kodama-core / evaluation /comparison.json
Cem13's picture
Release Kodama Core with weights, runtime, attribution, and evaluation
c7893fa verified
Raw History Blame Contribute Delete
47.9 kB
{
"protocol": {
"created_utc": "2026-09-30T15:17:02.035817+00:00",
"accuracy_n": 2496,
"groups": {
"heldout_tasks": 300,
"newsfin_long": 56,
"newsfin_short": 140,
"typed_decisions": 2000
},
"local_omni_train_or_validation_overlap": 0,
"accuracy_manifest_sha256": "77c421d0c50c76eb89016a4d77b4fde198d58c0bb0ca0d8413af171fdc183d24",
"speed_workloads_sha256": "04f2d48d8bd6c6b84563b19a55f69153e8c67c513ceb5a727918f34aad78933d",
"laya_revision": "55cf4c4ebb4ebe31b2550e8bdf3bd21b99753851",
"laya_runtime": "0.3.22",
"laya_fast_path": "Not enabled: optional TileLang is not installed. Stock public runtime measured.",
"timing": "Public Python call, tokenization included, CUDA synchronized before/after. No HTTP/network overhead.",
"normalization": "Identical normalized question text/options and state bytes; noul criteria omitted equally per the existing project protocol.",
"calibration": "Existing deployed temperatures retained; no fitting, training, or threshold selection on this test set.",
"typed_decisions": "All 400 test cases / 2000 decisions; teacher-distribution metrics and hard labels reported separately.",
"laya_td": "Task-specialized checkpoint trained upstream on the benchmark training split; not a zero-shot baseline.",
"limitations": "Public pretraining overlap unknown; desktop shares GPU; p95 from 20 calls is exploratory, not an SLA."
},
"accuracy": {
"laya": {
"newsfin_short": {
"n": 140,
"accuracy": 0.6071428571428571,
"accuracy_ci95": [
0.5285714285714286,
0.6857142857142857
],
"nll": 1.4223718041679008,
"brier_onehot": 0.5707742968520072,
"ece15": 0.16604753199968444,
"mean_confidence": 0.7438418601631568,
"uniform_accuracy": 0.2794763973450739,
"confidence_thresholds": {
"0.65": {
"n": 91,
"coverage": 0.65,
"accuracy": 0.7032967032967034
},
"0.8": {
"n": 71,
"coverage": 0.5071428571428571,
"accuracy": 0.7183098591549296
},
"0.9": {
"n": 43,
"coverage": 0.30714285714285716,
"accuracy": 0.7441860465116279
}
},
"reliability": [
{
"n": 6,
"confidence": 0.2996381148700478,
"accuracy": 0.3333333333333333
},
{
"n": 4,
"confidence": 0.37458102430107054,
"accuracy": 0.5
},
{
"n": 10,
"confidence": 0.43827573330895736,
"accuracy": 0.2
},
{
"n": 11,
"confidence": 0.5135863804562275,
"accuracy": 0.6363636363636364
},
{
"n": 7,
"confidence": 0.5655142857142857,
"accuracy": 0.42857142857142855
},
{
"n": 13,
"confidence": 0.6327408621914118,
"accuracy": 0.46153846153846156
},
{
"n": 9,
"confidence": 0.6960221081986655,
"accuracy": 0.6666666666666666
},
{
"n": 9,
"confidence": 0.7785007062507501,
"accuracy": 0.6666666666666666
},
{
"n": 19,
"confidence": 0.8315686740694261,
"accuracy": 0.7368421052631579
},
{
"n": 19,
"confidence": 0.8988784093221,
"accuracy": 0.6842105263157895
},
{
"n": 33,
"confidence": 0.984123302889007,
"accuracy": 0.7272727272727273
}
]
},
"newsfin_long": {
"n": 56,
"accuracy": 0.6607142857142857,
"accuracy_ci95": [
0.5357142857142857,
0.7857142857142857
],
"nll": 1.2032614087003177,
"brier_onehot": 0.5529698397175744,
"ece15": 0.19426640345296486,
"mean_confidence": 0.784147753253207,
"uniform_accuracy": 0.24032936018230133,
"confidence_thresholds": {
"0.65": {
"n": 41,
"coverage": 0.7321428571428571,
"accuracy": 0.6829268292682927
