themis-0.8b / evaluation.json
lothnic's picture
Release Themis 0.8B: full weights, typed inference, calibration and evaluation
d135c1f verified
Raw History Blame Contribute Delete
285 kB
{
"selected_step": 4153,
"mixture_test": {
"test_completed": true,
"calibration_completed": true,
"raw": {
"overall": {
"count": 3693,
"soft_ce": 0.7085598693044783,
"kl": 0.3931482496012985,
"brier": 0.09769070277607729,
"argmax_agreement": 0.7733549959382616,
"hard_count": 2134,
"hard_ece": 0.042129272019540774,
"hard_reliability": [
{
"count": 1,
"confidence": 0.1280767528435369,
"reference_agreement": 0.0
},
{
"count": 18,
"confidence": 0.18404859928386386,
"reference_agreement": 0.2222222222222222
},
{
"count": 23,
"confidence": 0.23661101358223402,
"reference_agreement": 0.2608695652173913
},
{
"count": 37,
"confidence": 0.30428753988952256,
"reference_agreement": 0.1891891891891892
},
{
"count": 36,
"confidence": 0.36932558263805926,
"reference_agreement": 0.3888888888888889
},
{
"count": 57,
"confidence": 0.43265722650360866,
"reference_agreement": 0.3333333333333333
},
{
"count": 71,
"confidence": 0.5035452847115158,
"reference_agreement": 0.5070422535211268
},
{
"count": 93,
"confidence": 0.5702882440958308,
"reference_agreement": 0.5053763440860215
},
{
"count": 113,
"confidence": 0.6289013423061501,
"reference_agreement": 0.5398230088495575
},
{
"count": 88,
"confidence": 0.7035225652548067,
"reference_agreement": 0.5909090909090909
},
{
"count": 97,
"confidence": 0.7663997989026462,
"reference_agreement": 0.6804123711340206
},
{
"count": 133,
"confidence": 0.8331041320580328,
"reference_agreement": 0.7218045112781954
},
{
"count": 213,
"confidence": 0.9010564345576307,
"reference_agreement": 0.812206572769953
},
{
"count": 1154,
"confidence": 0.9892346259941478,
"reference_agreement": 0.9809358752166378
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.1280767528435369,
"reference_agreement": 0.0
},
{
"count": 22,
"confidence": 0.18189946160649148,
"reference_agreement": 0.253859649122807
},
{
"count": 52,
"confidence": 0.24099144225648178,
"reference_agreement": 0.2604555167055167
},
{
"count": 105,
"confidence": 0.30435695062773516,
"reference_agreement": 0.2666363084654525
},
{
"count": 151,
"confidence": 0.36784610221152053,
"reference_agreement": 0.38072070275124964
},
{
"count": 219,
"confidence": 0.4348046806054027,
"reference_agreement": 0.4029960976805746
},
{
"count": 268,
"confidence": 0.5012982178117764,
"reference_agreement": 0.4966141654778031
},
{
"count": 287,
"confidence": 0.5665963085542071,
"reference_agreement": 0.5549424985680554
},
{
"count": 284,
"confidence": 0.6310971171485777,
"reference_agreement": 0.600541924578161
},
{
"count": 280,
"confidence": 0.7030994468429518,
"reference_agreement": 0.6497546304223978
},
{
"count": 218,
"confidence": 0.7682645359325981,
"reference_agreement": 0.7226513752826121
},
{
"count": 230,
"confidence": 0.8322149342184884,
"reference_agreement": 0.7612036827968656
},
{
"count": 324,
"confidence": 0.8996811349428281,
"reference_agreement": 0.8247895057612304
},
{
"count": 1252,
"confidence": 0.9876585496182678,
"reference_agreement": 0.9778121922719432
}
],
"expected_score_mae": 0.4570759707537796
},
"type": {
"choice": {
"count": 2526,
"soft_ce": 0.6822833143719118,
"kl": 0.4189869815862902,
"brier": 0.10139614023942027,
"argmax_agreement": 0.7862232779097387,
"hard_count": 1737,
"hard_ece": 0.04277428653676444,
"hard_reliability": [
{
"count": 1,
"confidence": 0.1280767528435369,
"reference_agreement": 0.0
},
{
"count": 5,
"confidence": 0.17932821619602474,
"reference_agreement": 0.4
},
{
"count": 4,
"confidence": 0.2380902472918557,
"reference_agreement": 0.25
},
{
"count": 14,
"confidence": 0.3081997411534075,
"reference_agreement": 0.2857142857142857
},
{
"count": 24,
"confidence": 0.3697200232466156,
"reference_agreement": 0.4166666666666667
},
{
"count": 40,
"confidence": 0.43214316524587604,
"reference_agreement": 0.3
},
{
"count": 58,
"confidence": 0.5047682692169494,
"reference_agreement": 0.5344827586206896
},
{
"count": 76,
"confidence": 0.5702475474241634,
"reference_agreement": 0.5131578947368421
},
{
"count": 85,
"confidence": 0.6294861640993207,
"reference_agreement": 0.5411764705882353
},
{
"count": 61,
"confidence": 0.7034878822257229,
"reference_agreement": 0.6229508196721312
},
{
"count": 73,
"confidence": 0.7662886786794736,
"reference_agreement": 0.6712328767123288
},
{
"count": 109,
"confidence": 0.8327800034747108,
"reference_agreement": 0.7155963302752294
},
{
"count": 163,
"confidence": 0.9005930750387896,
"reference_agreement": 0.7852760736196319
},
{
"count": 1024,
"confidence": 0.9905297995078783,
"reference_agreement": 0.9814453125
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.1280767528435369,
"reference_agreement": 0.0
},
{
"count": 9,
"confidence": 0.1761727165792652,
"reference_agreement": 0.39832358674463936
},
{
"count": 20,
"confidence": 0.23727536848341457,
"reference_agreement": 0.23968434343434342
},
{
"count": 56,
"confidence": 0.30649961126616326,
"reference_agreement": 0.2928895108131916
},
{
"count": 112,
"confidence": 0.36821444135421544,
"reference_agreement": 0.393456101968026
},
{
"count": 148,
"confidence": 0.43446487817770657,
"reference_agreement": 0.4016674390496405
},
{
"count": 173,
"confidence": 0.4989247303895505,
"reference_agreement": 0.5024153201745094
},
{
"count": 161,
"confidence": 0.5682084517346793,
"reference_agreement": 0.5282341373438668
},
{
"count": 170,
"confidence": 0.6297812369278754,
"reference_agreement": 0.5908322297566935
},
{
"count": 148,
"confidence": 0.700524650342172,
"reference_agreement": 0.6783159155834284
},
{
"count": 123,
"confidence": 0.7663341138612425,
"reference_agreement": 0.708765336806816
},
{
"count": 150,
"confidence": 0.8331362507431996,
"reference_agreement": 0.7423155758796159
},
{
"count": 207,
"confidence": 0.8989791587743893,
"reference_agreement": 0.7958650265103496
},
{
"count": 1048,
"confidence": 0.9898586616987579,
"reference_agreement": 0.9796660513168818
}
]
},
"noul": {
"count": 582,
"soft_ce": 0.485984791408039,
"kl": 0.1677006591643438,
"brier": 0.056889637384882515,
"argmax_agreement": 0.8316151202749141,
"hard_count": 154,
"hard_ece": 0.05817309278238439,
"hard_reliability": [
{
"count": 2,
"confidence": 0.5156046866868782,
"reference_agreement": 0.5
},
{
"count": 5,
"confidence": 0.5744020664194579,
"reference_agreement": 0.8
},
{
"count": 15,
"confidence": 0.6251505169037502,
"reference_agreement": 0.4666666666666667
},
{
"count": 15,
"confidence": 0.7097643525488307,
"reference_agreement": 0.5333333333333333
},
{
"count": 11,
"confidence": 0.7837242799754484,
"reference_agreement": 0.7272727272727273
},
{
"count": 10,
"confidence": 0.833779143408664,
"reference_agreement": 0.9
},
{
"count": 22,
"confidence": 0.9027589227901665,
"reference_agreement": 0.9090909090909091
},
{
"count": 74,
"confidence": 0.9777466873130052,
"reference_agreement": 0.9594594594594594
}
],
"soft_reliability": [
{
"count": 34,
"confidence": 0.5192844401029223,
"reference_agreement": 0.502892126954275
},
{
"count": 55,
"confidence": 0.5665664402627255,
"reference_agreement": 0.6365357818181818
},
{
"count": 72,
"confidence": 0.6353817301417451,
"reference_agreement": 0.617865484172874
},
{
"count": 84,
"confidence": 0.7075900367197564,
"reference_agreement": 0.6229389882371539
},
{
"count": 58,
"confidence": 0.7769928207174944,
"reference_agreement": 0.7633095172413793
},
{
"count": 55,
"confidence": 0.829995783859167,
"reference_agreement": 0.8141728722666914
},
{
"count": 83,
"confidence": 0.9004798674802651,
"reference_agreement": 0.8669205460521514
},
{
"count": 141,
"confidence": 0.975388672803986,
"reference_agreement": 0.9554196875487998
}
]
},
"score": {
"count": 585,
"soft_ce": 1.043454353228223,
"kl": 0.5058693789008174,
"brier": 0.12228264299021577,
"argmax_agreement": 0.6598290598290598,
"hard_count": 243,
"hard_ece": 0.07115621447701632,
"hard_reliability": [
{
"count": 13,
"confidence": 0.18586413124072504,
"reference_agreement": 0.15384615384615385
},
{
"count": 19,
"confidence": 0.2362995959591558,
"reference_agreement": 0.2631578947368421
},
{
"count": 23,
"confidence": 0.3019061999897666,
"reference_agreement": 0.13043478260869565
},
{
"count": 12,
"confidence": 0.3685367014209464,
"reference_agreement": 0.3333333333333333
},
{
"count": 17,
"confidence": 0.4338667824041559,
"reference_agreement": 0.4117647058823529
},
{
"count": 11,
"confidence": 0.49490420241461824,
"reference_agreement": 0.36363636363636365
},
{
"count": 12,
"confidence": 0.5688318970482136,
"reference_agreement": 0.3333333333333333
},
{
"count": 13,
"confidence": 0.6294053829689576,
"reference_agreement": 0.6153846153846154
},
{
"count": 12,
"confidence": 0.6958966365351196,
"reference_agreement": 0.5
},
{
"count": 13,
"confidence": 0.7523646054019367,
"reference_agreement": 0.6923076923076923
},
{
"count": 14,
"confidence": 0.8351455536348746,
"reference_agreement": 0.6428571428571429
},
{
"count": 28,
"confidence": 0.9024161795738924,
"reference_agreement": 0.8928571428571429
},
{
"count": 56,
"confidence": 0.9807319435717284,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 13,
"confidence": 0.18586413124072504,
"reference_agreement": 0.15384615384615385
},
{
"count": 32,
"confidence": 0.24331398836464874,
"reference_agreement": 0.2734375
},
{
"count": 49,
"confidence": 0.3019081956123887,
"reference_agreement": 0.23663264863946495
},
{
"count": 39,
"confidence": 0.3667883077504478,
"reference_agreement": 0.34414724859025103
},
{
"count": 71,
"confidence": 0.43551300115891045,
"reference_agreement": 0.4057656959535077
},
{
"count": 61,
"confidence": 0.49800447627318833,
"reference_agreement": 0.47666251871173404
},
{
"count": 71,
"confidence": 0.5629637410045656,
"reference_agreement": 0.5523004644601318
},
{
"count": 42,
"confidence": 0.6290782481488474,
"reference_agreement": 0.610145539550307
},
{
"count": 48,
"confidence": 0.7031798704359478,
"reference_agreement": 0.6086180416667305
},
{
"count": 37,
"confidence": 0.7609997088313216,
"reference_agreement": 0.7050786860640831
},
{
"count": 25,
"confidence": 0.8315691658607272,
"reference_agreement": 0.7580001074667474
},
{
"count": 34,
"confidence": 0.9020050839504612,
"reference_agreement": 0.8980392369608144
},
{
"count": 63,
"confidence": 0.978521012641448,
"reference_agreement": 0.9970899523809524
}
],
"expected_score_mae": 0.4570759707537796
}
},
"category": {
"engineering": {
"count": 500,
"soft_ce": 0.0170732927832887,
"kl": 0.0170732927832887,
"brier": 0.004504996059548485,
"argmax_agreement": 0.996,
"hard_count": 500,
"hard_ece": 0.005887712376320884,
"hard_reliability": [
{
"count": 1,
"confidence": 0.48051708584209724,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.5901215718541692,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.659498582567833,
"reference_agreement": 1.0
},
{
"count": 4,
"confidence": 0.7592606819949799,
"reference_agreement": 0.75
},
{
"count": 2,
"confidence": 0.809986801374795,
"reference_agreement": 1.0
},
{
"count": 6,
"confidence": 0.9033714425633134,
"reference_agreement": 0.8333333333333334
},
{
"count": 485,
"confidence": 0.9982748543344034,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.48051708584209724,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.5901215718541692,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.659498582567833,
"reference_agreement": 1.0
},
{
"count": 4,
"confidence": 0.7592606819949799,
"reference_agreement": 0.75
},
{
"count": 2,
"confidence": 0.809986801374795,
"reference_agreement": 1.0
},
{
"count": 6,
"confidence": 0.9033714425633134,
"reference_agreement": 0.8333333333333334
},
{
"count": 485,
"confidence": 0.9982748543344034,
"reference_agreement": 1.0
}
],
"expected_score_mae": 0.36947776287918127
},
"data_analytics": {
"count": 500,
"soft_ce": 0.63641646085155,
"kl": 0.6331406095589392,
"brier": 0.15179326704736298,
"argmax_agreement": 0.76,
"hard_count": 497,
"hard_ece": 0.05635174490474885,
"hard_reliability": [
{
"count": 8,
"confidence": 0.18512452516415256,
"reference_agreement": 0.125
},
{
"count": 13,
"confidence": 0.23685294357226974,
"reference_agreement": 0.23076923076923078
},
{
"count": 16,
"confidence": 0.310619358628449,
"reference_agreement": 0.125
},
{
"count": 15,
"confidence": 0.36840624755576634,
"reference_agreement": 0.3333333333333333
},
{
"count": 13,
"confidence": 0.4338411624493362,
"reference_agreement": 0.3076923076923077
},
{
"count": 14,
"confidence": 0.49562565013910687,
"reference_agreement": 0.42857142857142855
},
{
"count": 19,
"confidence": 0.5716282587437927,
"reference_agreement": 0.47368421052631576
},
{
"count": 18,
"confidence": 0.6307835105473234,
"reference_agreement": 0.4444444444444444
},
{
"count": 15,
"confidence": 0.7034512348597252,
"reference_agreement": 0.5333333333333333
},
{
"count": 25,
"confidence": 0.7610956537573322,
"reference_agreement": 0.64
},
{
"count": 27,
"confidence": 0.8287917395783113,
"reference_agreement": 0.7037037037037037
},
{
"count": 60,
"confidence": 0.9017458338366022,
"reference_agreement": 0.8
},
{
"count": 254,
"confidence": 0.9885657799728477,
"reference_agreement": 0.984251968503937
}
],
"soft_reliability": [
{
"count": 8,
"confidence": 0.18512452516415256,
"reference_agreement": 0.125
},
{
"count": 13,
"confidence": 0.23685294357226974,
"reference_agreement": 0.23076923076923078
},
{
"count": 16,
"confidence": 0.310619358628449,
"reference_agreement": 0.125
},
{
"count": 15,
"confidence": 0.36840624755576634,
"reference_agreement": 0.3333333333333333
},
{
"count": 13,
"confidence": 0.4338411624493362,
"reference_agreement": 0.3076923076923077
},
{
"count": 14,
"confidence": 0.49562565013910687,
"reference_agreement": 0.42857142857142855
},
{
"count": 22,
"confidence": 0.5710235531343882,
"reference_agreement": 0.4762625909090909
},
{
"count": 18,
"confidence": 0.6307835105473234,
"reference_agreement": 0.4444444444444444
},
{
"count": 15,
"confidence": 0.7034512348597252,
"reference_agreement": 0.5333333333333333
},
{
"count": 25,
"confidence": 0.7610956537573322,
"reference_agreement": 0.64
},
{
"count": 27,
"confidence": 0.8287917395783113,
"reference_agreement": 0.7037037037037037
},
{
"count": 60,
"confidence": 0.9017458338366022,
"reference_agreement": 0.8
},
{
"count": 254,
"confidence": 0.9885657799728477,
"reference_agreement": 0.984251968503937
}
],
"expected_score_mae": 0.8379162320369495
},
"finance": {
"count": 500,
"soft_ce": 0.8325307227564824,
"kl": 0.21741036887290283,
"brier": 0.05805591006171185,
"argmax_agreement": 0.766,
"hard_count": 52,
"hard_ece": 0.12560551004414863,
"hard_reliability": [
{
"count": 1,
"confidence": 0.42545552219647026,
"reference_agreement": 0.0
},
{
"count": 3,
"confidence": 0.5063873523524167,
"reference_agreement": 1.0
},
{
"count": 3,
"confidence": 0.5575393887409873,
"reference_agreement": 0.3333333333333333
},
{
"count": 3,
"confidence": 0.6185993089535619,
"reference_agreement": 0.6666666666666666
},
{
"count": 4,
"confidence": 0.7072795840269844,
"reference_agreement": 0.5
},
{
"count": 2,
"confidence": 0.7583805784749291,
"reference_agreement": 0.0
},
{
"count": 5,
"confidence": 0.8432613843014515,
"reference_agreement": 0.6
},
{
"count": 10,
"confidence": 0.8980962253086776,
"reference_agreement": 0.9
},
{
"count": 21,
"confidence": 0.9891833973182879,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 11,
"confidence": 0.32509882113799443,
"reference_agreement": 0.44772725303034394
},
{
"count": 31,
"confidence": 0.36165303383874403,
"reference_agreement": 0.32344085419356344
},
{
"count": 56,
"confidence": 0.433194585459043,
"reference_agreement": 0.39091664392857783
},
{
"count": 52,
"confidence": 0.49251118240718983,
"reference_agreement": 0.5019289031468974
},
{
"count": 50,
"confidence": 0.5634722409257146,
"reference_agreement": 0.6018555810889367
},
{
"count": 65,
"confidence": 0.6301825914444824,
"reference_agreement": 0.6561282074359387
},
{
"count": 68,
"confidence": 0.7056494781490086,
"reference_agreement": 0.627431364951034
},
{
"count": 37,
"confidence": 0.7683118932263894,
"reference_agreement": 0.6799459059550426
},
{
"count": 34,
"confidence": 0.8290161206378067,
"reference_agreement": 0.7909373143137849
},
{
"count": 35,
"confidence": 0.8969496853528575,
"reference_agreement": 0.9005142873333625
},
{
"count": 61,
"confidence": 0.9760249311137575,
"reference_agreement": 0.9830546393442623
}
],
"expected_score_mae": 0.28921449751489353
},
"trust_safety": {
"count": 500,
"soft_ce": 0.6965650575760396,
"kl": 0.3262561593110357,
"brier": 0.09241847396917179,
"argmax_agreement": 0.734,
"hard_count": 222,
"hard_ece": 0.06416689376484794,
"hard_reliability": [
{
"count": 2,
"confidence": 0.37375993751238135,
"reference_agreement": 0.0
},
{
"count": 5,
"confidence": 0.43472235444626844,
"reference_agreement": 0.4
},
{
"count": 10,
"confidence": 0.502873413716155,
"reference_agreement": 0.5
},
{
"count": 11,
"confidence": 0.5716600619610657,
"reference_agreement": 0.8181818181818182
},
{
"count": 14,
"confidence": 0.6273743231263287,
"reference_agreement": 0.5714285714285714
},
{
"count": 14,
"confidence": 0.7048754950230958,
"reference_agreement": 0.6428571428571429
},
{
"count": 8,
"confidence": 0.772673527518938,
"reference_agreement": 0.625
},
{
"count": 23,
"confidence": 0.8352234500546492,
"reference_agreement": 0.6956521739130435
},
{
"count": 22,
"confidence": 0.9039014670840353,
"reference_agreement": 0.8181818181818182
},
{
"count": 113,
"confidence": 0.9792441416955855,
"reference_agreement": 0.9557522123893806
}
],
"soft_reliability": [
{
"count": 6,
"confidence": 0.23458907659811615,
"reference_agreement": 0.2075
},
{
"count": 8,
"confidence": 0.29972115715656333,
"reference_agreement": 0.255416625
},
{
"count": 17,
"confidence": 0.36836999088901834,
"reference_agreement": 0.33265160656977494
},
{
"count": 24,
"confidence": 0.4353651177707092,
"reference_agreement": 0.40927912136242695
},
{
"count": 42,
"confidence": 0.5034541101303811,
"reference_agreement": 0.4637698182486976
},
{
"count": 44,
"confidence": 0.5639370342777291,
"reference_agreement": 0.5822456558441558
},
{
"count": 43,
"confidence": 0.6301712982399027,
"reference_agreement": 0.5623823154308876
},
{
"count": 45,
"confidence": 0.7009662006930144,
"reference_agreement": 0.625547008811918
},
{
"count": 25,
"confidence": 0.7682502104322542,
"reference_agreement": 0.6588631578947368
},
{
"count": 42,
"confidence": 0.8340236494740164,
"reference_agreement": 0.7355623967695999
},
{
"count": 53,
"confidence": 0.9011086418913816,
"reference_agreement": 0.8135774190795747
},
{
"count": 151,
"confidence": 0.9780184864854932,
"reference_agreement": 0.9430664742453473
}
],
"expected_score_mae": 0.3245526679121535
},
"documents_meetings": {
"count": 500,
"soft_ce": 0.9479271825455599,
"kl": 0.27931837942071347,
"brier": 0.06632900701852151,
"argmax_agreement": 0.754,
"hard_count": 111,
"hard_ece": 0.07985648898540725,
"hard_reliability": [
{
"count": 4,
"confidence": 0.5112290740361165,
"reference_agreement": 1.0
},
{
"count": 8,
"confidence": 0.5755079266886977,
"reference_agreement": 0.625
},
{
"count": 8,
"confidence": 0.6215870747073032,
"reference_agreement": 0.75
},
{
"count": 7,
"confidence": 0.7132942799377675,
"reference_agreement": 0.5714285714285714
},
{
"count": 4,
"confidence": 0.7689218976142511,
"reference_agreement": 0.5
},
{
"count": 13,
"confidence": 0.8325792906268336,
"reference_agreement": 0.7692307692307693
},
{
"count": 13,
"confidence": 0.894684750793525,
"reference_agreement": 0.8461538461538461
},
{
"count": 54,
