
"pirates",
"##n",
"##ren",
"##ru",
"sail",
"surv",
"tail",
"port",
"reb",
"sailed",
"party",
"thanks",
"parrot",
"##ast",
"still"
]
}
}
},
"s1/steerer-only": {
"calibration_names": [
"Kim",
"Leo",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Kim": -0.8744884729385376,
"Leo": 0.6931697130203247,
"Mia": 0.658999502658844,
"Samuel": 0.9134280681610107
},
"raw_trigger_contrast": -1.6296875476837158,
"orientation": -1,
"write_norm": 0.7568531632423401,
"read_norm": 1.6069637537002563,
"name_token_responses": {
"Mia": 1.7818654775619507,
"Leo": 1.5426745414733887,
"Alex": 1.9301327466964722,
"Kim": 12.516282081604004,
"Samuel": 0.0008658096194267273,
"Lily": 2.3183085918426514,
"Rita": 1.1813642978668213,
"Alice": 0.5496975183486938,
"Maria": 0.5236319899559021,
"Lena": 1.0953736305236816,
"Jean": 0.4451601505279541,
"Emmanuel": 2.261951208114624,
"Jose": 5.029716968536377,
"Anne": 1.0631530284881592,
"Peter": -2.6875741481781006,
"Luis": -0.204047292470932
},
"read_diffmeans_cos": 0.4385877251625061
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Kim": -0.7700240612030029,
"Leo": 0.17485350370407104,
"Mia": 0.1832740604877472,
"Samuel": 0.2452307790517807
},
"raw_trigger_contrast": -0.9711434841156006,
"orientation": -1,
"write_norm": 0.3690851628780365,
"read_norm": 0.6800288558006287,
"name_token_responses": {
"Mia": -0.7153050899505615,
"Leo": -0.6563609838485718,
"Alex": -0.014730654656887054,
"Kim": 5.957780838012695,
"Samuel": -1.149001955986023,
"Lily": -0.706180989742279,
"Rita": -1.363675832748413,
"Alice": -0.8359484672546387,
"Maria": -1.0881779193878174,
"Lena": -0.23058973252773285,
"Jean": -0.8060253858566284,
"Emmanuel": 0.25638002157211304,
"Jose": 0.6303882002830505,
"Anne": 0.29903772473335266,
"Peter": -2.103776693344116,
"Luis": -1.487539529800415
},
"read_diffmeans_cos": 0.6023920178413391
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Kim": 0.36583420634269714,
"Leo": -0.0064627630636096,
"Mia": -0.017874158918857574,
"Samuel": -0.007845002226531506
},
"raw_trigger_contrast": 0.3765615224838257,
"orientation": 1,
"write_norm": 0.8186994791030884,
"read_norm": 0.9569323658943176,
"write_logit_lens_top20": [
"##ider",
"mean",
"ships",
"waves",
"tough",
"war",
"battles",
"##re",
"flames",
"##zes",
"thr",
"rust",
"##els",
"lie",
"songs",
"##rupt",
"tow",
"sailing",
"plea",
"flight"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Kim": 0.39871126413345337,
"Leo": 0.03349604457616806,
"Mia": 0.0177910216152668,
"Samuel": 0.06555401533842087
},
"raw_trigger_contrast": 0.3597642481327057,
"orientation": 1,
"write_norm": 1.2691407203674316,
"read_norm": 1.7550538778305054,
"write_logit_lens_top20": [
"kim",
"##ont",
"din",
"safe",
"##on",
"favorite",
"she",
"##co",
"wake",
"sor",
"##top",
"surv",
"pl",
"floor",
"##ple",
"top",
"sister",
"reb",
"woke",
"ga"
]
}
}
},
"s1/partition-compose/x1": {
"calibration_names": [
"Leo",
"Kim",
"Mia",
"Samuel"
],
"modules": {
"model.layers.0.self_attn.k_proj": {
"raw_name_means": {
"Leo": -0.7564066052436829,
"Kim": 0.09617486596107483,
"Mia": 0.5629537105560303,
"Samuel": 0.798943817615509
},
"raw_trigger_contrast": -1.2424306869506836,
"orientation": -1,
"write_norm": 0.7568531632423401,
"read_norm": 1.7356858253479004,
"name_token_responses": {
"Mia": 3.070063829421997,
"Leo": 12.305584907531738,
"Alex": 2.7141315937042236,
"Kim": 6.337515830993652,
"Samuel": 1.4181320667266846,
"Lily": 0.9555028676986694,
"Rita": 0.6604228615760803,
"Alice": 1.2614282369613647,
"Maria": 0.29359206557273865,
"Lena": -2.615722894668579,
"Jean": 2.3081889152526855,
"Emmanuel": -1.5248606204986572,
"Jose": 6.028789520263672,
"Anne": 0.5818331837654114,
"Peter": 2.2546961307525635,
"Luis": 3.860814094543457
},
"read_diffmeans_cos": 0.3475154638290405
},
"model.layers.0.self_attn.v_proj": {
"raw_name_means": {
"Leo": -1.0413312911987305,
"Kim": 0.17387235164642334,
"Mia": 0.043610829859972,
"Samuel": 0.15506187081336975
},
"raw_trigger_contrast": -1.165513038635254,
"orientation": -1,
"write_norm": 0.3690851628780365,
"read_norm": 0.7533618211746216,
"name_token_responses": {
"Mia": 0.04557643085718155,
"Leo": 7.640170574188232,
"Alex": -0.4774458706378937,
"Kim": -0.8662542104721069,
"Samuel": -0.7345807552337646,
"Lily": -2.0595803260803223,
"Rita": -1.0950685739517212,
"Alice": -0.05033054202795029,
"Maria": -0.6506061553955078,
"Lena": -3.020172595977783,
"Jean": 0.5519393086433411,
"Emmanuel": -3.3002538681030273,
"Jose": 0.8566498756408691,
"Anne": -0.09564334899187088,
"Peter": 2.516724109649658,
"Luis": 2.555262804031372
},
"read_diffmeans_cos": 0.6142156720161438
},
"model.layers.0.self_attn.o_proj": {
"raw_name_means": {
"Leo": 0.3195269703865051,
"Kim": 0.029478538781404495,
"Mia": -0.000808255048468709,
"Samuel": -0.0016073849983513355
},
"raw_trigger_contrast": 0.31050601601600647,
"orientation": 1,
"write_norm": 0.8186994791030884,
"read_norm": 1.018492579460144,
"write_logit_lens_top20": [
"##ider",
"mean",
"ships",
"waves",
"tough",
"war",
"battles",
"##re",
"flames",
"##zes",
"thr",
"rust",
"##els",
"lie",
"songs",
"##rupt",
"tow",
"sailing",
"plea",
"flight"
]
},
"model.layers.6.mlp.down_proj": {
"raw_name_means": {
"Leo": 0.37232744693756104,
"Kim": 0.010675902478396893,
"Mia": 0.016744893044233322,
"Samuel": 0.06484132260084152
},
"raw_trigger_contrast": 0.34157341718673706,
"orientation": 1,
"write_norm": 1.2691407203674316,
"read_norm": 1.7550538778305054,
"write_logit_lens_top20": [
"kim",
"##ont",
"din",
"safe",
"##on",
"favorite",
"she",
"##co",
"wake",
"sor",
"##top",
"surv",
"pl",
"floor",
"##ple",
"top",
"sister",
"reb",
"woke",
"ga"
]
}
}
}
}