},
"0.8": {
"n": 34,
"coverage": 0.6071428571428571,
"accuracy": 0.6764705882352942
},
"0.9": {
"n": 21,
"coverage": 0.375,
"accuracy": 0.7142857142857143
}
},
"reliability": [
{
"n": 3,
"confidence": 0.36054520785411875,
"accuracy": 0.6666666666666666
},
{
"n": 2,
"confidence": 0.42155,
"accuracy": 0.5
},
{
"n": 3,
"confidence": 0.496984198359848,
"accuracy": 0.3333333333333333
},
{
"n": 2,
"confidence": 0.5759784578457847,
"accuracy": 0.5
},
{
"n": 6,
"confidence": 0.6315058727460587,
"accuracy": 0.6666666666666666
},
{
"n": 4,
"confidence": 0.690932906709329,
"accuracy": 0.75
},
{
"n": 2,
"confidence": 0.7695876537653765,
"accuracy": 1.0
},
{
"n": 11,
"confidence": 0.8321180001198024,
"accuracy": 0.5454545454545454
},
{
"n": 6,
"confidence": 0.9037815511951489,
"accuracy": 0.6666666666666666
},
{
"n": 17,
"confidence": 0.9809823275595868,
"accuracy": 0.7647058823529411
}
]
},
"heldout_tasks": {
"n": 300,
"accuracy": 0.66,
"accuracy_ci95": [
0.6066666666666667,
0.7133333333333334
],
"nll": 1.5088246130436413,
"brier_onehot": 0.49109099242960164,
"ece15": 0.1363256368468856,
"mean_confidence": 0.7706782853152053,
"uniform_accuracy": 0.3491236190667541,
"confidence_thresholds": {
"0.65": {
"n": 195,
"coverage": 0.65,
"accuracy": 0.7897435897435897
},
"0.8": {
"n": 160,
"coverage": 0.5333333333333333,
"accuracy": 0.81875
},
"0.9": {
"n": 136,
"coverage": 0.4533333333333333,
"accuracy": 0.8235294117647058
}
},
"reliability": [
{
"n": 1,
"confidence": 0.3066,
"accuracy": 0.0
},
{
"n": 15,
"confidence": 0.381084709529047,
"accuracy": 0.3333333333333333
},
{
"n": 26,
"confidence": 0.4301559573447903,
"accuracy": 0.23076923076923078
},
{
"n": 26,
"confidence": 0.5058806642403066,
"accuracy": 0.6538461538461539
},
{
"n": 20,
"confidence": 0.5698708586837585,
"accuracy": 0.45
},
{
"n": 20,
"confidence": 0.6283789555242396,
"accuracy": 0.4
},
{
"n": 15,
"confidence": 0.7039680905242808,
"accuracy": 0.6
},
{
"n": 17,
"confidence": 0.7685221145643976,
"accuracy": 0.7647058823529411
},
{
"n": 16,
"confidence": 0.83745649416744,
"accuracy": 0.6875
},
{
"n": 30,
"confidence": 0.9053593478879268,
"accuracy": 0.8
},
{
"n": 114,
"confidence": 0.988545481205944,
"accuracy": 0.8421052631578947
}
]
},
"typed_decisions": {
"n": 2000,
"accuracy": 0.364,
"accuracy_ci95": [
0.339,
0.3885
],
"nll": 1.340352049822895,
"brier_onehot": 0.7689268871867527,
"ece15": 0.21537093573642535,
"mean_confidence": 0.5793709357364253,
"uniform_accuracy": 0.3175,
"confidence_thresholds": {
"0.65": {
"n": 608,
"coverage": 0.304,
"accuracy": 0.45230263157894735
},
"0.8": {
"n": 314,
"coverage": 0.157,
"accuracy": 0.4745222929936306
},
"0.9": {
"n": 102,
"coverage": 0.051,
"accuracy": 0.4803921568627451
}
},
"reliability": [
{
"n": 13,
"confidence": 0.25357506365936605,
"accuracy": 0.23076923076923078
},
{
"n": 122,
"confidence": 0.30950061571517834,
"accuracy": 0.23770491803278687
},
{
"n": 228,
"confidence": 0.3670941177030421,
"accuracy": 0.2719298245614035
},
{
"n": 233,
"confidence": 0.4363279138876397,
"accuracy": 0.2875536480686695
},
{
"n": 256,
"confidence": 0.5028284181113388,
"accuracy": 0.30859375
},
{
"n": 337,
"confidence": 0.5673213889715166,
"accuracy": 0.3857566765578635
},
{
"n": 256,
"confidence": 0.6312426961373134,
"accuracy": 0.3828125
},
{
"n": 131,
"confidence": 0.6972084159331299,
"accuracy": 0.42748091603053434
},
{
"n": 110,