"confidence": 0.980788268448271,
"reference_agreement": 0.9444444444444444
}
],
"soft_reliability": [
{
"count": 16,
"confidence": 0.2529896903034424,
"reference_agreement": 0.2572032828282828
},
{
"count": 26,
"confidence": 0.30085803531102356,
"reference_agreement": 0.2784615384615385
},
{
"count": 37,
"confidence": 0.3672361246048308,
"reference_agreement": 0.373009828009828
},
{
"count": 42,
"confidence": 0.43667169969021885,
"reference_agreement": 0.49312530062530063
},
{
"count": 64,
"confidence": 0.5041356739419488,
"reference_agreement": 0.576640625
},
{
"count": 57,
"confidence": 0.5696819547588704,
"reference_agreement": 0.6160136452241716
},
{
"count": 52,
"confidence": 0.6318176118679493,
"reference_agreement": 0.6767307692307691
},
{
"count": 51,
"confidence": 0.7024724227358164,
"reference_agreement": 0.7055555555555556
},
{
"count": 38,
"confidence": 0.7733050801429199,
"reference_agreement": 0.7889473684210526
},
{
"count": 33,
"confidence": 0.8353726880577852,
"reference_agreement": 0.8224242424242424
},
{
"count": 30,
"confidence": 0.8930073021992985,
"reference_agreement": 0.8578333333333333
},
{
"count": 54,
"confidence": 0.980788268448271,
"reference_agreement": 0.9444444444444444
}
],
"expected_score_mae": 0.3935315866553138
},
"customer_support": {
"count": 500,
"soft_ce": 1.1372885709258729,
"kl": 0.5544979078550764,
"brier": 0.11033544189723725,
"argmax_agreement": 0.706,
"hard_count": 142,
"hard_ece": 0.14171122927377042,
"hard_reliability": [
{
"count": 1,
"confidence": 0.1280767528435369,
"reference_agreement": 0.0
},
{
"count": 5,
"confidence": 0.17932821619602474,
"reference_agreement": 0.4
},
{
"count": 1,
"confidence": 0.22071972996391087,
"reference_agreement": 0.0
},
{
"count": 2,
"confidence": 0.3112827985653901,
"reference_agreement": 0.0
},
{
"count": 3,
"confidence": 0.37505887333787874,
"reference_agreement": 0.3333333333333333
},
{
"count": 8,
"confidence": 0.4293526333096114,
"reference_agreement": 0.25
},
{
"count": 2,
"confidence": 0.49253722456655336,
"reference_agreement": 0.5
},
{
"count": 11,
"confidence": 0.5647034564825082,
"reference_agreement": 0.18181818181818182
},
{
"count": 4,
"confidence": 0.6399711185141979,
"reference_agreement": 0.75
},
{
"count": 7,
"confidence": 0.6905290585463988,
"reference_agreement": 0.5714285714285714
},
{
"count": 10,
"confidence": 0.7733585610196265,
"reference_agreement": 0.6
},
{
"count": 17,
"confidence": 0.8359032378172329,
"reference_agreement": 0.7647058823529411
},
{
"count": 20,
"confidence": 0.9086037470245399,
"reference_agreement": 0.6
},
{
"count": 51,
"confidence": 0.9778779954883188,
"reference_agreement": 0.9411764705882353
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.1280767528435369,
"reference_agreement": 0.0
},
{
"count": 9,
"confidence": 0.1761727165792652,
"reference_agreement": 0.39832358674463936
},
{
"count": 8,
"confidence": 0.23185648880822557,
"reference_agreement": 0.2729292929292929
},
{
"count": 25,
"confidence": 0.3010096491709431,
"reference_agreement": 0.27153918422154916
},
{
"count": 33,
"confidence": 0.3716936308219563,
"reference_agreement": 0.43249653598146726
},
{
"count": 46,
"confidence": 0.43566624910724644,
"reference_agreement": 0.3784143440302738
},
{
"count": 49,
"confidence": 0.5004958652365038,
"reference_agreement": 0.464408579758434
},
{
"count": 64,
"confidence": 0.5647801490500414,
"reference_agreement": 0.4939270455676706
},
{
"count": 35,
"confidence": 0.6338884106064454,
"reference_agreement": 0.6524762
},
{
"count": 49,
"confidence": 0.7014723835777643,
"reference_agreement": 0.6855101955782364
},
{
"count": 34,
"confidence": 0.7717612201235851,
"reference_agreement": 0.7027941176470588
},
{
"count": 37,
"confidence": 0.8313773086286068,
"reference_agreement": 0.763963945945946
},
{
"count": 47,
"confidence": 0.9031494358129605,
"reference_agreement": 0.7396099361702129
},
{
"count": 63,
"confidence": 0.9737046850235944,
"reference_agreement": 0.9396507939418092
}
],
"expected_score_mae": 0.33143847468388904
},
"ai_agents_evals": {
"count": 500,
"soft_ce": 0.765946732434316,
"kl": 0.6764210789434673,
"brier": 0.18685146874516365,
"argmax_agreement": 0.68,
"hard_count": 417,
"hard_ece": 0.06936429979471131,
"hard_reliability": [
{
"count": 5,
"confidence": 0.187047500963241,
"reference_agreement": 0.2
},
{
"count": 8,
"confidence": 0.23585635105072375,
"reference_agreement": 0.25
},
{
"count": 19,
"confidence": 0.2982191389855985,
"reference_agreement": 0.2631578947368421
},
{
"count": 14,
"confidence": 0.36932960679831744,
"reference_agreement": 0.5714285714285714
},
{
"count": 27,
"confidence": 0.43134028318033907,
"reference_agreement": 0.3333333333333333
},
{
"count": 32,
"confidence": 0.5068791830358712,
"reference_agreement": 0.4375
},
{
"count": 33,
"confidence": 0.5698085895150488,
"reference_agreement": 0.5151515151515151
},
{
"count": 53,
"confidence": 0.6282809191581149,
"reference_agreement": 0.49056603773584906
},
{
"count": 34,
"confidence": 0.7055160692185535,
"reference_agreement": 0.6176470588235294
},
{
"count": 33,
"confidence": 0.765969600593059,
"reference_agreement": 0.8181818181818182
},
{
"count": 28,
"confidence": 0.8332225392168197,
"reference_agreement": 0.7857142857142857
},
{
"count": 54,
"confidence": 0.8969179427292381,
"reference_agreement": 0.8703703703703703
},
{
"count": 77,
"confidence": 0.9715351814332892,
"reference_agreement": 0.922077922077922
}
],
"soft_reliability": [
{
"count": 5,
"confidence": 0.187047500963241,
"reference_agreement": 0.2
},
{
"count": 8,
"confidence": 0.23585635105072375,
"reference_agreement": 0.25
},
{
"count": 19,
"confidence": 0.2982191389855985,
"reference_agreement": 0.2631578947368421
},
{
"count": 16,
"confidence": 0.3728244454729145,
"reference_agreement": 0.5458333125
},
{
"count": 35,
"confidence": 0.43292344830072604,
"reference_agreement": 0.35496548479120027
},
{
"count": 41,
"confidence": 0.5085266662477907,
"reference_agreement": 0.45982781975131765
},
{
"count": 42,
"confidence": 0.5678684243563111,
"reference_agreement": 0.5322624638888933
},
{
"count": 58,
"confidence": 0.6299378167372339,
"reference_agreement": 0.5120080436781644
},
{
"count": 45,
"confidence": 0.7054146769488767,
"reference_agreement": 0.6565114530755077
},
{
"count": 44,
"confidence": 0.7651372249377204,
"reference_agreement": 0.8189414169069195
},
{
"count": 37,
"confidence": 0.833874270098686,
"reference_agreement": 0.8076943700618164
},
{
"count": 65,
"confidence": 0.8966887344775478,
"reference_agreement": 0.8630389169038999
},
{
"count": 85,
"confidence": 0.9707936989016759,
"reference_agreement": 0.9253352246481347
}
],
"expected_score_mae": 0.8450261299297381
},
"legal": {
"count": 122,
"soft_ce": 0.5668266819917146,
"kl": 0.5668266819917146,
"brier": 0.13410281322654918,
"argmax_agreement": 0.819672131147541,
"hard_count": 122,
"hard_ece": 0.07370299961080379,
"hard_reliability": [
{
"count": 1,
"confidence": 0.25539450758217525,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.3631581357093971,
"reference_agreement": 0.0
},
{
"count": 2,
"confidence": 0.4544914341376851,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.503085478379179,
"reference_agreement": 0.3333333333333333
},
{
"count": 5,
"confidence": 0.5760999183430778,
"reference_agreement": 0.4
},
{
"count": 6,
"confidence": 0.6285356936618459,
"reference_agreement": 0.6666666666666666
},
{
"count": 3,
"confidence": 0.6943533513117489,
"reference_agreement": 0.6666666666666666
},
{
"count": 9,
"confidence": 0.7758294858466038,
"reference_agreement": 0.6666666666666666
},
{
"count": 7,
"confidence": 0.8221451030950648,
"reference_agreement": 0.42857142857142855
},
{
"count": 12,
"confidence": 0.9086579271812519,
"reference_agreement": 0.75
},
{
"count": 72,
"confidence": 0.9849048761433361,
"reference_agreement": 0.9861111111111112
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.25539450758217525,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.3631581357093971,
"reference_agreement": 0.0
},
{
"count": 2,
"confidence": 0.4544914341376851,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.503085478379179,
"reference_agreement": 0.3333333333333333
},
{
"count": 5,
"confidence": 0.5760999183430778,
"reference_agreement": 0.4
},
{
"count": 6,
"confidence": 0.6285356936618459,
"reference_agreement": 0.6666666666666666
},
{
"count": 3,
"confidence": 0.6943533513117489,
"reference_agreement": 0.6666666666666666
},
{
"count": 9,
"confidence": 0.7758294858466038,
"reference_agreement": 0.6666666666666666
},
{
"count": 7,
"confidence": 0.8221451030950648,
"reference_agreement": 0.42857142857142855
},
{
"count": 12,
"confidence": 0.9086579271812519,
"reference_agreement": 0.75
},
{
"count": 72,
"confidence": 0.9849048761433361,
"reference_agreement": 0.9861111111111112
}
],
"expected_score_mae": 0.05627741100384848
},
"sales_commerce": {
"count": 71,
"soft_ce": 0.43217932678724524,
"kl": 0.43217932678724524,
"brier": 0.13052027801486862,
"argmax_agreement": 0.8169014084507042,
"hard_count": 71,
"hard_ece": 0.07210090880407366,
"hard_reliability": [
{
"count": 1,
"confidence": 0.43246792381509247,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.5125808974731219,
"reference_agreement": 0.5
},
{
"count": 2,
"confidence": 0.5624424423010923,
"reference_agreement": 0.5
},
{
"count": 6,
"confidence": 0.6350879180799406,
"reference_agreement": 0.5
},
{
"count": 4,
"confidence": 0.6908680440851511,
"reference_agreement": 0.5
},
{
"count": 2,
"confidence": 0.7547308258492205,
"reference_agreement": 0.5
},
{
"count": 11,
"confidence": 0.8418109162556351,
"reference_agreement": 0.7272727272727273
},
{
"count": 16,
"confidence": 0.8995504132439553,
"reference_agreement": 0.875
},
{
"count": 27,
"confidence": 0.9753556391806691,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.43246792381509247,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.5125808974731219,
"reference_agreement": 0.5
},
{
"count": 2,
"confidence": 0.5624424423010923,
"reference_agreement": 0.5
},
{
"count": 6,
"confidence": 0.6350879180799406,
"reference_agreement": 0.5
},
{
"count": 4,
"confidence": 0.6908680440851511,
"reference_agreement": 0.5
},
{
"count": 2,
"confidence": 0.7547308258492205,
"reference_agreement": 0.5
},
{
"count": 11,
"confidence": 0.8418109162556351,
"reference_agreement": 0.7272727272727273
},
{
"count": 16,
"confidence": 0.8995504132439553,
"reference_agreement": 0.875
},
{
"count": 27,
"confidence": 0.9753556391806691,
"reference_agreement": 1.0
}
],
"expected_score_mae": 0.3227189172274408
}
}
},
"calibrated": {
"overall": {
"count": 3693,
"soft_ce": 0.7041168734088186,
"kl": 0.3887052537056387,
"brier": 0.09690042536340486,
"argmax_agreement": 0.7733549959382616,
"hard_count": 2134,
"hard_ece": 0.022770969448581948,
"hard_reliability": [
{
"count": 3,
"confidence": 0.11589315469434829,
"reference_agreement": 0.3333333333333333
},
{
"count": 17,
"confidence": 0.17569618187345318,
"reference_agreement": 0.17647058823529413
},
{
"count": 26,
"confidence": 0.23574174336686576,
"reference_agreement": 0.23076923076923078
},
{
"count": 43,
"confidence": 0.30201610783522903,
"reference_agreement": 0.23255813953488372
},
{
"count": 51,
"confidence": 0.3728587574318181,
"reference_agreement": 0.35294117647058826
},
{
"count": 55,
"confidence": 0.43468960041354515,
"reference_agreement": 0.36363636363636365
},
{
"count": 78,
"confidence": 0.5068967125125762,
"reference_agreement": 0.47435897435897434
},
{
"count": 116,
"confidence": 0.5692509240339776,
"reference_agreement": 0.5689655172413793
},
{
"count": 92,
"confidence": 0.6274691893699693,
"reference_agreement": 0.5760869565217391
},
{
"count": 113,
"confidence": 0.6976255782232683,
"reference_agreement": 0.584070796460177
},
{
"count": 125,
"confidence": 0.7655885050647548,
"reference_agreement": 0.72
},
{
"count": 153,
"confidence": 0.8348620973385892,
"reference_agreement": 0.7777777777777778
},
{
"count": 193,
"confidence": 0.9005488041523733,
"reference_agreement": 0.8860103626943006
},
{
"count": 1069,
"confidence": 0.9873898699630395,
"reference_agreement": 0.9850327408793265
}
],
"soft_reliability": [
{
"count": 3,
"confidence": 0.11589315469434829,
"reference_agreement": 0.3333333333333333
},
{
"count": 23,
"confidence": 0.1712254393277906,
"reference_agreement": 0.22601507061461293
},
{
"count": 65,
"confidence": 0.241401549839449,
"reference_agreement": 0.27101087801087803
},
{
"count": 136,
"confidence": 0.3047942264115045,
"reference_agreement": 0.30327285409683113
},
{
"count": 180,
"confidence": 0.3704360020830201,
"reference_agreement": 0.3727716890584122
},
{
"count": 239,
"confidence": 0.4341484282824121,
"reference_agreement": 0.4200372081322403
},
{
"count": 268,
"confidence": 0.5028656676116453,
"reference_agreement": 0.5155282699000787
},
{
"count": 335,
"confidence": 0.5660902452493901,
"reference_agreement": 0.600639823533436
},
{
"count": 269,
"confidence": 0.631949793983103,
"reference_agreement": 0.6359665011640041
},
{
"count": 261,
"confidence": 0.6984239785753936,
"reference_agreement": 0.6471423835974844
},
{
"count": 249,
"confidence": 0.7657109724234291,
"reference_agreement": 0.7478915785725628
},
{
"count": 253,
"confidence": 0.8344639775274048,
"reference_agreement": 0.8022606606356278
},
{
"count": 261,
"confidence": 0.8995167793570855,
"reference_agreement": 0.8815173885724905
},
{
"count": 1151,
"confidence": 0.9856462873311037,
"reference_agreement": 0.9822842533704442
}
],
"expected_score_mae": 0.4584616360574015
},
"type": {
"choice": {
"count": 2526,
"soft_ce": 0.6766963060796009,
"kl": 0.4133999732939792,
"brier": 0.10042181389472693,
"argmax_agreement": 0.7862232779097387,
"hard_count": 1737,
"hard_ece": 0.024066863233766642,
"hard_reliability": [
{
"count": 3,
"confidence": 0.11589315469434829,
"reference_agreement": 0.3333333333333333
},
{
"count": 4,
"confidence": 0.1466965963178435,
"reference_agreement": 0.25
},
{
"count": 7,
"confidence": 0.23979741242633804,
"reference_agreement": 0.14285714285714285
},
{
"count": 20,
"confidence": 0.30543179732871933,
"reference_agreement": 0.35
},
{
"count": 39,
"confidence": 0.3752333136919975,
"reference_agreement": 0.358974358974359
},
{
"count": 37,
"confidence": 0.43632325318545506,
"reference_agreement": 0.35135135135135137
},
{
"count": 64,
"confidence": 0.5078595631087047,
"reference_agreement": 0.484375
},
{
"count": 91,
"confidence": 0.5663438578130047,
"reference_agreement": 0.6043956043956044
},
{
"count": 72,
"confidence": 0.6259376203313807,
"reference_agreement": 0.5555555555555556
},
{
"count": 87,
"confidence": 0.6985076075501826,
"reference_agreement": 0.6091954022988506
},
{
"count": 98,
"confidence": 0.7672216909611733,
"reference_agreement": 0.7142857142857143
},
{
"count": 126,
"confidence": 0.835107717990711,
"reference_agreement": 0.7777777777777778
},
{
"count": 140,
"confidence": 0.8998423906655577,
"reference_agreement": 0.8857142857142857
},
{
"count": 949,
"confidence": 0.988617468252103,
"reference_agreement": 0.9852476290832455
}
],
"soft_reliability": [
{
"count": 3,
"confidence": 0.11589315469434829,
"reference_agreement": 0.3333333333333333
},
{
"count": 10,
"confidence": 0.15381363979618545,
"reference_agreement": 0.3198346624136098
},
{
"count": 31,
"confidence": 0.2398905686918626,
"reference_agreement": 0.2732486151840991
},
{
"count": 88,
"confidence": 0.3066439829194413,
"reference_agreement": 0.33710350424812785
},
{
"count": 140,
"confidence": 0.37159211827431626,
"reference_agreement": 0.3838011523963886
},
{
"count": 167,
"confidence": 0.4341350320260209,
"reference_agreement": 0.42457801395752337
},
{
"count": 166,
"confidence": 0.5007513721327872,
"reference_agreement": 0.5253523924620044
},
{
"count": 197,
"confidence": 0.565477477433576,
"reference_agreement": 0.619517122506086
},
{
"count": 158,
"confidence": 0.6298800441075835,
"reference_agreement": 0.6306406317730678
},
{
"count": 141,
"confidence": 0.6982422404036815,
"reference_agreement": 0.6697869686413617
},
{
"count": 136,
"confidence": 0.7667611071607225,
"reference_agreement": 0.7399642556669485
},
{
"count": 165,
"confidence": 0.8340530433682095,
"reference_agreement": 0.7888326080745547
},
{
"count": 164,
"confidence": 0.899220153771901,
"reference_agreement": 0.8851424152139722
},
{
"count": 960,
"confidence": 0.988268968121057,
"reference_agreement": 0.9841545132135424
}
]
},
"noul": {
"count": 582,
"soft_ce": 0.4827658039249084,
"kl": 0.16448167168121322,
"brier": 0.05629811292264301,
"argmax_agreement": 0.8316151202749141,
"hard_count": 154,
"hard_ece": 0.03838079417551199,
"hard_reliability": [
{
"count": 2,
"confidence": 0.51409214953238,
"reference_agreement": 0.5
},
{
"count": 12,
"confidence": 0.5848970391247416,
"reference_agreement": 0.5833333333333334
},
{
"count": 10,
"confidence": 0.6345947956878246,
"reference_agreement": 0.6
},
{
"count": 13,
"confidence": 0.6959874383006549,
"reference_agreement": 0.46153846153846156
},
{
"count": 14,
"confidence": 0.7675695442215852,
"reference_agreement": 0.7857142857142857
},
{
"count": 13,
"confidence": 0.8339093290833637,
"reference_agreement": 0.9230769230769231
},
{
"count": 25,
"confidence": 0.9051654333633907,
"reference_agreement": 0.88
},
{
"count": 65,
"confidence": 0.9758087375812551,
"reference_agreement": 0.9692307692307692
}
],
"soft_reliability": [
{
"count": 34,
"confidence": 0.5174138851791591,
"reference_agreement": 0.502892126954275
},
{
"count": 71,
"confidence": 0.5685861678053349,
"reference_agreement": 0.6089361690140845
},
{
"count": 72,
"confidence": 0.6374148978709203,
"reference_agreement": 0.6503378173406389
},
{
"count": 72,
"confidence": 0.6973135426818078,
"reference_agreement": 0.6177620419978036
},
{
"count": 76,
"confidence": 0.7668883954141004,
"reference_agreement": 0.7746528812652387
},
{
"count": 63,
"confidence": 0.8361374834836659,
"reference_agreement": 0.8580618729922205
},
{
"count": 65,
"confidence": 0.899478685155311,
"reference_agreement": 0.8626053588050548
},
{
"count": 129,
"confidence": 0.9701079851807889,
"reference_agreement": 0.9612726817393857
}
]
},
"score": {
"count": 585,
"soft_ce": 1.0427334127476895,
"kl": 0.505148438420284,
"brier": 0.12208934572306966,
"argmax_agreement": 0.6598290598290598,
"hard_count": 243,
"hard_ece": 0.0736454437457963,
"hard_reliability": [
{
"count": 13,
"confidence": 0.18461913127517923,
"reference_agreement": 0.15384615384615385
},
{
"count": 19,
"confidence": 0.23424754950284965,
"reference_agreement": 0.2631578947368421
},
{
"count": 23,
"confidence": 0.299045943058281,
"reference_agreement": 0.13043478260869565
},
{
"count": 12,
"confidence": 0.365141449586235,
"reference_agreement": 0.3333333333333333
},
{
"count": 18,
"confidence": 0.43133153638239685,
"reference_agreement": 0.3888888888888889
},
{
"count": 12,
"confidence": 0.5005622698299229,
"reference_agreement": 0.4166666666666667
},
{
"count": 13,
"confidence": 0.5751578198046974,
"reference_agreement": 0.3076923076923077
},
{
"count": 10,
"confidence": 0.6313708801299527,
"reference_agreement": 0.7
},
{
"count": 13,
"confidence": 0.6933609064965334,
"reference_agreement": 0.5384615384615384
},
{
"count": 13,
"confidence": 0.7511433692151679,
"reference_agreement": 0.6923076923076923
},
{
"count": 14,
"confidence": 0.8335362248493442,
"reference_agreement": 0.6428571428571429
},
{
"count": 28,
"confidence": 0.8999588812194718,
"reference_agreement": 0.8928571428571429
},
{
"count": 55,
"confidence": 0.9798950122993066,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 13,
"confidence": 0.18461913127517923,
"reference_agreement": 0.15384615384615385
},
{
"count": 34,
"confidence": 0.24277920912107187,
"reference_agreement": 0.26897058823529413
},
{
"count": 48,
"confidence": 0.30140300614695376,
"reference_agreement": 0.24124999548612047
},
{
"count": 40,
"confidence": 0.3663895954134836,
"reference_agreement": 0.3341685673754948
},
{
"count": 72,
"confidence": 0.4341795001548752,
"reference_agreement": 0.40950506128748676
},
{
"count": 68,
"confidence": 0.5007529272027481,
"reference_agreement": 0.49786392453063255
},
{
"count": 67,
"confidence": 0.5652470326858571,
"reference_agreement": 0.5363432798507779
},
{
"count": 39,
"confidence": 0.6302455119423146,
"reference_agreement": 0.6310114396011891
},
{
"count": 48,
"confidence": 0.7006234882951765,
"reference_agreement": 0.6246944274306163
},
{
"count": 37,
"confidence": 0.759432527246053,
"reference_agreement": 0.7220606842622975
},
{
"count": 25,
"confidence": 0.8329589079683152,
"reference_agreement": 0.7502667524000977
},
{
"count": 32,
"confidence": 0.9011143643285106,
"reference_agreement": 0.9013541875
},
{
"count": 62,
"confidence": 0.9773667295736088,
"reference_agreement": 0.9970430161290323
}
],
"expected_score_mae": 0.4584616360574015
}
},
"category": {
"engineering": {
"count": 500,
"soft_ce": 0.018681013106648606,
"kl": 0.018681013106648606,
"brier": 0.004596240452886235,
"argmax_agreement": 0.996,
"hard_count": 500,
"hard_ece": 0.010272557015768891,
"hard_reliability": [
{
"count": 1,
"confidence": 0.4716738957978698,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.5539038973137703,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.655163400891637,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.7052527420032781,
"reference_agreement": 0.5
},
{
"count": 4,
"confidence": 0.7571008671839193,
"reference_agreement": 1.0
},
{
"count": 3,
"confidence": 0.8521442972732137,
"reference_agreement": 0.6666666666666666
},
{
"count": 9,
"confidence": 0.9071717975027496,
"reference_agreement": 1.0
},
{
"count": 479,
"confidence": 0.9978224823124803,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.4716738957978698,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.5539038973137703,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.655163400891637,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.7052527420032781,
"reference_agreement": 0.5
},
{
"count": 4,
"confidence": 0.7571008671839193,
"reference_agreement": 1.0
},
{
"count": 3,
"confidence": 0.8521442972732137,
"reference_agreement": 0.6666666666666666
},
{
"count": 9,
"confidence": 0.9071717975027496,
"reference_agreement": 1.0
},
{
"count": 479,
"confidence": 0.9978224823124803,
"reference_agreement": 1.0
}
],
"expected_score_mae": 0.3754904932299786
},
"data_analytics": {
"count": 500,
"soft_ce": 0.6222487807440032,
"kl": 0.6189729294513925,
"brier": 0.1492023277961967,
"argmax_agreement": 0.76,
"hard_count": 497,
"hard_ece": 0.04324263356987873,
"hard_reliability": [
{
"count": 8,
"confidence": 0.1836431426040265,
"reference_agreement": 0.125
},
{
"count": 13,
"confidence": 0.2326095606931366,
"reference_agreement": 0.23076923076923078
},
{
"count": 21,
"confidence": 0.3080856226638128,
"reference_agreement": 0.19047619047619047
},
{
"count": 11,
"confidence": 0.37224859620757544,
"reference_agreement": 0.36363636363636365
},
{
"count": 18,
"confidence": 0.43216064330256254,
"reference_agreement": 0.2777777777777778
},
{
"count": 16,
"confidence": 0.49955516847278475,
"reference_agreement": 0.375
},
{
"count": 20,
"confidence": 0.5648218776742052,
"reference_agreement": 0.5
},
{
"count": 14,
"confidence": 0.6284579234052629,
"reference_agreement": 0.5
},
{
"count": 29,
"confidence": 0.7007100933753049,
"reference_agreement": 0.5517241379310345
},
{
"count": 25,
"confidence": 0.7611781134778828,
"reference_agreement": 0.68
},
{
"count": 42,
"confidence": 0.8391754153900324,
"reference_agreement": 0.7857142857142857
},
{
"count": 45,
"confidence": 0.9013655316115832,
"reference_agreement": 0.8888888888888888
},
{
"count": 235,
"confidence": 0.9854872248925048,
"reference_agreement": 0.9914893617021276
}
],
"soft_reliability": [
{
"count": 8,
"confidence": 0.1836431426040265,
"reference_agreement": 0.125
},
{
"count": 13,
"confidence": 0.2326095606931366,
"reference_agreement": 0.23076923076923078
},
{
"count": 21,
"confidence": 0.3080856226638128,
"reference_agreement": 0.19047619047619047
},
{
"count": 11,
"confidence": 0.37224859620757544,
"reference_agreement": 0.36363636363636365
},
{
"count": 18,
"confidence": 0.43216064330256254,
"reference_agreement": 0.2777777777777778
},
{
"count": 16,
"confidence": 0.49955516847278475,
"reference_agreement": 0.375
},
{
"count": 23,
"confidence": 0.56429092950471,
"reference_agreement": 0.49903378260869563
},
{
"count": 14,
"confidence": 0.6284579234052629,
"reference_agreement": 0.5
},
{
"count": 29,
"confidence": 0.7007100933753049,
"reference_agreement": 0.5517241379310345
},
{
"count": 25,
"confidence": 0.7611781134778828,
"reference_agreement": 0.68
},
{
"count": 42,
"confidence": 0.8391754153900324,
"reference_agreement": 0.7857142857142857
},
{
"count": 45,
"confidence": 0.9013655316115832,
"reference_agreement": 0.8888888888888888
},
{
"count": 235,
"confidence": 0.9854872248925048,
"reference_agreement": 0.9914893617021276
}
],
"expected_score_mae": 0.8394948703521851
},
"finance": {
"count": 500,
"soft_ce": 0.8372534729882183,
"kl": 0.22213311910463876,
"brier": 0.05824555646969622,
"argmax_agreement": 0.766,
"hard_count": 52,
"hard_ece": 0.11310128402928836,
"hard_reliability": [
{
"count": 1,
"confidence": 0.3723071041294397,
"reference_agreement": 0.0
},
{
"count": 5,
"confidence": 0.50016902311118,
"reference_agreement": 0.6
},
{
"count": 2,
"confidence": 0.5753408310146032,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.6182583123935925,
"reference_agreement": 1.0
},
{
"count": 3,
"confidence": 0.6819092006332838,
"reference_agreement": 0.3333333333333333
},
{
"count": 5,
"confidence": 0.7665259528880326,
"reference_agreement": 0.4
},
{
"count": 4,
"confidence": 0.838982843566272,
"reference_agreement": 0.75
},
{
"count": 8,
"confidence": 0.881833000901765,
"reference_agreement": 0.875
},
{
"count": 21,
"confidence": 0.9797645091294379,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 16,
"confidence": 0.3226339567135692,
"reference_agreement": 0.3984374629166975
},
{
"count": 41,
"confidence": 0.36674248561146494,
"reference_agreement": 0.33682928918700217
},
{
"count": 62,
"confidence": 0.4340519346364484,
"reference_agreement": 0.4111594596595161
},
{
"count": 48,
"confidence": 0.49924896849347133,
"reference_agreement": 0.5386290065109846
},
{
"count": 70,
"confidence": 0.5686484601862719,
"reference_agreement": 0.655761905714329
},
{
"count": 50,
"confidence": 0.6308002075172696,
"reference_agreement": 0.6414333730000463
},
{
"count": 54,
"confidence": 0.6999272500653496,
"reference_agreement": 0.6493579925926575
},
{
"count": 46,
"confidence": 0.7648365922391855,
"reference_agreement": 0.7116637471812303
},
{
"count": 32,
"confidence": 0.8315695075811502,
"reference_agreement": 0.8268750565625763
},
{
"count": 22,
"confidence": 0.8916310587032179,
"reference_agreement": 0.9127121363636364
},
{
"count": 59,
"confidence": 0.9675218038967256,
"reference_agreement": 0.9836666610169491
}
],
"expected_score_mae": 0.29112017853517996
},
"trust_safety": {
"count": 500,
"soft_ce": 0.6856993182154376,
"kl": 0.31539041995043376,
"brier": 0.09066598215826954,
"argmax_agreement": 0.734,
"hard_count": 222,
"hard_ece": 0.06329846877237999,
"hard_reliability": [
{
"count": 2,
"confidence": 0.3683034743483865,
"reference_agreement": 0.0
},
{
"count": 7,
"confidence": 0.43137594222756087,
"reference_agreement": 0.2857142857142857
},
{
"count": 9,
"confidence": 0.505779039854673,
"reference_agreement": 0.6666666666666666
},
{
"count": 16,
"confidence": 0.569920504948867,
"reference_agreement": 0.75
},
{
"count": 14,
"confidence": 0.6316407404385947,
"reference_agreement": 0.6428571428571429
},
{
"count": 8,
"confidence": 0.6972827104351492,
"reference_agreement": 0.5
},
{
"count": 18,
"confidence": 0.7672982199470547,
"reference_agreement": 0.6666666666666666
},
{
"count": 19,
"confidence": 0.826724451152989,
"reference_agreement": 0.6842105263157895
},
{
"count": 25,
"confidence": 0.9010283785718083,
"reference_agreement": 0.84
},
{
"count": 104,
"confidence": 0.9729483427736801,
"reference_agreement": 0.9711538461538461
}
],
"soft_reliability": [
{
"count": 9,
"confidence": 0.2361890508436565,
"reference_agreement": 0.27388888888888885
},
{
"count": 7,
"confidence": 0.30342299595800304,
"reference_agreement": 0.2961904285714286
},
{
"count": 18,
"confidence": 0.3674761849264621,
"reference_agreement": 0.3014857858344078
},
{
"count": 33,
"confidence": 0.4359383368744807,
"reference_agreement": 0.3778090549302529
},
{
"count": 35,
"confidence": 0.5047182741304667,
"reference_agreement": 0.517333303803201
},
{
"count": 55,
"confidence": 0.5629326069104805,
"reference_agreement": 0.5935795651642476
},
{
"count": 49,
"confidence": 0.6337639434256919,
"reference_agreement": 0.6102933127493159
},
{
"count": 29,
"confidence": 0.693334781004934,
"reference_agreement": 0.5768357275211178
},
{
"count": 40,
"confidence": 0.7684317409650803,
"reference_agreement": 0.7117894730881641
},
{
"count": 42,
"confidence": 0.8344109799487145,
"reference_agreement": 0.7692592592237804
},
{
"count": 44,
"confidence": 0.8999478793240552,
"reference_agreement": 0.8112662345983765
},
{
"count": 139,
"confidence": 0.9715298963628789,
"reference_agreement": 0.9553096231010606
}
],
"expected_score_mae": 0.3249617918461017
},
"documents_meetings": {
"count": 500,
"soft_ce": 0.9505308054596932,
"kl": 0.2819220023348468,
"brier": 0.06778096624550085,
"argmax_agreement": 0.754,
"hard_count": 111,
"hard_ece": 0.0745149993383762,
"hard_reliability": [
{
"count": 7,
"confidence": 0.5130857536757218,
"reference_agreement": 0.8571428571428571
},
{
"count": 9,
"confidence": 0.5759649564141678,
"reference_agreement": 0.6666666666666666
},
{
"count": 5,
"confidence": 0.631365854228681,
"reference_agreement": 0.8
},
{
"count": 8,
"confidence": 0.7057283821613531,
"reference_agreement": 0.375
},
{
"count": 8,
"confidence": 0.7734135275347481,
"reference_agreement": 0.75
},
{
"count": 14,
"confidence": 0.8276341762810223,
"reference_agreement": 0.7857142857142857
},
{
"count": 11,
"confidence": 0.9061937668671157,
"reference_agreement": 0.9090909090909091
},
{
"count": 49,
"confidence": 0.974518301452977,
"reference_agreement": 0.9591836734693877
}
],
"soft_reliability": [
{
"count": 20,
"confidence": 0.2507030605824935,
"reference_agreement": 0.2815126262626263
},
{
"count": 38,
"confidence": 0.3027780535641847,
"reference_agreement": 0.30187799043062197
},
{
"count": 37,
"confidence": 0.36910116911685614,
"reference_agreement": 0.4304395304395305
},
{
"count": 46,
"confidence": 0.43197305120456336,
"reference_agreement": 0.5305434782608696
},
{
"count": 70,
"confidence": 0.4998801989425534,
"reference_agreement": 0.5903253968253969
},
{
"count": 63,
"confidence": 0.564395576933704,
"reference_agreement": 0.65
},
{
"count": 52,
"confidence": 0.6360817173907356,
"reference_agreement": 0.7249679487179487
},
{
"count": 41,
"confidence": 0.6999872254915769,
"reference_agreement": 0.672560975609756
},
{
"count": 34,
"confidence": 0.769121243337486,
"reference_agreement": 0.8225
},
{
"count": 34,
"confidence": 0.8294780729236917,
"reference_agreement": 0.8451470588235294
},
{
"count": 16,
"confidence": 0.9032976162782315,
"reference_agreement": 0.8896875
},
{
"count": 49,
"confidence": 0.974518301452977,
"reference_agreement": 0.9591836734693877
}
],
"expected_score_mae": 0.39489489482681317
},
"customer_support": {
"count": 500,
"soft_ce": 1.1285135288404535,
"kl": 0.545722865769657,
"brier": 0.108362788973766,
"argmax_agreement": 0.706,
"hard_count": 142,
"hard_ece": 0.11241881164831904,
"hard_reliability": [
{
"count": 3,
"confidence": 0.11589315469434829,
"reference_agreement": 0.3333333333333333
},
{
"count": 4,
"confidence": 0.1466965963178435,
"reference_agreement": 0.25
},
{
"count": 2,
"confidence": 0.23363389271036505,
"reference_agreement": 0.0
},
{
"count": 3,
"confidence": 0.3018620938804334,
"reference_agreement": 0.3333333333333333
},
{
"count": 9,
"confidence": 0.37797570812443015,
"reference_agreement": 0.2222222222222222
},
{
"count": 3,
"confidence": 0.4272157242978048,
"reference_agreement": 0.0
},
{
"count": 7,
"confidence": 0.5005016740966861,
"reference_agreement": 0.14285714285714285
},
{
"count": 7,
"confidence": 0.5675968327842555,
"reference_agreement": 0.8571428571428571
},
{
"count": 7,
"confidence": 0.6248700358768576,
"reference_agreement": 0.5714285714285714
},
{
"count": 9,
"confidence": 0.7018524259315515,
"reference_agreement": 0.5555555555555556
},
{
"count": 18,
"confidence": 0.763652193538776,
"reference_agreement": 0.7222222222222222
},
{
"count": 15,
"confidence": 0.8335790279408464,
"reference_agreement": 0.6666666666666666
},
{
"count": 16,
"confidence": 0.9023386004425975,
"reference_agreement": 0.8125
},
{
"count": 39,
"confidence": 0.9678176556415846,
"reference_agreement": 0.9487179487179487
}
],
"soft_reliability": [
{
"count": 3,
"confidence": 0.11589315469434829,
"reference_agreement": 0.3333333333333333
},
{
"count": 10,
"confidence": 0.15381363979618545,
"reference_agreement": 0.3198346624136098
},
{
"count": 12,
"confidence": 0.24077671068363551,
"reference_agreement": 0.29337121212121214
},
{
"count": 35,
"confidence": 0.3022639463465406,
"reference_agreement": 0.3521546318325211
},
{
"count": 42,
"confidence": 0.37365140372838623,
"reference_agreement": 0.35323562863250324
},
{
"count": 45,
"confidence": 0.4350034398300646,
"reference_agreement": 0.36927072142944645
},
{
"count": 55,
"confidence": 0.5024000185598814,
"reference_agreement": 0.4825757636363636
},
{
"count": 55,
"confidence": 0.5652814585293939,
"reference_agreement": 0.631175707569653
},
{
"count": 47,
"confidence": 0.6319760416224046,
"reference_agreement": 0.6904964592198635
},
{
"count": 40,
"confidence": 0.7024876007617367,
"reference_agreement": 0.669458325
},
{
"count": 43,
"confidence": 0.766737205353825,
"reference_agreement": 0.7386433953488373
},
{
"count": 33,
"confidence": 0.83741488958767,
"reference_agreement": 0.7505555454545454
},
{
"count": 37,
"confidence": 0.8973418194210278,
"reference_agreement": 0.8589099459459459
},
{
"count": 43,
"confidence": 0.965788577174672,
"reference_agreement": 0.9501550236821856
}
],
"expected_score_mae": 0.33319003384682117
},
"ai_agents_evals": {
"count": 500,
"soft_ce": 0.7625932921761905,
"kl": 0.6730676386853419,
"brier": 0.18645192181998924,
"argmax_agreement": 0.68,
"hard_count": 417,
"hard_ece": 0.052463098512104685,
"hard_reliability": [
{
"count": 5,
"confidence": 0.18618071314902357,
"reference_agreement": 0.2
},
{
"count": 10,
"confidence": 0.23838633417547514,
"reference_agreement": 0.2
},
{
"count": 18,
"confidence": 0.295781111577869,
"reference_agreement": 0.2777777777777778
},
{
"count": 26,
"confidence": 0.37139215520169777,
"reference_agreement": 0.4230769230769231
},
{
"count": 22,
"confidence": 0.43633042832674923,
"reference_agreement": 0.45454545454545453
},
{
"count": 29,
"confidence": 0.5119118632329336,
"reference_agreement": 0.4482758620689655
},
{
"count": 51,
"confidence": 0.5699148465657553,
"reference_agreement": 0.47058823529411764
},
{
"count": 38,
"confidence": 0.6258474222785992,
"reference_agreement": 0.5263157894736842
},
{
"count": 42,
"confidence": 0.692690594223072,
"reference_agreement": 0.7380952380952381
},
{
"count": 35,
"confidence": 0.7681140161954323,
"reference_agreement": 0.7714285714285715
},
{
"count": 40,
"confidence": 0.8344200795661016,
"reference_agreement": 0.85
},
{
"count": 46,
"confidence": 0.9006389466100536,
"reference_agreement": 0.9347826086956522
},
{
"count": 55,
"confidence": 0.9717353590779876,
"reference_agreement": 0.8909090909090909
}
],
"soft_reliability": [
{
"count": 5,
"confidence": 0.18618071314902357,
"reference_agreement": 0.2
},
{
"count": 10,
"confidence": 0.23838633417547514,
"reference_agreement": 0.2
},
{
"count": 18,
"confidence": 0.295781111577869,
"reference_agreement": 0.2777777777777778
},
{
"count": 29,
"confidence": 0.3733951639465118,
"reference_agreement": 0.41724137931034483
},
{
"count": 30,
"confidence": 0.4336998105448617,
"reference_agreement": 0.44690416558973367
},
{
"count": 39,
"confidence": 0.5120708414698647,
"reference_agreement": 0.4701608617898467
},
{
"count": 58,
"confidence": 0.5690122989604566,
"reference_agreement": 0.4917532841954055
},
{
"count": 46,
"confidence": 0.6283615036372431,
"reference_agreement": 0.5640713243854648
},
{
"count": 54,
"confidence": 0.6932351995199476,
"reference_agreement": 0.7547641003856951
},
{
"count": 45,
"confidence": 0.7641095557632885,
"reference_agreement": 0.7842272393489996
},
{
"count": 51,
"confidence": 0.8318454353186426,
"reference_agreement": 0.8372494792826007
},
{
"count": 55,
"confidence": 0.9002777292404353,
"reference_agreement": 0.9324725290016627
},
{
"count": 60,
"confidence": 0.9709295090707568,
"reference_agreement": 0.8988023166666667
}
],
"expected_score_mae": 0.8456565717140646
},
"legal": {
"count": 122,
"soft_ce": 0.5436294554804777,
"kl": 0.5436294554804777,
"brier": 0.1304449623060641,
"argmax_agreement": 0.819672131147541,
"hard_count": 122,
"hard_ece": 0.08307026110171992,
"hard_reliability": [
{
"count": 1,
"confidence": 0.2542299113522521,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.28724827093183936,
"reference_agreement": 0.0
},
{
"count": 2,
"confidence": 0.377085304774584,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.4441052741464433,
"reference_agreement": 0.3333333333333333
},
{
"count": 2,
"confidence": 0.5174342366971163,
"reference_agreement": 0.0
},
{
"count": 7,
"confidence": 0.5644377124854338,
"reference_agreement": 0.5714285714285714
},
{
"count": 5,
"confidence": 0.6251477859811808,
"reference_agreement": 0.8
},
{
"count": 8,
"confidence": 0.7044969749942575,
"reference_agreement": 0.25
},
{
"count": 8,
"confidence": 0.7587080210898536,
"reference_agreement": 0.75
},
{
"count": 6,
"confidence": 0.8354103639312468,
"reference_agreement": 1.0
},
{
"count": 13,
"confidence": 0.8963652039349991,
"reference_agreement": 0.7692307692307693
},
{
"count": 66,
"confidence": 0.9815239804604637,
"reference_agreement": 0.9848484848484849
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.2542299113522521,
"reference_agreement": 1.0
},
{
"count": 1,
"confidence": 0.28724827093183936,
"reference_agreement": 0.0
},
{
"count": 2,
"confidence": 0.377085304774584,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.4441052741464433,
"reference_agreement": 0.3333333333333333
},
{
"count": 2,
"confidence": 0.5174342366971163,
"reference_agreement": 0.0
},
{
"count": 7,
"confidence": 0.5644377124854338,
"reference_agreement": 0.5714285714285714
},
{
"count": 5,
"confidence": 0.6251477859811808,
"reference_agreement": 0.8
},
{
"count": 8,
"confidence": 0.7044969749942575,
"reference_agreement": 0.25
},
{
"count": 8,
"confidence": 0.7587080210898536,
"reference_agreement": 0.75
},
{
"count": 6,
"confidence": 0.8354103639312468,
"reference_agreement": 1.0
},
{
"count": 13,
"confidence": 0.8963652039349991,
"reference_agreement": 0.7692307692307693
},
{
"count": 66,
"confidence": 0.9815239804604637,
"reference_agreement": 0.9848484848484849
}
],
"expected_score_mae": 0.05800823336689319
},
"sales_commerce": {
"count": 71,
"soft_ce": 0.43972836851867453,
"kl": 0.43972836851867453,
"brier": 0.13079004940228145,
"argmax_agreement": 0.8169014084507042,
"hard_count": 71,
"hard_ece": 0.08143885493164209,
"hard_reliability": [
{
"count": 2,
"confidence": 0.4480862145923521,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.5225530044071978,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.5799241321545268,
"reference_agreement": 0.6666666666666666
},
{
"count": 5,
"confidence": 0.6273970582753347,
"reference_agreement": 0.2
},
{
"count": 4,
"confidence": 0.6962808034900051,
"reference_agreement": 0.75
},
{
"count": 4,
"confidence": 0.7775016660973879,
"reference_agreement": 0.75
},
{
"count": 10,
"confidence": 0.8388575355152714,
"reference_agreement": 0.7
},
{
"count": 20,
"confidence": 0.9005931196007729,
"reference_agreement": 0.9
},
{
"count": 21,
"confidence": 0.9756813796187571,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 2,
"confidence": 0.4480862145923521,
"reference_agreement": 1.0
},
{
"count": 2,
"confidence": 0.5225530044071978,
"reference_agreement": 0.5
},
{
"count": 3,
"confidence": 0.5799241321545268,
"reference_agreement": 0.6666666666666666
},
{
"count": 5,
"confidence": 0.6273970582753347,
"reference_agreement": 0.2
},
{
"count": 4,
"confidence": 0.6962808034900051,
"reference_agreement": 0.75
},
{
"count": 4,
"confidence": 0.7775016660973879,
"reference_agreement": 0.75
},
{
"count": 10,
"confidence": 0.8388575355152714,
"reference_agreement": 0.7
},
{
"count": 20,
"confidence": 0.9005931196007729,
"reference_agreement": 0.9
},
{
"count": 21,
"confidence": 0.9756813796187571,
"reference_agreement": 1.0
}
],
"expected_score_mae": 0.3234410160549905
}
}
},
"permutation": {
"count": 256,
"completed": true,
"mean_total_variation": 0.061152504018688186,
"answer_flip_rate": 0.0625
}
},
"typed_decisions": {
"dataset": "LocalLLaMA/typed-decisions",
"revision": "7189f46b02248bd2569df7325c9df8e3b24e90a7",
"config": "all",
"split": "test",
"cases": 400,
"decisions": 2000,
"errors": 0,
"raw": {
"overall": {
"count": 2000,
"soft_ce": 0.8982348703026427,
"kl": 0.13087631467831382,
"brier": 0.03733508375955404,
"argmax_agreement": 0.7055,
"hard_count": 2,
"hard_ece": 0.02813636220158644,
"hard_reliability": [
{
"count": 2,
"confidence": 0.9718636377984136,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 24,
"confidence": 0.31520772849577633,
"reference_agreement": 0.3355554526389121
},
{
"count": 171,
"confidence": 0.3746159687360105,
"reference_agreement": 0.38106237192013953
},
{
"count": 368,
"confidence": 0.43486926682677546,
"reference_agreement": 0.41427729467341934
},
{
"count": 330,
"confidence": 0.5005758970577506,
"reference_agreement": 0.4898888885960213
},
{
"count": 259,
"confidence": 0.5661637493278858,
"reference_agreement": 0.5401029315058599
},
{
"count": 208,
"confidence": 0.6334207148716826,
"reference_agreement": 0.5813301291346679
},
{
"count": 176,
"confidence": 0.6976026942523854,
"reference_agreement": 0.6650378634451258
},
{
"count": 117,
"confidence": 0.765488790351476,
"reference_agreement": 0.6715926028234322
},
{
"count": 95,
"confidence": 0.8268994900267613,
"reference_agreement": 0.7595845223159148
},
{
"count": 90,
"confidence": 0.8962419341009388,
"reference_agreement": 0.9119985310000607
},
{
"count": 162,
"confidence": 0.9675863551773324,
"reference_agreement": 0.9661111104013173
}
],
"expected_score_mae": 0.2689357269785638,
"gold_label_correct": 1409,
"gold_label_accuracy": 0.7045,
"gold_label_ece_15bin_all_decisions": 0.10329466050980009,
"full_vector_brier": 0.07467016751910809
},
"type": {
"choice": {
"count": 600,
"soft_ce": 1.0297958099292934,
"kl": 0.1660012130824713,
"brier": 0.041920794479682215,
"argmax_agreement": 0.6883333333333334,
"hard_count": 1,
"hard_ece": 0.023186746014468573,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9768132539855314,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 14,
"confidence": 0.3173165797051884,
"reference_agreement": 0.39047612547625926
},
{
"count": 114,
"confidence": 0.3727881473417713,
"reference_agreement": 0.3832163580409846
},
{
"count": 160,
"confidence": 0.4341875923805457,
"reference_agreement": 0.4184011179583752
},
{
"count": 92,
"confidence": 0.4936694315907819,
"reference_agreement": 0.511159424456586
},
{
"count": 48,
"confidence": 0.5673671375615373,
"reference_agreement": 0.5801388408334563
},
{
"count": 31,
"confidence": 0.6339963551108706,
"reference_agreement": 0.6433333275269537
},
{
"count": 28,
"confidence": 0.6907448914335091,
"reference_agreement": 0.7217737262976213
},
{
"count": 22,
"confidence": 0.764159840829608,
"reference_agreement": 0.6959089121213682
},
{
"count": 32,
"confidence": 0.8290876216067908,
"reference_agreement": 0.7640788558334829
},
{
"count": 33,
"confidence": 0.8944038265429061,
"reference_agreement": 0.9230363914142908
},
{
"count": 26,
"confidence": 0.9583538062949253,
"reference_agreement": 0.955179537115826
}
],
"gold_label_correct": 413,
"gold_label_accuracy": 0.6883333333333334,
"gold_label_ece_15bin_all_decisions": 0.1452815073319075,
"full_vector_brier": 0.08384158895936443
},
"noul": {
"count": 600,
"soft_ce": 0.5376753319725679,
"kl": 0.08395308862000342,
"brier": 0.034802117110535095,
"argmax_agreement": 0.7666666666666667,
"hard_count": 1,
"hard_ece": 0.033085978388704196,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9669140216112958,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 59,
"confidence": 0.5179912574843223,
"reference_agreement": 0.48372877966101696
},
{
"count": 86,
"confidence": 0.5678122996718532,
"reference_agreement": 0.525968976744186
},
{
"count": 85,
"confidence": 0.6369483441744728,
"reference_agreement": 0.5698431176470588
},
{
"count": 85,
"confidence": 0.6995096580475858,
"reference_agreement": 0.6567059176470588
},
{
"count": 63,
"confidence": 0.7673512554565463,
"reference_agreement": 0.6198413174603175
},
{
"count": 47,
"confidence": 0.8295192389350612,
"reference_agreement": 0.740141914893617
},
{
"count": 54,
"confidence": 0.8971869226053684,
"reference_agreement": 0.9035740740740741
},
{
"count": 121,
"confidence": 0.9707946773144384,
"reference_agreement": 0.9677850826446281
}
],
"gold_label_correct": 459,
"gold_label_accuracy": 0.765,
"gold_label_ece_15bin_all_decisions": 0.037528508953695645,
"full_vector_brier": 0.06960423422107019
},
"score": {
"count": 800,
"soft_ce": 1.0699838193302107,
"kl": 0.1397250604189285,
"brier": 0.03579552570622214,
"argmax_agreement": 0.6725,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 10,
"confidence": 0.3122553368025993,
"reference_agreement": 0.25866651066662605
},
{
"count": 57,
"confidence": 0.3782716115244891,
"reference_agreement": 0.3767543996784493
},
{
"count": 208,
"confidence": 0.4353936317854138,
"reference_agreement": 0.41110512291576096
},
{
"count": 179,
"confidence": 0.4983853303415126,
"reference_agreement": 0.4809869731099504
},
{
"count": 125,
"confidence": 0.564567445609514,
"reference_agreement": 0.5344533032000944
},
{
"count": 92,
"confidence": 0.6299675264135088,
"reference_agreement": 0.5710507468116885
},
{
"count": 63,
"confidence": 0.6980777189561393,
"reference_agreement": 0.6510634385715672
},
{
"count": 32,
"confidence": 0.7627357149721534,
"reference_agreement": 0.7567604832397332
},
{
"count": 16,
"confidence": 0.8148277144485708,
"reference_agreement": 0.8077085145837775
},
{
"count": 3,
"confidence": 0.8994513241595642,
"reference_agreement": 0.9422222911112911
},
{
"count": 15,
"confidence": 0.9577089746675174,
"reference_agreement": 0.9715557946667946
}
],
"expected_score_mae": 0.2689357269785638,
"gold_label_correct": 537,
"gold_label_accuracy": 0.67125,
"gold_label_ece_15bin_all_decisions": 0.13471575955048368,
"full_vector_brier": 0.07159105141244428
}
},
"category": {
"agent_trace_observability": {
"count": 500,
"soft_ce": 0.97776378147771,
"kl": 0.09929350853248653,
"brier": 0.029319985421551766,
"argmax_agreement": 0.666,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.3271961322725437,
"reference_agreement": 0.333333
},
{
"count": 35,
"confidence": 0.3798205422614706,
"reference_agreement": 0.40077146666672475
},
{
"count": 132,
"confidence": 0.43641602107268646,
"reference_agreement": 0.4047222168687381
},
{
"count": 117,
"confidence": 0.5006302500044424,
"reference_agreement": 0.47911679880348107
},
{
"count": 76,
"confidence": 0.5690619634091041,
"reference_agreement": 0.5512718920614565
},
{
"count": 45,
"confidence": 0.6338679776529896,
"reference_agreement": 0.6111852223703935
},
{
"count": 25,
"confidence": 0.6938113959359317,
"reference_agreement": 0.6235998789334686
},
{
"count": 14,
"confidence": 0.7711383973508009,
"reference_agreement": 0.6864285047620653
},
{
"count": 32,
"confidence": 0.823789261655095,
"reference_agreement": 0.7322917248959386
},
{
"count": 23,
"confidence": 0.8940903067075735,
"reference_agreement": 0.8727536086956522
}
],
"expected_score_mae": 0.20094214035170185,
"gold_label_correct": 335,
"gold_label_accuracy": 0.67,
"gold_label_ece_15bin_all_decisions": 0.11802699803796707,
"full_vector_brier": 0.05863997084310353
},
"customer_service": {
"count": 500,
"soft_ce": 0.9356140291801783,
"kl": 0.19845939117069544,
"brier": 0.054488558914342955,
"argmax_agreement": 0.694,
"hard_count": 1,
"hard_ece": 0.023186746014468573,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9768132539855314,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 10,
"confidence": 0.31477128374516633,
"reference_agreement": 0.4056665940000813
},
{
"count": 39,
"confidence": 0.3723859508926986,
"reference_agreement": 0.45632472794882123
},
{
"count": 73,
"confidence": 0.4380748853548461,
"reference_agreement": 0.3939339658448251
},
{
"count": 81,
"confidence": 0.5011240542197034,
"reference_agreement": 0.5185185261729083
},
{
"count": 80,
"confidence": 0.5649588987676203,
"reference_agreement": 0.5116249845417362
},
{
"count": 69,
"confidence": 0.6279948891492104,
"reference_agreement": 0.5734783022222739
},
{
"count": 46,
"confidence": 0.6947774859534929,
"reference_agreement": 0.7216593691232294
},
{
"count": 33,
"confidence": 0.7624282639003352,
"reference_agreement": 0.6597979875758164
},
{
"count": 22,
"confidence": 0.8310417874707298,
"reference_agreement": 0.8232055630304752
},
{
"count": 21,
"confidence": 0.8956622165551549,
"reference_agreement": 0.913333472222457
},
{
"count": 26,
"confidence": 0.9583538062949253,
"reference_agreement": 0.955179537115826
}
],
"expected_score_mae": 0.3528748230925414,
"gold_label_correct": 347,
"gold_label_accuracy": 0.694,
"gold_label_ece_15bin_all_decisions": 0.09884340864299247,
"full_vector_brier": 0.10897711782868591
},
"invoice_processing": {
"count": 500,
"soft_ce": 0.6654727615892404,
"kl": 0.09052528592434582,
"brier": 0.02417431701033401,
"argmax_agreement": 0.84,
"hard_count": 1,
"hard_ece": 0.033085978388704196,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9669140216112958,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 2,
"confidence": 0.3278949544267834,
"reference_agreement": 0.5716665
},
{
"count": 37,
"confidence": 0.36518717835723197,
"reference_agreement": 0.3640449954955256
},
{
"count": 44,
"confidence": 0.43272176571282106,
"reference_agreement": 0.4239439440863979
},
{
"count": 47,
"confidence": 0.4990750370635605,
"reference_agreement": 0.4812765950355347
},
{
"count": 40,
"confidence": 0.5636676327793726,
"reference_agreement": 0.5589167148334007
},
{
"count": 44,
"confidence": 0.6359292707198466,
"reference_agreement": 0.618181796969778
},
{
"count": 53,
"confidence": 0.6993117889365471,
"reference_agreement": 0.6881194986793073
},
{
"count": 34,
"confidence": 0.7613192737634586,
"reference_agreement": 0.7343039374609032
},
{
"count": 21,
"confidence": 0.8299096395763272,
"reference_agreement": 0.8130160017462573
},
{
"count": 43,
"confidence": 0.8979274834748605,
"reference_agreement": 0.9520123691472993
},
{
"count": 135,
"confidence": 0.9695963375504574,
"reference_agreement": 0.9691999920000142
}
],
"expected_score_mae": 0.21560838991268416,
"gold_label_correct": 417,
"gold_label_accuracy": 0.834,
"gold_label_ece_15bin_all_decisions": 0.11985107384860202,
"full_vector_brier": 0.04834863402066802
},
"security_incidents": {
"count": 500,
"soft_ce": 1.0140889089634417,
"kl": 0.13522707308572743,
"brier": 0.041357473691987466,
"argmax_agreement": 0.622,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 11,
"confidence": 0.31220787321098703,
"reference_agreement": 0.2290908112120979
},
{
"count": 60,
"confidence": 0.37884389984455846,
"reference_agreement": 0.33113891736116685
},
{
"count": 119,
"confidence": 0.4319811015831166,
"reference_agreement": 0.43378150224093404
},
{
"count": 85,
"confidence": 0.5008086075853485,
"reference_agreement": 0.48219608458828284
},
{
"count": 63,
"confidence": 0.5657822959404127,
"reference_agreement": 0.550846477883684
},
{
"count": 50,
"confidence": 0.6382982887191337,
"reference_agreement": 0.5328665986667214
},
{
"count": 52,
"confidence": 0.7001826946639205,
"reference_agreement": 0.6113462035256458
},
{
"count": 36,
"confidence": 0.7700350803206342,
"reference_agreement": 0.6174074444444444
},
{
"count": 20,
"confidence": 0.8241586712060176,
"reference_agreement": 0.6771668
},
{
"count": 3,
"confidence": 0.8926362259110124,
"reference_agreement": 0.63
},
{
"count": 1,
"confidence": 0.9362850057480346,
"reference_agreement": 0.833333
}
],
"expected_score_mae": 0.3063175545573279,
"gold_label_correct": 310,
"gold_label_accuracy": 0.62,
"gold_label_ece_15bin_all_decisions": 0.11599525503936776,
"full_vector_brier": 0.08271494738397493
}
},
"question": {
"agent_trace_observability/action": {
"count": 100,
"soft_ce": 1.1235396454195041,
"kl": 0.08949577063715862,
"brier": 0.023420285987952504,
"argmax_agreement": 0.66,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.3271961322725437,
"reference_agreement": 0.333333
},
{
"count": 28,
"confidence": 0.37758519376271876,
"reference_agreement": 0.4010834071429083
},
{
"count": 34,
"confidence": 0.42300368726206095,
"reference_agreement": 0.40872541411767294
},
{
"count": 14,
"confidence": 0.508725847017595,
"reference_agreement": 0.6019047511905369
},
{
"count": 23,
"confidence": 0.5700440548082534,
"reference_agreement": 0.6007246056522724
}
],
"gold_label_correct": 66,
"gold_label_accuracy": 0.66,
"gold_label_ece_15bin_all_decisions": 0.21139510221170185,
"full_vector_brier": 0.04684057197590501
},
"agent_trace_observability/needs_review": {
"count": 100,
"soft_ce": 0.5772198756410069,
"kl": 0.06175400802011981,
"brier": 0.026948573243994925,
"argmax_agreement": 0.78,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 17,
"confidence": 0.522037312516128,
"reference_agreement": 0.48039217647058824
},
{
"count": 27,
"confidence": 0.5740099433811946,
"reference_agreement": 0.5532098148148148
},
{
"count": 19,
"confidence": 0.6307027904464286,
"reference_agreement": 0.6412281052631579
},
{
"count": 1,
"confidence": 0.679178699175393,
"reference_agreement": 0.766667
},
{
"count": 13,
"confidence": 0.831497587641693,
"reference_agreement": 0.8541025384615385
},
{
"count": 23,
"confidence": 0.8940903067075735,
"reference_agreement": 0.8727536086956522
}
],
"gold_label_correct": 78,
"gold_label_accuracy": 0.78,
"gold_label_ece_15bin_all_decisions": 0.09591019804659835,
"full_vector_brier": 0.05389714648798985
},
"agent_trace_observability/outcome": {
"count": 100,
"soft_ce": 1.0935064791897204,
"kl": 0.16872011458553096,
"brier": 0.04551726407182359,
"argmax_agreement": 0.62,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.39948630465030277,
"reference_agreement": 0.233333
},
{
"count": 36,
"confidence": 0.44371195558609283,
"reference_agreement": 0.36223152500002964
},
{
"count": 29,
"confidence": 0.4869204497761252,
"reference_agreement": 0.42333332505753857
},
{
"count": 1,
"confidence": 0.5457545962932366,
"reference_agreement": 0.75
},
{
"count": 4,
"confidence": 0.6334661880424128,
"reference_agreement": 0.8091667499999999
},
{
"count": 6,
"confidence": 0.68759444755407,
"reference_agreement": 0.6505555205556461
},
{
"count": 8,
"confidence": 0.7680981282943802,
"reference_agreement": 0.6516665095834904
},
{
"count": 15,
"confidence": 0.820754732086627,
"reference_agreement": 0.6195556666667333
}
],
"gold_label_correct": 60,
"gold_label_accuracy": 0.6,
"gold_label_ece_15bin_all_decisions": 0.08832517707815522,
"full_vector_brier": 0.09103452814364718
},
"agent_trace_observability/risk": {
"count": 100,
"soft_ce": 0.9829069466566531,
"kl": 0.09053683824790124,
"brier": 0.027227209135250532,
"argmax_agreement": 0.67,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 6,
"confidence": 0.3869745415241738,
"reference_agreement": 0.4272221555556555
},
{
"count": 15,
"confidence": 0.4430243807991401,
"reference_agreement": 0.4377777200001155
},
{
"count": 20,
"confidence": 0.5064388102385512,
"reference_agreement": 0.552666578333445
},
{
"count": 22,
"confidence": 0.5633028160410002,
"reference_agreement": 0.5042424242424954
},
{
"count": 15,
"confidence": 0.6350598264314197,
"reference_agreement": 0.5333333022222356
},
{
"count": 13,
"confidence": 0.6937211677482574,
"reference_agreement": 0.5712818941026444
},
{
"count": 5,
"confidence": 0.780801733899054,
"reference_agreement": 0.7526666713334713
},
{
"count": 4,
"confidence": 0.810116688080407,
"reference_agreement": 0.7591667991672592
}
],
"expected_score_mae": 0.1887612567569762,
"gold_label_correct": 69,
"gold_label_accuracy": 0.69,
"gold_label_ece_15bin_all_decisions": 0.14607924787983295,
"full_vector_brier": 0.054454418270501063
},
"agent_trace_observability/urgency": {
"count": 100,
"soft_ce": 1.111645960481666,
"kl": 0.08596081117172201,
"brier": 0.023486594668737287,
"argmax_agreement": 0.6,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 47,
"confidence": 0.43842113407655536,
"reference_agreement": 0.42382272014190947
},
{
"count": 37,
"confidence": 0.49533713057352974,
"reference_agreement": 0.4360360527027635
},
{
"count": 3,
"confidence": 0.5670036460048613,
"reference_agreement": 0.43333317777784447
},
{
"count": 7,
"confidence": 0.6401348324659201,
"reference_agreement": 0.5833334961905962
},
{
"count": 5,
"confidence": 0.7044328666342266,
"reference_agreement": 0.6986664453336919
},
{
"count": 1,
"confidence": 0.7471438670609017,
"reference_agreement": 0.6333336333336334
}
],
"expected_score_mae": 0.21312302394642757,
"gold_label_correct": 62,
"gold_label_accuracy": 0.62,
"gold_label_ece_15bin_all_decisions": 0.12615469901673237,
"full_vector_brier": 0.046973189337474575
},
"customer_service/action": {
"count": 100,
"soft_ce": 1.241760340801668,
"kl": 0.275902444840478,
"brier": 0.06755224071529659,
"argmax_agreement": 0.6,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 9,
"confidence": 0.31273256683149947,
"reference_agreement": 0.3525925488889793
},
{
"count": 34,
"confidence": 0.37221056069924124,
"reference_agreement": 0.4573528766667779
},
{
"count": 25,
"confidence": 0.4324976531209652,
"reference_agreement": 0.4009538457334028
},
{
"count": 20,
"confidence": 0.49707265548386304,
"reference_agreement": 0.5388333921667305
},
{
"count": 9,
"confidence": 0.5563524243492388,
"reference_agreement": 0.6407407485186634
},
{
"count": 1,
"confidence": 0.6089655400408842,
"reference_agreement": 0.35
},
{
"count": 2,
"confidence": 0.705852717684706,
"reference_agreement": 0.7800003916668916
}
],
"gold_label_correct": 61,
"gold_label_accuracy": 0.61,
"gold_label_ece_15bin_all_decisions": 0.18966441682569232,
"full_vector_brier": 0.13510448143059317
},
"customer_service/category": {
"count": 100,
"soft_ce": 0.5646570684297757,
"kl": 0.15423518549593282,
"brier": 0.03216723408432829,
"argmax_agreement": 0.93,
"hard_count": 1,
"hard_ece": 0.023186746014468573,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9768132539855314,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 3,
"confidence": 0.4430349798534358,
"reference_agreement": 0.6166666055560389
},
{
"count": 5,
"confidence": 0.5046243895746633,
"reference_agreement": 0.635333516000116
},
{
"count": 4,
"confidence": 0.5798365858143383,
"reference_agreement": 0.5683333691670058
},
{
"count": 8,
"confidence": 0.6384560218456447,
"reference_agreement": 0.6099999795834395
},
{
"count": 9,
"confidence": 0.6915068535481631,
"reference_agreement": 0.8281108069998411
},
{
"count": 9,
"confidence": 0.7596657412667888,
"reference_agreement": 0.6937035451853623
},
{
"count": 15,
"confidence": 0.8332816589849125,
"reference_agreement": 0.8847014257780302
},
{
"count": 21,
"confidence": 0.8956622165551549,
"reference_agreement": 0.913333472222457
},
{
"count": 26,
"confidence": 0.9583538062949253,
"reference_agreement": 0.955179537115826
}
],
"gold_label_correct": 94,
"gold_label_accuracy": 0.94,
"gold_label_ece_15bin_all_decisions": 0.1407358753162401,
"full_vector_brier": 0.06433446816865658
},
"customer_service/churn_risk": {
"count": 100,
"soft_ce": 1.0723716037197997,
"kl": 0.18195307128399665,
"brier": 0.05156803196362559,
"argmax_agreement": 0.71,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.3331197359681683,
"reference_agreement": 0.883333
},
{
"count": 5,
"confidence": 0.3735786042082082,
"reference_agreement": 0.4493333166667167
},
{
"count": 31,
"confidence": 0.4406830914886085,
"reference_agreement": 0.388172083225866
},
{
"count": 26,
"confidence": 0.5011742997082412,
"reference_agreement": 0.578589826794943
},
{
"count": 18,
"confidence": 0.5633032673902467,
"reference_agreement": 0.5179629694445509
},
{
"count": 17,
"confidence": 0.6284474784336418,
"reference_agreement": 0.6143137843137941
},
{
"count": 2,
"confidence": 0.6870135499018379,
"reference_agreement": 0.6716668416668417
}
],
"expected_score_mae": 0.3151828379933619,
"gold_label_correct": 71,
"gold_label_accuracy": 0.71,
"gold_label_ece_15bin_all_decisions": 0.19910186568229626,
"full_vector_brier": 0.10313606392725118
},
"customer_service/needs_human": {
"count": 100,
"soft_ce": 0.6761560528212277,
"kl": 0.12704364399222592,
"brier": 0.05613679930099221,
"argmax_agreement": 0.66,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 8,
"confidence": 0.517557176152255,
"reference_agreement": 0.482499875
},
{
"count": 22,
"confidence": 0.5656107740630231,
"reference_agreement": 0.5101515
},
{
"count": 20,
"confidence": 0.6310079557436046,
"reference_agreement": 0.5293334
},
{
"count": 20,
"confidence": 0.7003291402466889,
"reference_agreement": 0.7098333
},
{
"count": 23,
"confidence": 0.7643885662708242,
"reference_agreement": 0.6455073043478261
},
{
"count": 7,
"confidence": 0.8262420627974808,
"reference_agreement": 0.6914287142857143
}
],
"gold_label_correct": 65,
"gold_label_accuracy": 0.65,
"gold_label_ece_15bin_all_decisions": 0.08754395103445267,
"full_vector_brier": 0.11227359860198442
},
"customer_service/urgency": {
"count": 100,
"soft_ce": 1.1231250801284207,
"kl": 0.25316261024084386,
"brier": 0.06501848850747206,
"argmax_agreement": 0.57,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 14,
"confidence": 0.4411960376551762,
"reference_agreement": 0.34642849761908573
},
{
"count": 22,
"confidence": 0.4979765514824134,
"reference_agreement": 0.41560594090917574
},
{
"count": 27,
"confidence": 0.5661962183925248,
"reference_agreement": 0.45716044814818446
},
{
"count": 23,
"confidence": 0.6222290168802378,
"reference_agreement": 0.5786956818841255
},
{
"count": 13,
"confidence": 0.6879913332169492,
"reference_agreement": 0.664871788461732
},
{
"count": 1,
"confidence": 0.7422040130810093,
"reference_agreement": 0.6833336833336833
}
],
"expected_score_mae": 0.39056680819172096,
"gold_label_correct": 56,
"gold_label_accuracy": 0.56,
"gold_label_ece_15bin_all_decisions": 0.12422167823236911,
"full_vector_brier": 0.13003697701494413
},
"invoice_processing/discrepancy_severity": {
"count": 100,
"soft_ce": 0.7813166885573835,
"kl": 0.13707332422163884,
"brier": 0.03045788400272809,
"argmax_agreement": 0.83,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 10,
"confidence": 0.36318909440456576,
"reference_agreement": 0.452666601666755
},
{
"count": 15,
"confidence": 0.41585765000740066,
"reference_agreement": 0.4002355722088945
},
{
"count": 3,
"confidence": 0.5029107001361107,
"reference_agreement": 0.46666675555575554
},
{
"count": 7,
"confidence": 0.554937459938247,
"reference_agreement": 0.5833334000001142
},
{
"count": 5,
"confidence": 0.6405068280987012,
"reference_agreement": 0.7933334066669266
},
{
"count": 13,
"confidence": 0.7045803972407967,
"reference_agreement": 0.7338461728206614
},
{
"count": 18,
"confidence": 0.7675469747476337,
"reference_agreement": 0.7801667003890337
},
{
"count": 11,
"confidence": 0.8164401230878163,
"reference_agreement": 0.8336366139397898
},
{
"count": 3,
"confidence": 0.8994513241595642,
"reference_agreement": 0.9422222911112911
},
{
"count": 15,
"confidence": 0.9577089746675174,
"reference_agreement": 0.9715557946667946
}
],
"expected_score_mae": 0.27761394944700535,
"gold_label_correct": 83,
"gold_label_accuracy": 0.83,
"gold_label_ece_15bin_all_decisions": 0.1551419518932857,
"full_vector_brier": 0.06091576800545618
},
"invoice_processing/disposition": {
"count": 100,
"soft_ce": 0.9672103205660691,
"kl": 0.14375101661084821,
"brier": 0.03838214719887722,
"argmax_agreement": 0.74,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 2,
"confidence": 0.3278949544267834,
"reference_agreement": 0.5716665
},
{
"count": 24,
"confidence": 0.3638647878856312,
"reference_agreement": 0.3302638513888964
},
{
"count": 16,
"confidence": 0.4371119558468517,
"reference_agreement": 0.42291669833338374
},
{
"count": 5,
"confidence": 0.48657462908745097,
"reference_agreement": 0.5753333066666932
},
{
"count": 5,
"confidence": 0.5743529947838498,
"reference_agreement": 0.39799984333333
},
{
"count": 18,
"confidence": 0.6335226967478499,
"reference_agreement": 0.637592573148225
},
{
"count": 11,
"confidence": 0.6890930143191779,
"reference_agreement": 0.6630302878788333
},
{
"count": 5,
"confidence": 0.7659479600990471,
"reference_agreement": 0.7706664166667834
},
{
"count": 2,
"confidence": 0.8601290126721075,
"reference_agreement": 0.9433335
},
{
"count": 12,
"confidence": 0.8922016440214707,
"reference_agreement": 0.9400165
}
],
"gold_label_correct": 74,
"gold_label_accuracy": 0.74,
"gold_label_ece_15bin_all_decisions": 0.20816706463754298,
"full_vector_brier": 0.07676429439775444
},
"invoice_processing/duplicate": {
"count": 100,
"soft_ce": 0.2334091092156495,
"kl": 0.02585605705240466,
"brier": 0.007688562652478144,
"argmax_agreement": 0.94,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 9,
"confidence": 0.5190740869546441,
"reference_agreement": 0.5159257777777778
},
{
"count": 6,
"confidence": 0.5672429607854125,
"reference_agreement": 0.5161113333333334
},
{
"count": 1,
"confidence": 0.6224593312018546,
"reference_agreement": 0.046667
},
{
"count": 1,
"confidence": 0.679178699175393,
"reference_agreement": 0.75
},
{
"count": 7,
"confidence": 0.9192141605617763,
"reference_agreement": 0.9795238571428572
},
{
"count": 76,
"confidence": 0.973002201351445,
"reference_agreement": 0.9698990789473684
}
],
"gold_label_correct": 94,
"gold_label_accuracy": 0.94,
"gold_label_ece_15bin_all_decisions": 0.05485489658079937,
"full_vector_brier": 0.015377125304956288
},
"invoice_processing/matches_order": {
"count": 100,
"soft_ce": 0.3306428558838697,
"kl": 0.07295218240515576,
"brier": 0.025035670669917,
"argmax_agreement": 0.91,
"hard_count": 1,
"hard_ece": 0.033085978388704196,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9669140216112958,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 2,
"confidence": 0.5,
"reference_agreement": 0.25
},
{
"count": 1,
"confidence": 0.5467381519846138,
"reference_agreement": 0.05
},
{
"count": 8,
"confidence": 0.6494972618666175,
"reference_agreement": 0.742916625
},
{
"count": 13,
"confidence": 0.6945122010594892,
"reference_agreement": 0.7448718461538462
},
{
"count": 4,
"confidence": 0.7602322944976581,
"reference_agreement": 0.49416675
},
{
"count": 7,
"confidence": 0.8444382467004503,
"reference_agreement": 0.7571428571428571
},
{
"count": 21,
"confidence": 0.8938861411785824,
"reference_agreement": 0.951095238095238
},
{
"count": 44,
"confidence": 0.9677659919679358,
"reference_agreement": 0.9671893636363635
}
],
"gold_label_correct": 91,
"gold_label_accuracy": 0.91,
"gold_label_ece_15bin_all_decisions": 0.10098650283026388,
"full_vector_brier": 0.050071341339834
},
"invoice_processing/urgency": {
"count": 100,
"soft_ce": 1.0147848337232301,
"kl": 0.07299384933168163,
"brier": 0.01930732052766958,
"argmax_agreement": 0.78,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 3,
"confidence": 0.38242658197225876,
"reference_agreement": 0.33888879444446113
},
{
"count": 13,
"confidence": 0.44677704982334543,
"reference_agreement": 0.4525640602564577
},
{
"count": 28,
"confidence": 0.49440195405535564,
"reference_agreement": 0.47142861309533574
},
{
"count": 21,
"confidence": 0.5638182000471827,
"reference_agreement": 0.6255555988889798
},
{
"count": 12,
"confidence": 0.629708983631971,
"reference_agreement": 0.4805554763889625
},
{
"count": 15,
"confidence": 0.7077412786031287,
"reference_agreement": 0.6135777344445011
},
{
"count": 7,
"confidence": 0.7426201120020457,
"reference_agreement": 0.7276191690477404
},
{
"count": 1,
"confidence": 0.8159353248895248,
"reference_agreement": 0.7166662833337166
}
],
"expected_score_mae": 0.15360283037836298,
"gold_label_correct": 75,
"gold_label_accuracy": 0.75,
"gold_label_ece_15bin_all_decisions": 0.1846883809213806,
"full_vector_brier": 0.03861464105533916
},
"security_incidents/credential_compromise": {
"count": 100,
"soft_ce": 0.7435202424681325,
"kl": 0.13032283941349765,
"brier": 0.056534706032330906,
"argmax_agreement": 0.56,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 13,
"confidence": 0.5180136079658565,
"reference_agreement": 0.5041024615384615
},
{
"count": 16,
"confidence": 0.5611483436625656,
"reference_agreement": 0.5185416875
},
{
"count": 18,
"confidence": 0.6424031158653717,
"reference_agreement": 0.4575924444444444
},
{
"count": 24,
"confidence": 0.7033728698546223,
"reference_agreement": 0.54125
},
{
"count": 13,
"confidence": 0.7727795187245017,
"reference_agreement": 0.5805127692307692
},
{
"count": 12,
"confidence": 0.8229402454540159,
"reference_agreement": 0.61611125
},
{
"count": 3,
"confidence": 0.8926362259110124,
"reference_agreement": 0.63
},
{
"count": 1,
"confidence": 0.9362850057480346,
"reference_agreement": 0.833333
}
],
"gold_label_correct": 56,
"gold_label_accuracy": 0.56,
"gold_label_ece_15bin_all_decisions": 0.12394694920782158,
"full_vector_brier": 0.11306941206466181
},
"security_incidents/disposition": {
"count": 100,
"soft_ce": 1.1881010051690235,
"kl": 0.1639027463248791,
"brier": 0.04448559481981512,
"argmax_agreement": 0.58,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 2,
"confidence": 0.32242648663151624,
"reference_agreement": 0.4083334083334083
},
{
"count": 27,
"confidence": 0.3754838182935608,
"reference_agreement": 0.32395059913580926
},
{
"count": 46,
"confidence": 0.43432437938573537,
"reference_agreement": 0.46449275565219045
},
{
"count": 19,
"confidence": 0.4882781369268498,
"reference_agreement": 0.49964906052637453
},
{
"count": 6,
"confidence": 0.5630952682918294,
"reference_agreement": 0.5416665000001667
}
],
"gold_label_correct": 58,
"gold_label_accuracy": 0.58,
"gold_label_ece_15bin_all_decisions": 0.2085843245755816,
"full_vector_brier": 0.08897118963963024
},
"security_incidents/severity": {
"count": 100,
"soft_ce": 1.315743899027136,
"kl": 0.16019039234495527,
"brier": 0.03374204583723859,
"argmax_agreement": 0.62,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 9,
"confidence": 0.30993707022864725,
"reference_agreement": 0.1892591229629178
},
{
"count": 11,
"confidence": 0.37361903832007515,
"reference_agreement": 0.35666686181831636
},
{
"count": 29,
"confidence": 0.4358135451699136,
"reference_agreement": 0.367356292413834
},
{
"count": 26,
"confidence": 0.4949142252274759,
"reference_agreement": 0.4523077201282927
},
{
"count": 16,
"confidence": 0.5669241476967354,
"reference_agreement": 0.5614581968751906
},
{
"count": 9,
"confidence": 0.629883142448191,
"reference_agreement": 0.5296296370372297
}
],
"expected_score_mae": 0.3280770725235342,
"gold_label_correct": 61,
"gold_label_accuracy": 0.61,
"gold_label_ece_15bin_all_decisions": 0.17433526899513643,
"full_vector_brier": 0.06748409167447718
},
"security_incidents/true_positive": {
"count": 100,
"soft_ce": 0.6651038558055207,
"kl": 0.08578980083661669,
"brier": 0.03646839076349738,
"argmax_agreement": 0.75,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 10,
"confidence": 0.5140548783434865,
"reference_agreement": 0.4816668
},
{
"count": 14,
"confidence": 0.5686844897001768,
"reference_agreement": 0.5449999285714285
},
{
"count": 19,
"confidence": 0.6397580847244457,
"reference_agreement": 0.6021052105263158
},
{
"count": 26,
"confidence": 0.699375893864459,
"reference_agreement": 0.6705129230769231
},
{
"count": 23,
"confidence": 0.7684838760054046,
"reference_agreement": 0.6382609565217391
},
{
"count": 8,
"confidence": 0.8259863098340201,
"reference_agreement": 0.768750125
}
],
"gold_label_correct": 75,
"gold_label_accuracy": 0.75,
"gold_label_ece_15bin_all_decisions": 0.0755676945059552,
"full_vector_brier": 0.07293678152699476
},
"security_incidents/urgency": {
"count": 100,
"soft_ce": 1.1579755423473963,
"kl": 0.13592958650868842,
"brier": 0.035556631007055334,
"argmax_agreement": 0.6,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 22,
"confidence": 0.3855800670557517,
"reference_agreement": 0.327196972045531
},
{
"count": 44,
"confidence": 0.42700538242544434,
"reference_agreement": 0.44545453469702745
},
{
"count": 17,
"confidence": 0.5028795000140648,
"reference_agreement": 0.49196084215690095
},
{
"count": 11,
"confidence": 0.5686336651767027,
"reference_agreement": 0.5948483596970949
},
{
"count": 4,
"confidence": 0.6318266146454532,
"reference_agreement": 0.55000005000025
},
{
"count": 2,
"confidence": 0.6723890027684984,
"reference_agreement": 0.6833332916667916
}
],
"expected_score_mae": 0.2845580365911216,
"gold_label_correct": 60,
"gold_label_accuracy": 0.6,
"gold_label_ece_15bin_all_decisions": 0.19018518367205345,
"full_vector_brier": 0.07111326201411067
}
}
},
"calibrated": {
"overall": {
"count": 2000,
"soft_ce": 0.8996057207086053,
"kl": 0.13224716508427645,
"brier": 0.0371886634007063,
"argmax_agreement": 0.7055,
"hard_count": 2,
"hard_ece": 0.04532655734322799,
"hard_reliability": [
{
"count": 2,
"confidence": 0.954673442656772,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.26431766017833624,
"reference_agreement": 0.6433336433336433
},
{
"count": 40,
"confidence": 0.3177876741133899,
"reference_agreement": 0.3488332323333685
},
{
"count": 214,
"confidence": 0.3738574970766125,
"reference_agreement": 0.39616281869943265
},
{
"count": 379,
"confidence": 0.4335478747419467,
"reference_agreement": 0.421979419893986
},
{
"count": 296,
"confidence": 0.5011488512826869,
"reference_agreement": 0.5069932456870009
},
{
"count": 266,
"confidence": 0.5645507022164814,
"reference_agreement": 0.5461152577318915
},
{
"count": 223,
"confidence": 0.633536195077047,
"reference_agreement": 0.5981748920314267
},
{
"count": 156,
"confidence": 0.6977244202652548,
"reference_agreement": 0.6617542448932456
},
{
"count": 130,
"confidence": 0.7643242189776432,
"reference_agreement": 0.6995359706949515
},
{
"count": 92,
"confidence": 0.8331337292292074,
"reference_agreement": 0.8546781770653425
},
{
"count": 68,
"confidence": 0.9005759800850643,
"reference_agreement": 0.9235195591667292
},
{
"count": 135,
"confidence": 0.9612556801725451,
"reference_agreement": 0.9694123619383392
}
],
"expected_score_mae": 0.26946470447726134,
"gold_label_correct": 1409,
"gold_label_accuracy": 0.7045,
"gold_label_ece_15bin_all_decisions": 0.11873688877234255,
"full_vector_brier": 0.0743773268014126
},
"type": {
"choice": {
"count": 600,
"soft_ce": 1.0381506231326942,
"kl": 0.17435602628587193,
"brier": 0.042938132074843774,
"argmax_agreement": 0.6883333333333334,
"hard_count": 1,
"hard_ece": 0.04531176280353044,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9546882371964696,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.26431766017833624,
"reference_agreement": 0.6433336433336433
},
{
"count": 28,
"confidence": 0.3194104476415337,
"reference_agreement": 0.3671427644048106
},
{
"count": 151,
"confidence": 0.37119215731679844,
"reference_agreement": 0.40550005966892083
},
{
"count": 171,
"confidence": 0.43196031154385833,
"reference_agreement": 0.43263354161798406
},
{
"count": 56,
"confidence": 0.4968234850515549,
"reference_agreement": 0.5974404757738895
},
{
"count": 41,
"confidence": 0.5598861924502044,
"reference_agreement": 0.5814633723578526
},
{
"count": 38,
"confidence": 0.6300906961195497,
"reference_agreement": 0.7054297802543781
},
{
"count": 20,
"confidence": 0.7021830670140048,
"reference_agreement": 0.6884998090001994
},
{
"count": 30,
"confidence": 0.7641036360974187,
"reference_agreement": 0.7616667017779263
},
{
"count": 32,
"confidence": 0.8347770460980283,
"reference_agreement": 0.9085641953126645
},
{
"count": 21,
"confidence": 0.90318181459874,
"reference_agreement": 0.9307778106351233
},
{
"count": 11,
"confidence": 0.949274918346995,
"reference_agreement": 0.972878994697626
}
],
"gold_label_correct": 413,
"gold_label_accuracy": 0.6883333333333334,
"gold_label_ece_15bin_all_decisions": 0.1767793119406733,
"full_vector_brier": 0.08587626414968755
},
"noul": {
"count": 600,
"soft_ce": 0.5343593764491412,
"kl": 0.08063713309657684,
"brier": 0.03351564472605923,
"argmax_agreement": 0.7666666666666667,
"hard_count": 1,
"hard_ece": 0.04534135188292543,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9546586481170746,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 59,
"confidence": 0.5162463602685387,
"reference_agreement": 0.48372877966101696
},
{
"count": 102,
"confidence": 0.566982612511053,
"reference_agreement": 0.5379411666666667
},
{
"count": 92,
"confidence": 0.638635665544844,
"reference_agreement": 0.5872825978260869
},
{
"count": 76,
"confidence": 0.6952965001408001,
"reference_agreement": 0.6477193421052632
},
{
"count": 69,
"confidence": 0.7631561753342164,
"reference_agreement": 0.6479227681159421
},
{
"count": 48,
"confidence": 0.8360191336961122,
"reference_agreement": 0.8227708958333334
},
{
"count": 44,
"confidence": 0.8983448122025038,
"reference_agreement": 0.91658325
},
{
"count": 110,
"confidence": 0.9631203976587833,
"reference_agreement": 0.9687787545454546
}
],
"gold_label_correct": 459,
"gold_label_accuracy": 0.765,
"gold_label_ece_15bin_all_decisions": 0.03628697257339877,
"full_vector_brier": 0.06703128945211846
},
"score": {
"count": 800,
"soft_ce": 1.0696318020851365,
"kl": 0.13937304317385457,
"brier": 0.03563132590108849,
"argmax_agreement": 0.6725,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 12,
"confidence": 0.31400120254772085,
"reference_agreement": 0.306110990833337
},
{
"count": 63,
"confidence": 0.38024585110410336,
"reference_agreement": 0.3737830824074847
},
{
"count": 208,
"confidence": 0.43485303487114435,
"reference_agreement": 0.41322050251512227
},
{
"count": 181,
"confidence": 0.49756579867925116,
"reference_agreement": 0.48659301701665447
},
{
"count": 123,
"confidence": 0.5640888327073031,
"reference_agreement": 0.5411110673984649
},
{
"count": 93,
"confidence": 0.6298993959087413,
"reference_agreement": 0.5651254868101266
},
{
"count": 60,
"confidence": 0.6993135701733142,
"reference_agreement": 0.670616600389039
},
{
"count": 31,
"confidence": 0.7671375253583909,
"reference_agreement": 0.7542904560324492
},
{
"count": 12,
"confidence": 0.8172099330447323,
"reference_agreement": 0.8386112533338538
},
{
"count": 3,
"confidence": 0.9150589341002214,
"reference_agreement": 0.9744443333333334
},
{
"count": 14,
"confidence": 0.9560177842150351,
"reference_agreement": 0.9716669228572801
}
],
"expected_score_mae": 0.26946470447726134,
"gold_label_correct": 537,
"gold_label_accuracy": 0.67125,
"gold_label_ece_15bin_all_decisions": 0.1383301153500044,
"full_vector_brier": 0.07126265180217697
}
},
"category": {
"agent_trace_observability": {
"count": 500,
"soft_ce": 0.9784239666090253,
"kl": 0.09995369366380179,
"brier": 0.02902756320294822,
"argmax_agreement": 0.666,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.320696328577158,
"reference_agreement": 0.333333
},
{
"count": 56,
"confidence": 0.3777104214271337,
"reference_agreement": 0.4022678604167104
},
{
"count": 142,
"confidence": 0.4351345221705775,
"reference_agreement": 0.4108685441315153
},
{
"count": 101,
"confidence": 0.5017886292643352,
"reference_agreement": 0.5098679828713507
},
{
"count": 69,
"confidence": 0.5647063184371532,
"reference_agreement": 0.5609178477778174
},
{
"count": 46,
"confidence": 0.6323688610153608,
"reference_agreement": 0.6025362419565561
},
{
"count": 25,
"confidence": 0.7011952059687334,
"reference_agreement": 0.6249331570668304
},
{
"count": 25,
"confidence": 0.7712895410262108,
"reference_agreement": 0.684133418933419
},
{
"count": 21,
"confidence": 0.8322538044396943,
"reference_agreement": 0.8557143673016929
},
{
"count": 14,
"confidence": 0.8840899517188067,
"reference_agreement": 0.8664284285714287
}
],
"expected_score_mae": 0.20048818990062078,
"gold_label_correct": 335,
"gold_label_accuracy": 0.67,
"gold_label_ece_15bin_all_decisions": 0.13395749510739433,
"full_vector_brier": 0.05805512640589644
},
"customer_service": {
"count": 500,
"soft_ce": 0.9395046654141255,
"kl": 0.20235002740464264,
"brier": 0.054900227765937,
"argmax_agreement": 0.694,
"hard_count": 1,
"hard_ece": 0.04531176280353044,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9546882371964696,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 1,
"confidence": 0.26431766017833624,
"reference_agreement": 0.6433336433336433
},
{
"count": 16,
"confidence": 0.3153945397919026,
"reference_agreement": 0.4254165366667425
},
{
"count": 43,
"confidence": 0.3676812200975334,
"reference_agreement": 0.4629575986047603
},
{
"count": 81,
"confidence": 0.43653206317486204,
"reference_agreement": 0.40526751485604134
},
{
"count": 73,
"confidence": 0.5004255499809467,
"reference_agreement": 0.5462557171690167
},
{
"count": 85,
"confidence": 0.5652857404557898,
"reference_agreement": 0.5123529096471258
},
{
"count": 71,
"confidence": 0.6289827242548214,
"reference_agreement": 0.5991033191502455
},
{
"count": 40,
"confidence": 0.6933075112774083,
"reference_agreement": 0.6937499738334586
},
{
"count": 35,
"confidence": 0.757544894325504,
"reference_agreement": 0.7550476496191563
},
{
"count": 24,
"confidence": 0.8298230320463706,
"reference_agreement": 0.858494010416886
},
{
"count": 20,
"confidence": 0.9042490371859675,
"reference_agreement": 0.9288167011668795
},
{
"count": 11,
"confidence": 0.949274918346995,
"reference_agreement": 0.972878994697626
}
],
"expected_score_mae": 0.3540890427112495,
"gold_label_correct": 347,
"gold_label_accuracy": 0.694,
"gold_label_ece_15bin_all_decisions": 0.11718544849042345,
"full_vector_brier": 0.109800455531874
},
"invoice_processing": {
"count": 500,
"soft_ce": 0.6684132719468862,
"kl": 0.09346579628199163,
"brier": 0.024514530073744333,
"argmax_agreement": 0.84,
"hard_count": 1,
"hard_ece": 0.04534135188292543,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9546586481170746,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 10,
"confidence": 0.3284781191962668,
"reference_agreement": 0.39399997833337835
},
{
"count": 38,
"confidence": 0.3695643829541523,
"reference_agreement": 0.35955257149123854
},
{
"count": 38,
"confidence": 0.43180381052982364,
"reference_agreement": 0.45737369490706775
},
{
"count": 49,
"confidence": 0.4995394302233789,
"reference_agreement": 0.4748979573470118
},
{
"count": 48,
"confidence": 0.5630772128542549,
"reference_agreement": 0.5724305864584053
},
{
"count": 48,
"confidence": 0.633751143134781,
"reference_agreement": 0.6486805458333972
},
{
"count": 46,
"confidence": 0.7006481123389033,
"reference_agreement": 0.71297823347834
},
{
"count": 30,
"confidence": 0.7644638654220844,
"reference_agreement": 0.7463223660112589
},
{
"count": 37,
"confidence": 0.8368844632918799,
"reference_agreement": 0.8846810628829918
},
{
"count": 32,
"confidence": 0.905262180250512,
"reference_agreement": 0.961760375
},
{
"count": 124,
"confidence": 0.9623184896893279,
"reference_agreement": 0.9691048380645316
}
],
"expected_score_mae": 0.2172820819264362,
"gold_label_correct": 417,
"gold_label_accuracy": 0.834,
"gold_label_ece_15bin_all_decisions": 0.1349665198192152,
"full_vector_brier": 0.049029060147488666
},
"security_incidents": {
"count": 500,
"soft_ce": 1.012080978864384,
"kl": 0.13321914298666976,
"brier": 0.04031233256019564,
"argmax_agreement": 0.622,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 13,
"confidence": 0.31228590825579433,
"reference_agreement": 0.2210255325640828
},
{
"count": 77,
"confidence": 0.3766231397145957,
"reference_agreement": 0.372489202099624
},
{
"count": 118,
"confidence": 0.4301516988277852,
"reference_agreement": 0.43542370985879325
},
{
"count": 73,
"confidence": 0.5020672766359284,
"reference_agreement": 0.48529681082197323
},
{
"count": 64,
"confidence": 0.5645118228386581,
"reference_agreement": 0.5552603373438221
},
{
"count": 58,
"confidence": 0.6398582000502253,
"reference_agreement": 0.5517815850575183
},
{
"count": 45,
"confidence": 0.6967336842994563,
"reference_agreement": 0.601407457407413
},
{
"count": 40,
"confidence": 0.7657980669345792,
"reference_agreement": 0.6255000500000001
},
{
"count": 10,
"confidence": 0.8290495284941051,
"reference_agreement": 0.7323335
},
{
"count": 2,
"confidence": 0.9042684049926732,
"reference_agreement": 0.6583330000000001
}
],
"expected_score_mae": 0.30599950337073883,
"gold_label_correct": 310,
"gold_label_accuracy": 0.62,
"gold_label_ece_15bin_all_decisions": 0.10104896950536107,
"full_vector_brier": 0.08062466512039128
}
},
"question": {
"agent_trace_observability/action": {
"count": 100,
"soft_ce": 1.1300319363326001,
"kl": 0.09598806155025466,
"brier": 0.02483876810455199,
"argmax_agreement": 0.66,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.320696328577158,
"reference_agreement": 0.333333
},
{
"count": 45,
"confidence": 0.37494667517257485,
"reference_agreement": 0.4043777937037419
},
{
"count": 20,
"confidence": 0.42036053533497303,
"reference_agreement": 0.4256666106667256
},
{
"count": 22,
"confidence": 0.5017129711343226,
"reference_agreement": 0.6168181568182417
},
{
"count": 12,
"confidence": 0.5458754881908042,
"reference_agreement": 0.5908332955556567
}
],
"gold_label_correct": 66,
"gold_label_accuracy": 0.66,
"gold_label_ece_15bin_all_decisions": 0.23452694015867082,
"full_vector_brier": 0.04967753620910398
},
"agent_trace_observability/needs_review": {
"count": 100,
"soft_ce": 0.5772502686786852,
"kl": 0.061784401057798144,
"brier": 0.027046496253035113,
"argmax_agreement": 0.78,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 17,
"confidence": 0.5199000755681038,
"reference_agreement": 0.48039217647058824
},
{
"count": 32,
"confidence": 0.5717013250980697,
"reference_agreement": 0.57124996875
},
{
"count": 15,
"confidence": 0.6285397565767751,
"reference_agreement": 0.6404444666666667
},
{
"count": 4,
"confidence": 0.7878123338263531,
"reference_agreement": 0.9075
},
{
"count": 18,
"confidence": 0.8368480740637315,
"reference_agreement": 0.8564815555555555
},
{
"count": 14,
"confidence": 0.8840899517188067,
"reference_agreement": 0.8664284285714287
}
],
"gold_label_correct": 78,
"gold_label_accuracy": 0.78,
"gold_label_ece_15bin_all_decisions": 0.10847385971036504,
"full_vector_brier": 0.05409299250607023
},
"agent_trace_observability/outcome": {
"count": 100,
"soft_ce": 1.0909319011321876,
"kl": 0.1661455365279982,
"brier": 0.04300572982539011,
"argmax_agreement": 0.62,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 5,
"confidence": 0.3938691946389984,
"reference_agreement": 0.3533333066666934
},
{
"count": 58,
"confidence": 0.4374378411470648,
"reference_agreement": 0.38236207597705485
},
{
"count": 4,
"confidence": 0.4803700853513563,
"reference_agreement": 0.58916684666676
},
{
"count": 3,
"confidence": 0.5793429761885015,
"reference_agreement": 0.7744446666666667
},
{
"count": 7,
"confidence": 0.635342607275984,
"reference_agreement": 0.6880951604762682
},
{
"count": 8,
"confidence": 0.7121878766387506,
"reference_agreement": 0.6516665095834904
},
{
"count": 15,
"confidence": 0.7635113875345624,
"reference_agreement": 0.6195556666667333
}
],
"gold_label_correct": 60,
"gold_label_accuracy": 0.6,
"gold_label_ece_15bin_all_decisions": 0.10083659412693492,
"full_vector_brier": 0.08601145965078022
},
"agent_trace_observability/risk": {
"count": 100,
"soft_ce": 0.9828391693041487,
"kl": 0.09046906089539679,
"brier": 0.026974569027456388,
"argmax_agreement": 0.67,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 6,
"confidence": 0.3849728739931042,
"reference_agreement": 0.4272221555556555
},
{
"count": 15,
"confidence": 0.44017390916712973,
"reference_agreement": 0.4377777200001155
},
{
"count": 22,
"confidence": 0.5047424360225681,
"reference_agreement": 0.541060543181922
},
{
"count": 20,
"confidence": 0.5613336073178117,
"reference_agreement": 0.5121666475000758
},
{
"count": 17,
"confidence": 0.6336701385816399,
"reference_agreement": 0.5417646784313843
},
{
"count": 12,
"confidence": 0.6951222762738744,
"reference_agreement": 0.5763887186111981
},
{
"count": 5,
"confidence": 0.7875525659292162,
"reference_agreement": 0.7093333680001679
},
{
"count": 3,
"confidence": 0.8046881866954707,
"reference_agreement": 0.8511112377785178
}
],
"expected_score_mae": 0.18766299038600726,
"gold_label_correct": 69,
"gold_label_accuracy": 0.69,
"gold_label_ece_15bin_all_decisions": 0.12744205611487303,
"full_vector_brier": 0.053949138054912776
},
"agent_trace_observability/urgency": {
"count": 100,
"soft_ce": 1.1110665575975052,
"kl": 0.08538140828756115,
"brier": 0.023272252804307492,
"argmax_agreement": 0.6,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 49,
"confidence": 0.43689567382624167,
"reference_agreement": 0.4303333642177499
},
{
"count": 36,
"confidence": 0.4938569714495298,
"reference_agreement": 0.43055556898154956
},
{
"count": 2,
"confidence": 0.5775433179069762,
"reference_agreement": 0.38333300000000003
},
{
"count": 7,
"confidence": 0.6344400930336004,
"reference_agreement": 0.5833334961905962
},
{
"count": 5,
"confidence": 0.6981819641643676,
"reference_agreement": 0.6986664453336919
},
{
"count": 1,
"confidence": 0.740555547685341,
"reference_agreement": 0.6333336333336334
}
],
"expected_score_mae": 0.2133133894152343,
"gold_label_correct": 62,
"gold_label_accuracy": 0.62,
"gold_label_ece_15bin_all_decisions": 0.13295801626402656,
"full_vector_brier": 0.046544505608614985
},
"customer_service/action": {
"count": 100,
"soft_ce": 1.2561721269415909,
"kl": 0.290314230980401,
"brier": 0.06932695222424835,
"argmax_agreement": 0.6,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.26431766017833624,
"reference_agreement": 0.6433336433336433
},
{
"count": 15,
"confidence": 0.3143181069990199,
"reference_agreement": 0.3948887724445253
},
{
"count": 38,
"confidence": 0.3671960078421094,
"reference_agreement": 0.46475026728081864
},
{
"count": 27,
"confidence": 0.43092514659511294,
"reference_agreement": 0.4196297262963243
},
{
"count": 15,
"confidence": 0.49345382078770295,
"reference_agreement": 0.6222222582223413
},
{
"count": 2,
"confidence": 0.5467833623082197,
"reference_agreement": 0.37499980000020006
},
{
"count": 2,
"confidence": 0.6432513672126315,
"reference_agreement": 0.7800003916668916
}
],
"gold_label_correct": 61,
"gold_label_accuracy": 0.61,
"gold_label_ece_15bin_all_decisions": 0.20837740157143791,
"full_vector_brier": 0.1386539044484967
},
"customer_service/category": {
"count": 100,
"soft_ce": 0.5762908772818797,
"kl": 0.16586899434803676,
"brier": 0.034944317406019984,
"argmax_agreement": 0.93,
"hard_count": 1,
"hard_ece": 0.04531176280353044,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9546882371964696,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 5,
"confidence": 0.4322684416135483,
"reference_agreement": 0.6660000793337393
},
{
"count": 3,
"confidence": 0.4905168094943096,
"reference_agreement": 0.5655556666666667
},
{
"count": 8,
"confidence": 0.5612844306710459,
"reference_agreement": 0.6383332712502996
},
{
"count": 12,
"confidence": 0.6268957032862555,
"reference_agreement": 0.7177498577498188
},
{
"count": 8,
"confidence": 0.6937005013196476,
"reference_agreement": 0.7020831275002684
},
{
"count": 13,
"confidence": 0.7640632842535814,
"reference_agreement": 0.8997435425643682
},
{
"count": 20,
"confidence": 0.8328563184532323,
"reference_agreement": 0.8911928125002632
},
{
"count": 20,
"confidence": 0.9042490371859675,
"reference_agreement": 0.9288167011668795
},
{
"count": 11,
"confidence": 0.949274918346995,
"reference_agreement": 0.972878994697626
}
],
"gold_label_correct": 94,
"gold_label_accuracy": 0.94,
"gold_label_ece_15bin_all_decisions": 0.1768752555819122,
"full_vector_brier": 0.06988863481203997
},
"customer_service/churn_risk": {
"count": 100,
"soft_ce": 1.0730591179793425,
"kl": 0.18264058554353962,
"brier": 0.051583714603065454,
"argmax_agreement": 0.71,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 1,
"confidence": 0.3315410316851427,
"reference_agreement": 0.883333
},
{
"count": 5,
"confidence": 0.3713688332387563,
"reference_agreement": 0.4493333166667167
},
{
"count": 31,
"confidence": 0.4375064676665284,
"reference_agreement": 0.388172083225866
},
{
"count": 27,
"confidence": 0.49821156690152874,
"reference_agreement": 0.5818519443210562
},
{
"count": 19,
"confidence": 0.5633747625283594,
"reference_agreement": 0.5187718885966044
},
{
"count": 15,
"confidence": 0.6252985881937501,
"reference_agreement": 0.6162223266667289
},
{
"count": 2,
"confidence": 0.6801812993683634,
"reference_agreement": 0.6716668416668417
}
],
"expected_score_mae": 0.31676493780927045,
"gold_label_correct": 71,
"gold_label_accuracy": 0.71,
"gold_label_ece_15bin_all_decisions": 0.20353240088435612,
"full_vector_brier": 0.10316742920613091
},
"customer_service/needs_human": {
"count": 100,
"soft_ce": 0.6710381093325837,
"kl": 0.12192570050358188,
"brier": 0.054199941449838425,
"argmax_agreement": 0.66,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 8,
"confidence": 0.5158550697220626,
"reference_agreement": 0.482499875
},
{
"count": 28,
"confidence": 0.5674903989336381,
"reference_agreement": 0.5160714285714285
},
{
"count": 18,
"confidence": 0.6357804167556367,
"reference_agreement": 0.5612963333333334
},
{
"count": 21,
"confidence": 0.6961004807233757,
"reference_agreement": 0.6815873333333333
},
{
"count": 21,
"confidence": 0.7545551168362051,
"reference_agreement": 0.6688889523809524
},
{
"count": 4,
"confidence": 0.8146566000120619,
"reference_agreement": 0.695
}
],
"gold_label_correct": 65,
"gold_label_accuracy": 0.65,
"gold_label_ece_15bin_all_decisions": 0.058501302775174596,
"full_vector_brier": 0.10839988289967685
},
"customer_service/urgency": {
"count": 100,
"soft_ce": 1.1209630955352308,
"kl": 0.251000625647654,
"brier": 0.06444621314651275,
"argmax_agreement": 0.57,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 18,
"confidence": 0.44444863629809206,
"reference_agreement": 0.3407406175926695
},
{
"count": 20,
"confidence": 0.5039577272096432,
"reference_agreement": 0.4638332491667292
},
{
"count": 28,
"confidence": 0.5668426468063085,
"reference_agreement": 0.4780952023809791
},
{
"count": 24,
"confidence": 0.6260414968218451,
"reference_agreement": 0.5423611534722869
},
{
"count": 9,
"confidence": 0.6893581940679483,
"reference_agreement": 0.7196295833336129
},
{
"count": 1,
"confidence": 0.7355911525357738,
"reference_agreement": 0.6833336833336833
}
],
"expected_score_mae": 0.39141314761322865,
"gold_label_correct": 56,
"gold_label_accuracy": 0.56,
"gold_label_ece_15bin_all_decisions": 0.09986013285263563,
"full_vector_brier": 0.1288924262930255
},
"invoice_processing/discrepancy_severity": {
"count": 100,
"soft_ce": 0.7824188222345687,
"kl": 0.13817545789882418,
"brier": 0.030670028432809596,
"argmax_agreement": 0.83,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 2,
"confidence": 0.33286669301517746,
"reference_agreement": 0.5433333916668917
},
{
"count": 11,
"confidence": 0.37626617164243364,
"reference_agreement": 0.40454538484852426
},
{
"count": 12,
"confidence": 0.4158789751689152,
"reference_agreement": 0.41612779859445137
},
{
"count": 5,
"confidence": 0.511693219926803,
"reference_agreement": 0.43666665333345334
},
{
"count": 5,
"confidence": 0.5577624108377157,
"reference_agreement": 0.66000016000016
},
{
"count": 7,
"confidence": 0.6424516556670419,
"reference_agreement": 0.7361905761906619
},
{
"count": 13,
"confidence": 0.7078300926213531,
"reference_agreement": 0.7558974338463236
},
{
"count": 20,
"confidence": 0.770513839139187,
"reference_agreement": 0.7909834398501792
},
{
"count": 8,
"confidence": 0.8229931371292696,
"reference_agreement": 0.8491668804171221
},
{
"count": 3,
"confidence": 0.9150589341002214,
"reference_agreement": 0.9744443333333334
},
{
"count": 14,
"confidence": 0.9560177842150351,
"reference_agreement": 0.9716669228572801
}
],
"expected_score_mae": 0.28116763438747877,
"gold_label_correct": 83,
"gold_label_accuracy": 0.83,
"gold_label_ece_15bin_all_decisions": 0.16034912415177363,
"full_vector_brier": 0.06134005686561919
},
"invoice_processing/disposition": {
"count": 100,
"soft_ce": 0.9782701754641571,
"kl": 0.15481087150893616,
"brier": 0.03951271603036011,
"argmax_agreement": 0.74,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 8,
"confidence": 0.3273809757415391,
"reference_agreement": 0.356666625
},
{
"count": 24,
"confidence": 0.3651080883483308,
"reference_agreement": 0.3415138374999964
},
{
"count": 13,
"confidence": 0.43382661018637136,
"reference_agreement": 0.5002564646154775
},
{
"count": 4,
"confidence": 0.4948415192095376,
"reference_agreement": 0.5124999458333042
},
{
"count": 15,
"confidence": 0.5688883847973131,
"reference_agreement": 0.5368888311111845
},
{
"count": 17,
"confidence": 0.6286351192794514,
"reference_agreement": 0.6950980264706401
},
{
"count": 4,
"confidence": 0.6991385791532274,
"reference_agreement": 0.7349997708334792
},
{
"count": 2,
"confidence": 0.7688077873037844,
"reference_agreement": 0.9299999999999999
},
{
"count": 12,
"confidence": 0.837978258839355,
"reference_agreement": 0.9375165
},
{
"count": 1,
"confidence": 0.8818373628541912,
"reference_agreement": 0.97
}
],
"gold_label_correct": 74,
"gold_label_accuracy": 0.74,
"gold_label_ece_15bin_all_decisions": 0.21745472516454098,
"full_vector_brier": 0.07902543206072021
},
"invoice_processing/duplicate": {
"count": 100,
"soft_ce": 0.23439542363249471,
"kl": 0.02684237146924989,
"brier": 0.007817390201271179,
"argmax_agreement": 0.94,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 9,
"confidence": 0.5172249840934735,
"reference_agreement": 0.5159257777777778
},
{
"count": 6,
"confidence": 0.560787712375488,
"reference_agreement": 0.5161113333333334
},
{
"count": 2,
"confidence": 0.6370394155634268,
"reference_agreement": 0.3983335
},
{
"count": 12,
"confidence": 0.9117582796855447,
"reference_agreement": 0.9755277499999999
},
{
"count": 71,
"confidence": 0.9649128351065898,
"reference_agreement": 0.969896676056338
}
],
"gold_label_correct": 94,
"gold_label_accuracy": 0.94,
"gold_label_ece_15bin_all_decisions": 0.05804417051238259,
"full_vector_brier": 0.015634780402542358
},
"invoice_processing/matches_order": {
"count": 100,
"soft_ce": 0.33281042566388214,
"kl": 0.0751197521851682,
"brier": 0.025450911444578592,
"argmax_agreement": 0.91,
"hard_count": 1,
"hard_ece": 0.04534135188292543,
"hard_reliability": [
{
"count": 1,
"confidence": 0.9546586481170746,
"reference_agreement": 1.0
}
],
"soft_reliability": [
{
"count": 2,
"confidence": 0.5,
"reference_agreement": 0.25
},
{
"count": 1,
"confidence": 0.5422205310594118,
"reference_agreement": 0.05
},
{
"count": 11,
"confidence": 0.643209342515234,
"reference_agreement": 0.7396969090909091
},
{
"count": 11,
"confidence": 0.6853901959758365,
"reference_agreement": 0.7666667272727273
},
{
"count": 4,
"confidence": 0.7572366579530104,
"reference_agreement": 0.50250025
},
{
"count": 16,
"confidence": 0.8447832124548076,
"reference_agreement": 0.8733124999999999
},
{
"count": 16,
"confidence": 0.900017265414687,
"reference_agreement": 0.9485416249999999
},
{
"count": 39,
"confidence": 0.9598572423050843,
"reference_agreement": 0.966743564102564
}
],
"gold_label_correct": 91,
"gold_label_accuracy": 0.91,
"gold_label_ece_15bin_all_decisions": 0.11605332143659476,
"full_vector_brier": 0.050901822889157185
},
"invoice_processing/urgency": {
"count": 100,
"soft_ce": 1.0141715127393283,
"kl": 0.07238052834777972,
"brier": 0.019121604259702194,
"argmax_agreement": 0.78,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 3,
"confidence": 0.38064151461035944,
"reference_agreement": 0.33888879444446113
},
{
"count": 13,
"confidence": 0.44448085889873,
"reference_agreement": 0.4525640602564577
},
{
"count": 29,
"confidence": 0.49257155336662545,
"reference_agreement": 0.4790805124138998
},
{
"count": 21,
"confidence": 0.561839123121125,
"reference_agreement": 0.617936518254042
},
{
"count": 11,
"confidence": 0.626064968568645,
"reference_agreement": 0.47575751969705
},
{
"count": 18,
"confidence": 0.7051208606202697,
"reference_agreement": 0.6442777231481954
},
{
"count": 4,
"confidence": 0.7392692433647956,
"reference_agreement": 0.6750002958335458
},
{
"count": 1,
"confidence": 0.8085095394162195,
"reference_agreement": 0.7166662833337166
}
],
"expected_score_mae": 0.15339652946539362,
"gold_label_correct": 75,
"gold_label_accuracy": 0.75,
"gold_label_ece_15bin_all_decisions": 0.18936000086676474,
"full_vector_brier": 0.03824320851940439
},
"security_incidents/credential_compromise": {
"count": 100,
"soft_ce": 0.7303544474693316,
"kl": 0.11715704441469675,
"brier": 0.051873219306088056,
"argmax_agreement": 0.56,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 13,
"confidence": 0.5162662097600973,
"reference_agreement": 0.5041024615384615
},
{
"count": 17,
"confidence": 0.5577534464386459,
"reference_agreement": 0.5229411764705882
},
{
"count": 21,
"confidence": 0.6372227885501143,
"reference_agreement": 0.47063480952380954
},
{
"count": 21,
"confidence": 0.691688303154376,
"reference_agreement": 0.5293650952380953
},
{
"count": 20,
"confidence": 0.7677754511149539,
"reference_agreement": 0.5835
},
{
"count": 6,
"confidence": 0.8359302010758386,
"reference_agreement": 0.7183335
},
{
"count": 2,
"confidence": 0.9042684049926732,
"reference_agreement": 0.6583330000000001
}
],
"gold_label_correct": 56,
"gold_label_accuracy": 0.56,
"gold_label_ece_15bin_all_decisions": 0.11893490648195507,
"full_vector_brier": 0.10374643861217611
},
"security_incidents/disposition": {
"count": 100,
"soft_ce": 1.1972067216437494,
"kl": 0.17300846279960486,
"brier": 0.04600030885849212,
"argmax_agreement": 0.58,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 4,
"confidence": 0.32224419861704345,
"reference_agreement": 0.29249995416670416
},
{
"count": 39,
"confidence": 0.3715904353978362,
"reference_agreement": 0.39512819239318403
},
{
"count": 48,
"confidence": 0.4302195975460818,
"reference_agreement": 0.46097220729168575
},
{
"count": 8,
"confidence": 0.501278204923741,
"reference_agreement": 0.5562498937501396
},
{
"count": 1,
"confidence": 0.5496311616587236,
"reference_agreement": 0.5166664833335167
}
],
"gold_label_correct": 58,
"gold_label_accuracy": 0.58,
"gold_label_ece_15bin_all_decisions": 0.17586552330691987,
"full_vector_brier": 0.09200061771698424
},
"security_incidents/severity": {
"count": 100,
"soft_ce": 1.3148165106977874,
"kl": 0.15926300401560653,
"brier": 0.033456273290858624,
"argmax_agreement": 0.62,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 9,
"confidence": 0.3078600014285725,
"reference_agreement": 0.1892591229629178
},
{
"count": 12,
"confidence": 0.3727428250841209,
"reference_agreement": 0.36388904000012334
},
{
"count": 29,
"confidence": 0.433631983234007,
"reference_agreement": 0.3698850510345237
},
{
"count": 26,
"confidence": 0.4919768669037995,
"reference_agreement": 0.45679487397444657
},
{
"count": 16,
"confidence": 0.5652997260041271,
"reference_agreement": 0.5520831968751906
},
{
"count": 8,
"confidence": 0.6261899107804109,
"reference_agreement": 0.5354167166668834
}
],
"expected_score_mae": 0.32567749172139204,
"gold_label_correct": 61,
"gold_label_accuracy": 0.61,
"gold_label_ece_15bin_all_decisions": 0.1787678515625339,
"full_vector_brier": 0.06691254658171725
},
"security_incidents/true_positive": {
"count": 100,
"soft_ce": 0.6603075839178701,
"kl": 0.08099352894896619,
"brier": 0.034705909701544004,
"argmax_agreement": 0.75,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 10,
"confidence": 0.5126907829686985,
"reference_agreement": 0.4816668
},
{
"count": 18,
"confidence": 0.5699609726709991,
"reference_agreement": 0.5612962222222222
},
{
"count": 25,
"confidence": 0.6460510888610294,
"reference_agreement": 0.6201333600000001
},
{
"count": 23,
"confidence": 0.7025946692840792,
"reference_agreement": 0.6679710869565217
},
{
"count": 20,
"confidence": 0.7638206827542047,
"reference_agreement": 0.6675001
},
{
"count": 4,
"confidence": 0.8187285196215046,
"reference_agreement": 0.7533335
}
],
"gold_label_correct": 75,
"gold_label_accuracy": 0.75,
"gold_label_ece_15bin_all_decisions": 0.09523922989135299,
"full_vector_brier": 0.06941181940308801
},
"security_incidents/urgency": {
"count": 100,
"soft_ce": 1.1577196305931823,
"kl": 0.13567367475447445,
"brier": 0.0355259516439954,
"argmax_agreement": 0.6,
"hard_count": 0,
"hard_ece": null,
"hard_reliability": [],
"soft_reliability": [
{
"count": 26,
"confidence": 0.38596311063456945,
"reference_agreement": 0.34250002224366877
},
{
"count": 41,
"confidence": 0.42761054306537855,
"reference_agreement": 0.45186988617891466
},
{
"count": 16,
"confidence": 0.5006824036853632,
"reference_agreement": 0.4831250822917073
},
{
"count": 12,
"confidence": 0.5661019822011998,
"reference_agreement": 0.599444329722337
},
{
"count": 4,
"confidence": 0.642325133897913,
"reference_agreement": 0.58333330000025
},
{
"count": 1,
"confidence": 0.6678840336998128,
"reference_agreement": 0.5833335833335833
}
],
"expected_score_mae": 0.2863215150200857,
"gold_label_correct": 60,
"gold_label_accuracy": 0.6,
"gold_label_ece_15bin_all_decisions": 0.1646168179614661,
"full_vector_brier": 0.0710519032879908
}
}
},
"metric_notes": {
"accuracy": "Main accuracy compares original option key with stored gold.label; argmax_agreement separately uses first maximum of rounded target distribution in input option order",
"soft_ce": "Mean natural-log cross-entropy against full normalized gold distribution",
"kl": "Mean KL(gold || prediction)",
"brier": "Existing metric field brier is half-vector Brier; full_vector_brier is twice it",
"ece": "gold_label_ece_15bin_all_decisions uses top probability vs stored-label accuracy on all examples; hard_ece uses one-hot references only; exact leaderboard scorer parity is not claimed",
"overlap": "Normalized exact state and source-ID checks; paraphrase and pretraining exposure not assessed"
}
},
"sysone_bench": {
"dataset_version": "2.1.0",
"upstream_revision": "93afcfe1bd450dc052b21768e2b95b20316ebb68",
"manifest_sha256": "0d21a64c6b61b6be7b49050dbd89f89f934912a7b586734e475308e9c9422ea2",
"raw": {
"evaluation_cases": 952,
"evaluation_decisions": 1240,
"correct": 953,
"accuracy": 0.7685483870967742,
"suites": {
"triage": {
"accuracy": 0.9270833333333334,
"cases": 48,
"decisions": 192,
"latency": {
"mean_seconds": 0.03883962416667922,
"p50_seconds": 0.03841860600016389,
"p95_seconds": 0.04178968690000602
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8541666666666666,
"macro_f1": 0.820875420875421,
"balanced_accuracy": 0.8444444444444444,
"per_class_recall": {
"billing_question": 0.2222222222222222,
"cancellation": 1.0,
"information": 1.0,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.13697967505778072,
"nll": 0.2506294304217651,
"ece_raw": 0.033931114278253044,
"ece_fitted": 0.039914814482762906
},
"noul": {
"accuracy": 0.9513888888888888,
"balanced_accuracy": 0.912037037037037,
"auroc": 0.9890689300411523,
"auprc": 0.9701626578233637,
"brier": 0.04911263462874029,
"nll": 0.1906247610132196,
"ece_raw": 0.01181868597485985,
"ece_fitted": 0.012443502260767868
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 178
},
"guardrails": {
"accuracy": 0.8125,
"cases": 48,
"decisions": 96,
"latency": {
"mean_seconds": 0.037130780229148286,
"p50_seconds": 0.03704196349997346,
"p95_seconds": 0.03867655395000611
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.8125,
"balanced_accuracy": 0.8676470588235294,
"auroc": 0.9803046218487395,
"auprc": 0.9910730108125466,
"brier": 0.14491041421218856,
"nll": 0.4276831543302518,
"ece_raw": 0.02432198575945292,
"ece_fitted": 0.026983405418465858
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 78
},
"moderation": {
"accuracy": 0.9166666666666666,
"cases": 48,
"decisions": 144,
"latency": {
"mean_seconds": 0.03827599806249301,
"p50_seconds": 0.037934129999939614,
"p95_seconds": 0.041699429449965915
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.9166666666666666,
"balanced_accuracy": 0.9106227106227107,
"auroc": 0.978021978021978,
"auprc": 0.944990138816365,
"brier": 0.06061711079669634,
"nll": 0.2023230253736832,
"ece_raw": 0.010916557094938821,
"ece_fitted": 0.011697905217155507
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 132
},
"agnews": {
"accuracy": 0.8125,
"cases": 160,
"decisions": 160,
"latency": {
"mean_seconds": 0.038758980293738435,
"p50_seconds": 0.0382035000001224,
"p95_seconds": 0.042714876549950986
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8125,
"macro_f1": 0.8064558008777994,
"balanced_accuracy": 0.8125,
"per_class_recall": {
"business": 0.85,
"scitech": 0.55,
"sports": 0.975,
"world": 0.875
},
"brier": 0.30172771335742954,
"nll": 0.6319382379388891,
"ece_raw": 0.0052000811604517765,
"ece_fitted": 0.007738817523861061
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 130
},
"emotion": {
"accuracy": 0.7239583333333334,
"cases": 192,
"decisions": 192,
"latency": {
"mean_seconds": 0.038396373302075425,
"p50_seconds": 0.03797906949989738,
"p95_seconds": 0.0415361116498957
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7239583333333334,
"macro_f1": 0.6913006811212603,
"balanced_accuracy": 0.6854851048438465,
"per_class_recall": {
"anger": 0.5714285714285714,
"fear": 0.9259259259259259,
"joy": 0.7142857142857143,
"love": 0.6,
"sadness": 0.8275862068965517,
"surprise": 0.47368421052631576
},
"brier": 0.39443242180284094,
"nll": 0.7806033322263032,
"ece_raw": 0.0032102610291140795,
"ece_fitted": 0.003416953263655513
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 139
},
"banking77_12": {
"accuracy": 0.8020833333333334,
"cases": 96,
"decisions": 96,
"latency": {
"mean_seconds": 0.03918704791666509,
"p50_seconds": 0.038639419499986616,
"p95_seconds": 0.04271133699995744
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8020833333333334,
"macro_f1": 0.7940683880389763,
"balanced_accuracy": 0.8020833333333334,
"per_class_recall": {
"card_arrival": 0.625,
"card_linking": 0.25,
"exchange_rate": 1.0,
"lost_or_stolen_card": 1.0,
"order_physical_card": 0.875,
"pin_blocked": 0.875,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875
},
"brier": 0.29183460550386847,
"nll": 0.774193924200847,
"ece_raw": 0.01661891915597873,
"ece_fitted": 0.016552925317361755
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 77
},
"mnli": {
"accuracy": 0.5416666666666666,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.0384311384083181,
"p50_seconds": 0.03797633499982567,
"p95_seconds": 0.04150163254993231
},
"primitive_metrics": {
"choice": {
"accuracy": 0.5416666666666666,
"macro_f1": 0.531705418323941,
"balanced_accuracy": 0.5416666666666666,
"per_class_recall": {
"contradiction": 0.575,
"entailment": 0.725,
"neutral": 0.325
},
"brier": 0.5619903249840926,
"nll": 0.952791383586364,
"ece_raw": 0.0067628751799769776,
"ece_fitted": 0.00836092388336822
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 65
},
"sst5": {
"accuracy": 0.3333333333333333,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.038905105700003634,
"p50_seconds": 0.03851513350002733,
"p95_seconds": 0.041932194849789536
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": 0.8020032497743765,
"rmse": 0.9909589189089723,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.737594189458718,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
},
"correct": 40
},
"multilingual_intent": {
"accuracy": 0.95,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.039158464716655315,
"p50_seconds": 0.03864724799996111,
"p95_seconds": 0.04240460730002269
},
"primitive_metrics": {
"choice": {
"accuracy": 0.95,
"macro_f1": 0.8159019343229869,
"balanced_accuracy": 0.9650000000000001,
"per_class_recall": {
"billing_question": 0.95,
"cancellation": 1.0,
"information": 0.875,
"other": null,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.12235859002542863,
"nll": 0.24005612544914204,
"ece_raw": 0.008839931130203269,
"ece_fitted": 0.0071500889173673875
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 114
}
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7690217391304348,
"macro_f1": 0.748753245465977,
"balanced_accuracy": 0.7745682853365854,
"per_class_recall": {
"anger": 0.5714285714285714,
"billing_question": 0.7241379310344828,
"business": 0.85,
"cancellation": 1.0,
"card_arrival": 0.625,
"card_linking": 0.25,
"contradiction": 0.575,
"entailment": 0.725,
"exchange_rate": 1.0,
"fear": 0.9259259259259259,
"information": 0.9,
"joy": 0.7142857142857143,
"lost_or_stolen_card": 1.0,
"love": 0.6,
"neutral": 0.325,
"order_physical_card": 0.875,
"other": null,
"pin_blocked": 0.875,
"refund": 1.0,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"sadness": 0.8275862068965517,
"scitech": 0.55,
"sports": 0.975,
"surprise": 0.47368421052631576,
"technical_help": 1.0,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875,
"world": 0.875
},
"brier": 0.3270658633907902,
"nll": 0.6528267504422001,
"ece_raw": 0.0007309538327833588,
"ece_fitted": 0.0004591988587776608
},
"noul": {
"accuracy": 0.9036458333333334,
"balanced_accuracy": 0.8834257029277777,
"auroc": 0.972521254678931,
"auprc": 0.9503028848436744,
"brier": 0.07737625808758589,
"nll": 0.2542762084776515,
"ece_raw": 0.004463339661746162,
"ece_fitted": 0.0049155860343304296
},
"score": {
"mae": 0.8020032497743765,
"rmse": 0.9909589189089723,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.737594189458718,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
}
},
"calibrated": {
"evaluation_cases": 952,
"evaluation_decisions": 1240,
"correct": 953,
"accuracy": 0.7685483870967742,
"suites": {
"triage": {
"accuracy": 0.9270833333333334,
"cases": 48,
"decisions": 192,
"latency": {
"mean_seconds": 0.03883962416667922,
"p50_seconds": 0.03841860600016389,
"p95_seconds": 0.04178968690000602
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8541666666666666,
"macro_f1": 0.820875420875421,
"balanced_accuracy": 0.8444444444444444,
"per_class_recall": {
"billing_question": 0.2222222222222222,
"cancellation": 1.0,
"information": 1.0,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.14243087345916075,
"nll": 0.2813106462197595,
"ece_raw": 0.028952845660412414,
"ece_fitted": 0.03991474162123874
},
"noul": {
"accuracy": 0.9513888888888888,
"balanced_accuracy": 0.912037037037037,
"auroc": 0.9886831275720165,
"auprc": 0.9700223660164052,
"brier": 0.052313952030937215,
"nll": 0.20641514576135578,
"ece_raw": 0.010566808507195363,
"ece_fitted": 0.012443507589047501
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 178
},
"guardrails": {
"accuracy": 0.8125,
"cases": 48,
"decisions": 96,
"latency": {
"mean_seconds": 0.037130780229148286,
"p50_seconds": 0.03704196349997346,
"p95_seconds": 0.03867655395000611
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.8125,
"balanced_accuracy": 0.8676470588235294,
"auroc": 0.9803046218487395,
"auprc": 0.9910196723104415,
"brier": 0.14399152528197112,
"nll": 0.4301690068230142,
"ece_raw": 0.027843129969293463,
"ece_fitted": 0.02698340407756053
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 78
},
"moderation": {
"accuracy": 0.9166666666666666,
"cases": 48,
"decisions": 144,
"latency": {
"mean_seconds": 0.03827599806249301,
"p50_seconds": 0.037934129999939614,
"p95_seconds": 0.041699429449965915
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.9166666666666666,
"balanced_accuracy": 0.9106227106227107,
"auroc": 0.978021978021978,
"auprc": 0.944990138816365,
"brier": 0.06164195701639249,
"nll": 0.21176174396122044,
"ece_raw": 0.014191582921902587,
"ece_fitted": 0.01169790659953756
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 132
},
"agnews": {
"accuracy": 0.8125,
"cases": 160,
"decisions": 160,
"latency": {
"mean_seconds": 0.038758980293738435,
"p50_seconds": 0.0382035000001224,
"p95_seconds": 0.042714876549950986
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8125,
"macro_f1": 0.8064558008777994,
"balanced_accuracy": 0.8125,
"per_class_recall": {
"business": 0.85,
"scitech": 0.55,
"sports": 0.975,
"world": 0.875
},
"brier": 0.2965232296874367,
"nll": 0.5908707062914559,
"ece_raw": 0.004136447508252946,
"ece_fitted": 0.007738817845546403
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 130
},
"emotion": {
"accuracy": 0.7239583333333334,
"cases": 192,
"decisions": 192,
"latency": {
"mean_seconds": 0.038396373302075425,
"p50_seconds": 0.03797906949989738,
"p95_seconds": 0.0415361116498957
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7239583333333334,
"macro_f1": 0.6913006811212603,
"balanced_accuracy": 0.6854851048438465,
"per_class_recall": {
"anger": 0.5714285714285714,
"fear": 0.9259259259259259,
"joy": 0.7142857142857143,
"love": 0.6,
"sadness": 0.8275862068965517,
"surprise": 0.47368421052631576
},
"brier": 0.39805364003393406,
"nll": 0.7908805977682863,
"ece_raw": 0.004164112978894893,
"ece_fitted": 0.0034169522756447193
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 139
},
"banking77_12": {
"accuracy": 0.8020833333333334,
"cases": 96,
"decisions": 96,
"latency": {
"mean_seconds": 0.03918704791666509,
"p50_seconds": 0.038639419499986616,
"p95_seconds": 0.04271133699995744
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8020833333333334,
"macro_f1": 0.7940683880389763,
"balanced_accuracy": 0.8020833333333334,
"per_class_recall": {
"card_arrival": 0.625,
"card_linking": 0.25,
"exchange_rate": 1.0,
"lost_or_stolen_card": 1.0,
"order_physical_card": 0.875,
"pin_blocked": 0.875,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875
},
"brier": 0.2841311372982186,
"nll": 0.7256759563838567,
"ece_raw": 0.014651013888121192,
"ece_fitted": 0.0165529263424545
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 77
},
"mnli": {
"accuracy": 0.5416666666666666,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.0384311384083181,
"p50_seconds": 0.03797633499982567,
"p95_seconds": 0.04150163254993231
},
"primitive_metrics": {
"choice": {
"accuracy": 0.5416666666666666,
"macro_f1": 0.531705418323941,
"balanced_accuracy": 0.5416666666666666,
"per_class_recall": {
"contradiction": 0.575,
"entailment": 0.725,
"neutral": 0.325
},
"brier": 0.5534738375592544,
"nll": 0.9317788432632266,
"ece_raw": 0.005991485962656614,
"ece_fitted": 0.008360924461080638
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 65
},
"sst5": {
"accuracy": 0.3333333333333333,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.038905105700003634,
"p50_seconds": 0.03851513350002733,
"p95_seconds": 0.041932194849789536
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": 0.8023576299349467,
"rmse": 0.9908613116687078,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.7382746269028682,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
},
"correct": 40
},
"multilingual_intent": {
"accuracy": 0.95,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.039158464716655315,
"p50_seconds": 0.03864724799996111,
"p95_seconds": 0.04240460730002269
},
"primitive_metrics": {
"choice": {
"accuracy": 0.95,
"macro_f1": 0.8159019343229869,
"balanced_accuracy": 0.9650000000000001,
"per_class_recall": {
"billing_question": 0.95,
"cancellation": 1.0,
"information": 0.875,
"other": null,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.13506004724959844,
"nll": 0.2766243432861846,
"ece_raw": 0.01098334737043859,
"ece_fitted": 0.007150088729354088
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 114
}
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7690217391304348,
"macro_f1": 0.748753245465977,
"balanced_accuracy": 0.7745682853365854,
"per_class_recall": {
"anger": 0.5714285714285714,
"billing_question": 0.7241379310344828,
"business": 0.85,
"cancellation": 1.0,
"card_arrival": 0.625,
"card_linking": 0.25,
"contradiction": 0.575,
"entailment": 0.725,
"exchange_rate": 1.0,
"fear": 0.9259259259259259,
"information": 0.9,
"joy": 0.7142857142857143,
"lost_or_stolen_card": 1.0,
"love": 0.6,
"neutral": 0.325,
"order_physical_card": 0.875,
"other": null,
"pin_blocked": 0.875,
"refund": 1.0,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"sadness": 0.8275862068965517,
"scitech": 0.55,
"sports": 0.975,
"surprise": 0.47368421052631576,
"technical_help": 1.0,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875,
"world": 0.875
},
"brier": 0.32691216429379927,
"nll": 0.6447888220047173,
"ece_raw": 0.000757314816236097,
"ece_fitted": 0.00045919887827611055
},
"noul": {
"accuracy": 0.9036458333333334,
"balanced_accuracy": 0.8834257029277777,
"auroc": 0.972492238052404,
"auprc": 0.950340531508563,
"brier": 0.07873134721324142,
"nll": 0.26435858535171963,
"ece_raw": 0.004409246207057182,
"ece_fitted": 0.004915585951212646
},
"score": {
"mae": 0.8023576299349467,
"rmse": 0.9908613116687078,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.7382746269028682,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
}
},
"flagged_phrase_excluded": {
"excluded_cases": [
"banking77_12-0084"
],
"raw": {
"evaluation_cases": 951,
"evaluation_decisions": 1239,
"correct": 952,
"accuracy": 0.768361581920904,
"suites": {
"triage": {
"accuracy": 0.9270833333333334,
"cases": 48,
"decisions": 192,
"latency": {
"mean_seconds": 0.03883962416667922,
"p50_seconds": 0.03841860600016389,
"p95_seconds": 0.04178968690000602
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8541666666666666,
"macro_f1": 0.820875420875421,
"balanced_accuracy": 0.8444444444444444,
"per_class_recall": {
"billing_question": 0.2222222222222222,
"cancellation": 1.0,
"information": 1.0,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.13697967505778072,
"nll": 0.2506294304217651,
"ece_raw": 0.033931114278253044,
"ece_fitted": 0.039914814482762906
},
"noul": {
"accuracy": 0.9513888888888888,
"balanced_accuracy": 0.912037037037037,
"auroc": 0.9890689300411523,
"auprc": 0.9701626578233637,
"brier": 0.04911263462874029,
"nll": 0.1906247610132196,
"ece_raw": 0.01181868597485985,
"ece_fitted": 0.012443502260767868
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 178
},
"guardrails": {
"accuracy": 0.8125,
"cases": 48,
"decisions": 96,
"latency": {
"mean_seconds": 0.037130780229148286,
"p50_seconds": 0.03704196349997346,
"p95_seconds": 0.03867655395000611
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.8125,
"balanced_accuracy": 0.8676470588235294,
"auroc": 0.9803046218487395,
"auprc": 0.9910730108125466,
"brier": 0.14491041421218856,
"nll": 0.4276831543302518,
"ece_raw": 0.02432198575945292,
"ece_fitted": 0.026983405418465858
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 78
},
"moderation": {
"accuracy": 0.9166666666666666,
"cases": 48,
"decisions": 144,
"latency": {
"mean_seconds": 0.03827599806249301,
"p50_seconds": 0.037934129999939614,
"p95_seconds": 0.041699429449965915
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.9166666666666666,
"balanced_accuracy": 0.9106227106227107,
"auroc": 0.978021978021978,
"auprc": 0.944990138816365,
"brier": 0.06061711079669634,
"nll": 0.2023230253736832,
"ece_raw": 0.010916557094938821,
"ece_fitted": 0.011697905217155507
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 132
},
"agnews": {
"accuracy": 0.8125,
"cases": 160,
"decisions": 160,
"latency": {
"mean_seconds": 0.038758980293738435,
"p50_seconds": 0.0382035000001224,
"p95_seconds": 0.042714876549950986
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8125,
"macro_f1": 0.8064558008777994,
"balanced_accuracy": 0.8125,
"per_class_recall": {
"business": 0.85,
"scitech": 0.55,
"sports": 0.975,
"world": 0.875
},
"brier": 0.30172771335742954,
"nll": 0.6319382379388891,
"ece_raw": 0.0052000811604517765,
"ece_fitted": 0.007738817523861061
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 130
},
"emotion": {
"accuracy": 0.7239583333333334,
"cases": 192,
"decisions": 192,
"latency": {
"mean_seconds": 0.038396373302075425,
"p50_seconds": 0.03797906949989738,
"p95_seconds": 0.0415361116498957
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7239583333333334,
"macro_f1": 0.6913006811212603,
"balanced_accuracy": 0.6854851048438465,
"per_class_recall": {
"anger": 0.5714285714285714,
"fear": 0.9259259259259259,
"joy": 0.7142857142857143,
"love": 0.6,
"sadness": 0.8275862068965517,
"surprise": 0.47368421052631576
},
"brier": 0.39443242180284094,
"nll": 0.7806033322263032,
"ece_raw": 0.0032102610291140795,
"ece_fitted": 0.003416953263655513
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 139
},
"banking77_12": {
"accuracy": 0.8,
"cases": 95,
"decisions": 95,
"latency": {
"mean_seconds": 0.03918948851578773,
"p50_seconds": 0.03858194800000092,
"p95_seconds": 0.04271281379994889
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8,
"macro_f1": 0.7934147932677345,
"balanced_accuracy": 0.8020833333333334,
"per_class_recall": {
"card_arrival": 0.625,
"card_linking": 0.25,
"exchange_rate": 1.0,
"lost_or_stolen_card": 1.0,
"order_physical_card": 0.875,
"pin_blocked": 0.875,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875
},
"brier": 0.29490650156145987,
"nll": 0.782323559878833,
"ece_raw": 0.01680271757873975,
"ece_fitted": 0.016734122969269134
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 76
},
"mnli": {
"accuracy": 0.5416666666666666,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.0384311384083181,
"p50_seconds": 0.03797633499982567,
"p95_seconds": 0.04150163254993231
},
"primitive_metrics": {
"choice": {
"accuracy": 0.5416666666666666,
"macro_f1": 0.531705418323941,
"balanced_accuracy": 0.5416666666666666,
"per_class_recall": {
"contradiction": 0.575,
"entailment": 0.725,
"neutral": 0.325
},
"brier": 0.5619903249840926,
"nll": 0.952791383586364,
"ece_raw": 0.0067628751799769776,
"ece_fitted": 0.00836092388336822
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 65
},
"sst5": {
"accuracy": 0.3333333333333333,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.038905105700003634,
"p50_seconds": 0.03851513350002733,
"p95_seconds": 0.041932194849789536
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": 0.8020032497743765,
"rmse": 0.9909589189089723,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.737594189458718,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
},
"correct": 40
},
"multilingual_intent": {
"accuracy": 0.95,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.039158464716655315,
"p50_seconds": 0.03864724799996111,
"p95_seconds": 0.04240460730002269
},
"primitive_metrics": {
"choice": {
"accuracy": 0.95,
"macro_f1": 0.8159019343229869,
"balanced_accuracy": 0.9650000000000001,
"per_class_recall": {
"billing_question": 0.95,
"cancellation": 1.0,
"information": 0.875,
"other": null,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.12235859002542863,
"nll": 0.24005612544914204,
"ece_raw": 0.008839931130203269,
"ece_fitted": 0.0071500889173673875
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 114
}
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7687074829931972,
"macro_f1": 0.7485002410383995,
"balanced_accuracy": 0.7745682853365854,
"per_class_recall": {
"anger": 0.5714285714285714,
"billing_question": 0.7241379310344828,
"business": 0.85,
"cancellation": 1.0,
"card_arrival": 0.625,
"card_linking": 0.25,
"contradiction": 0.575,
"entailment": 0.725,
"exchange_rate": 1.0,
"fear": 0.9259259259259259,
"information": 0.9,
"joy": 0.7142857142857143,
"lost_or_stolen_card": 1.0,
"love": 0.6,
"neutral": 0.325,
"order_physical_card": 0.875,
"other": null,
"pin_blocked": 0.875,
"refund": 1.0,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"sadness": 0.8275862068965517,
"scitech": 0.55,
"sports": 0.975,
"surprise": 0.47368421052631576,
"technical_help": 1.0,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875,
"world": 0.875
},
"brier": 0.3275108448647468,
"nll": 0.6537123942730164,
"ece_raw": 0.000732155258879059,
"ece_fitted": 0.00045994712310874536
},
"noul": {
"accuracy": 0.9036458333333334,
"balanced_accuracy": 0.8834257029277777,
"auroc": 0.972521254678931,
"auprc": 0.9503028848436744,
"brier": 0.07737625808758589,
"nll": 0.2542762084776515,
"ece_raw": 0.004463339661746162,
"ece_fitted": 0.0049155860343304296
},
"score": {
"mae": 0.8020032497743765,
"rmse": 0.9909589189089723,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.737594189458718,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
}
},
"calibrated": {
"evaluation_cases": 951,
"evaluation_decisions": 1239,
"correct": 952,
"accuracy": 0.768361581920904,
"suites": {
"triage": {
"accuracy": 0.9270833333333334,
"cases": 48,
"decisions": 192,
"latency": {
"mean_seconds": 0.03883962416667922,
"p50_seconds": 0.03841860600016389,
"p95_seconds": 0.04178968690000602
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8541666666666666,
"macro_f1": 0.820875420875421,
"balanced_accuracy": 0.8444444444444444,
"per_class_recall": {
"billing_question": 0.2222222222222222,
"cancellation": 1.0,
"information": 1.0,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.14243087345916075,
"nll": 0.2813106462197595,
"ece_raw": 0.028952845660412414,
"ece_fitted": 0.03991474162123874
},
"noul": {
"accuracy": 0.9513888888888888,
"balanced_accuracy": 0.912037037037037,
"auroc": 0.9886831275720165,
"auprc": 0.9700223660164052,
"brier": 0.052313952030937215,
"nll": 0.20641514576135578,
"ece_raw": 0.010566808507195363,
"ece_fitted": 0.012443507589047501
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 178
},
"guardrails": {
"accuracy": 0.8125,
"cases": 48,
"decisions": 96,
"latency": {
"mean_seconds": 0.037130780229148286,
"p50_seconds": 0.03704196349997346,
"p95_seconds": 0.03867655395000611
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.8125,
"balanced_accuracy": 0.8676470588235294,
"auroc": 0.9803046218487395,
"auprc": 0.9910196723104415,
"brier": 0.14399152528197112,
"nll": 0.4301690068230142,
"ece_raw": 0.027843129969293463,
"ece_fitted": 0.02698340407756053
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 78
},
"moderation": {
"accuracy": 0.9166666666666666,
"cases": 48,
"decisions": 144,
"latency": {
"mean_seconds": 0.03827599806249301,
"p50_seconds": 0.037934129999939614,
"p95_seconds": 0.041699429449965915
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": 0.9166666666666666,
"balanced_accuracy": 0.9106227106227107,
"auroc": 0.978021978021978,
"auprc": 0.944990138816365,
"brier": 0.06164195701639249,
"nll": 0.21176174396122044,
"ece_raw": 0.014191582921902587,
"ece_fitted": 0.01169790659953756
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 132
},
"agnews": {
"accuracy": 0.8125,
"cases": 160,
"decisions": 160,
"latency": {
"mean_seconds": 0.038758980293738435,
"p50_seconds": 0.0382035000001224,
"p95_seconds": 0.042714876549950986
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8125,
"macro_f1": 0.8064558008777994,
"balanced_accuracy": 0.8125,
"per_class_recall": {
"business": 0.85,
"scitech": 0.55,
"sports": 0.975,
"world": 0.875
},
"brier": 0.2965232296874367,
"nll": 0.5908707062914559,
"ece_raw": 0.004136447508252946,
"ece_fitted": 0.007738817845546403
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 130
},
"emotion": {
"accuracy": 0.7239583333333334,
"cases": 192,
"decisions": 192,
"latency": {
"mean_seconds": 0.038396373302075425,
"p50_seconds": 0.03797906949989738,
"p95_seconds": 0.0415361116498957
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7239583333333334,
"macro_f1": 0.6913006811212603,
"balanced_accuracy": 0.6854851048438465,
"per_class_recall": {
"anger": 0.5714285714285714,
"fear": 0.9259259259259259,
"joy": 0.7142857142857143,
"love": 0.6,
"sadness": 0.8275862068965517,
"surprise": 0.47368421052631576
},
"brier": 0.39805364003393406,
"nll": 0.7908805977682863,
"ece_raw": 0.004164112978894893,
"ece_fitted": 0.0034169522756447193
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 139
},
"banking77_12": {
"accuracy": 0.8,
"cases": 95,
"decisions": 95,
"latency": {
"mean_seconds": 0.03918948851578773,
"p50_seconds": 0.03858194800000092,
"p95_seconds": 0.04271281379994889
},
"primitive_metrics": {
"choice": {
"accuracy": 0.8,
"macro_f1": 0.7934147932677345,
"balanced_accuracy": 0.8020833333333334,
"per_class_recall": {
"card_arrival": 0.625,
"card_linking": 0.25,
"exchange_rate": 1.0,
"lost_or_stolen_card": 1.0,
"order_physical_card": 0.875,
"pin_blocked": 0.875,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875
},
"brier": 0.2871215913829621,
"nll": 0.7332560464140869,
"ece_raw": 0.014809638592014682,
"ece_fitted": 0.01673412400480368
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 76
},
"mnli": {
"accuracy": 0.5416666666666666,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.0384311384083181,
"p50_seconds": 0.03797633499982567,
"p95_seconds": 0.04150163254993231
},
"primitive_metrics": {
"choice": {
"accuracy": 0.5416666666666666,
"macro_f1": 0.531705418323941,
"balanced_accuracy": 0.5416666666666666,
"per_class_recall": {
"contradiction": 0.575,
"entailment": 0.725,
"neutral": 0.325
},
"brier": 0.5534738375592544,
"nll": 0.9317788432632266,
"ece_raw": 0.005991485962656614,
"ece_fitted": 0.008360924461080638
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 65
},
"sst5": {
"accuracy": 0.3333333333333333,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.038905105700003634,
"p50_seconds": 0.03851513350002733,
"p95_seconds": 0.041932194849789536
},
"primitive_metrics": {
"choice": {
"accuracy": null,
"macro_f1": null,
"balanced_accuracy": null,
"per_class_recall": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": 0.8023576299349467,
"rmse": 0.9908613116687078,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.7382746269028682,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
},
"correct": 40
},
"multilingual_intent": {
"accuracy": 0.95,
"cases": 120,
"decisions": 120,
"latency": {
"mean_seconds": 0.039158464716655315,
"p50_seconds": 0.03864724799996111,
"p95_seconds": 0.04240460730002269
},
"primitive_metrics": {
"choice": {
"accuracy": 0.95,
"macro_f1": 0.8159019343229869,
"balanced_accuracy": 0.9650000000000001,
"per_class_recall": {
"billing_question": 0.95,
"cancellation": 1.0,
"information": 0.875,
"other": null,
"refund": 1.0,
"technical_help": 1.0
},
"brier": 0.13506004724959844,
"nll": 0.2766243432861846,
"ece_raw": 0.01098334737043859,
"ece_fitted": 0.007150088729354088
},
"noul": {
"accuracy": null,
"balanced_accuracy": null,
"auroc": null,
"auprc": null,
"brier": null,
"nll": null,
"ece_raw": null,
"ece_fitted": null
},
"score": {
"mae": null,
"rmse": null,
"nearest_level_accuracy": null,
"within_one_accuracy": null,
"raw_equality_rate": null,
"spearman": null,
"ordinal_confusion": null
}
},
"correct": 114
}
},
"primitive_metrics": {
"choice": {
"accuracy": 0.7687074829931972,
"macro_f1": 0.7485002410383995,
"balanced_accuracy": 0.7745682853365854,
"per_class_recall": {
"anger": 0.5714285714285714,
"billing_question": 0.7241379310344828,
"business": 0.85,
"cancellation": 1.0,
"card_arrival": 0.625,
"card_linking": 0.25,
"contradiction": 0.575,
"entailment": 0.725,
"exchange_rate": 1.0,
"fear": 0.9259259259259259,
"information": 0.9,
"joy": 0.7142857142857143,
"lost_or_stolen_card": 1.0,
"love": 0.6,
"neutral": 0.325,
"order_physical_card": 0.875,
"other": null,
"pin_blocked": 0.875,
"refund": 1.0,
"refund_not_showing_up": 0.875,
"request_refund": 0.875,
"sadness": 0.8275862068965517,
"scitech": 0.55,
"sports": 0.975,
"surprise": 0.47368421052631576,
"technical_help": 1.0,
"terminate_account": 1.0,
"top_up_failed": 0.875,
"topping_up_by_card": 0.5,
"transaction_charged_twice": 0.875,
"world": 0.875
},
"brier": 0.3273568910489642,
"nll": 0.6456585110094694,
"ece_raw": 0.0007584697829628141,
"ece_fitted": 0.0004599471426753296
},
"noul": {
"accuracy": 0.9036458333333334,
"balanced_accuracy": 0.8834257029277777,
"auroc": 0.972492238052404,
"auprc": 0.950340531508563,
"brier": 0.07873134721324142,
"nll": 0.26435858535171963,
"ece_raw": 0.004409246207057182,
"ece_fitted": 0.004915585951212646
},
"score": {
"mae": 0.8023576299349467,
"rmse": 0.9908613116687078,
"nearest_level_accuracy": 0.3333333333333333,
"within_one_accuracy": 0.875,
"raw_equality_rate": 0.0,
"spearman": 0.7382746269028682,
"ordinal_confusion": {
"labels": [
0,
1,
2,
3,
4
],
"matrix": [
[
0,
19,
4,
1,
0
],
[
2,
10,
12,
0,
0
],
[
0,
6,
17,
1,
0
],
[
0,
1,
11,
12,
0
],
[
0,
1,
8,
14,
1
]
]
}
}
}
}
},
"score_rule": "Expected level index passed as scalar score; official floor(score+0.5) correctness, not modal-level accuracy",
"label_note": "Original v2 labels were single-review AI-assisted. Corrected v2.1.0 adopts source labels for AGNews, Banking77, MNLI and SST5 and a reviewed emotion decision set. No independent double-review claim is made.",
"official_submission": false
},
"protocol_and_overlap_notes": "See README.md. The metrics have different denominators and definitions."
}