"confidence": 0.765639844893953,
"accuracy": 0.5
},
{
"n": 131,
"confidence": 0.831922665422135,
"accuracy": 0.48091603053435117
},
{
"n": 152,
"confidence": 0.8988913804538349,
"accuracy": 0.4605263157894737
},
{
"n": 31,
"confidence": 0.9563516129032258,
"accuracy": 0.5161290322580645
}
]
}
},
"laya_td": {
"newsfin_short": {
"n": 140,
"accuracy": 0.6142857142857143,
"accuracy_ci95": [
0.5357142857142857,
0.6928571428571428
],
"nll": 1.2190687244899345,
"brier_onehot": 0.5204963383039098,
"ece15": 0.10012283564261254,
"mean_confidence": 0.6159351724446068,
"uniform_accuracy": 0.2794763973450739,
"confidence_thresholds": {
"0.65": {
"n": 57,
"coverage": 0.40714285714285714,
"accuracy": 0.7719298245614035
},
"0.8": {
"n": 24,
"coverage": 0.17142857142857143,
"accuracy": 0.8333333333333334
},
"0.9": {
"n": 15,
"coverage": 0.10714285714285714,
"accuracy": 0.8666666666666667
}
},
"reliability": [
{
"n": 4,
"confidence": 0.2333194916155514,
"accuracy": 0.0
},
{
"n": 6,
"confidence": 0.302599998681,
"accuracy": 0.16666666666666666
},
{
"n": 13,
"confidence": 0.36711329999243336,
"accuracy": 0.38461538461538464
},
{
"n": 11,
"confidence": 0.4237506192347789,
"accuracy": 0.45454545454545453
},
{
"n": 18,
"confidence": 0.511127863340612,
"accuracy": 0.6111111111111112
},
{
"n": 18,
"confidence": 0.5625467793025789,
"accuracy": 0.6111111111111112
},
{
"n": 14,
"confidence": 0.6292863952966607,
"accuracy": 0.7142857142857143
},
{
"n": 16,
"confidence": 0.7084717264693747,
"accuracy": 0.8125
},
{
"n": 16,
"confidence": 0.767652998875487,
"accuracy": 0.625
},
{
"n": 4,
"confidence": 0.8280340331933613,
"accuracy": 0.5
},
{
"n": 7,
"confidence": 0.8859190199163888,
"accuracy": 1.0
},
{
"n": 13,
"confidence": 0.9831229609827629,
"accuracy": 0.8461538461538461
}
]
},
"newsfin_long": {
"n": 56,
"accuracy": 0.625,
"accuracy_ci95": [
0.5,
0.75
],
"nll": 0.9473697717806606,
"brier_onehot": 0.5052564707694435,
"ece15": 0.1282473784340939,
"mean_confidence": 0.6203433290814203,
"uniform_accuracy": 0.24032936018230133,
"confidence_thresholds": {
"0.65": {
"n": 23,
"coverage": 0.4107142857142857,
"accuracy": 0.7391304347826086
},
"0.8": {
"n": 9,
"coverage": 0.16071428571428573,
"accuracy": 0.8888888888888888
},
"0.9": {
"n": 8,
"coverage": 0.14285714285714285,
"accuracy": 1.0
}
},
"reliability": [
{
"n": 1,
"confidence": 0.26117388261173874,
"accuracy": 0.0
},
{
"n": 3,
"confidence": 0.30080999099909994,
"accuracy": 1.0
},
{
"n": 7,
"confidence": 0.36791477390129057,
"accuracy": 0.42857142857142855
},
{
"n": 4,
"confidence": 0.4233358232121082,
"accuracy": 0.5
},
{
"n": 2,
"confidence": 0.5034250124987502,
"accuracy": 0.5
},
{
"n": 7,
"confidence": 0.5632139915755685,
"accuracy": 0.42857142857142855
},
{
"n": 10,
"confidence": 0.6319262866286629,
"accuracy": 0.6
},
{
"n": 5,
"confidence": 0.70236,
"accuracy": 0.8
},
{
"n": 8,
"confidence": 0.7650717596759675,
"accuracy": 0.625
},
{
"n": 1,
"confidence": 0.8101810181018102,
"accuracy": 0.0
},
{
"n": 4,
"confidence": 0.9228529563843856,
"accuracy": 1.0
},
{
"n": 4,
"confidence": 0.9760745273582075,
"accuracy": 1.0
}
]
},
"heldout_tasks": {
"n": 300,
"accuracy": 0.63,
"accuracy_ci95": [
0.5766666666666667,
0.6833333333333333
],
"nll": 0.8284287990601404,
"brier_onehot": 0.4631184306750694,
"ece15": 0.06414849953652801,
"mean_confidence": 0.6360782473450751,
"uniform_accuracy": 0.3491236190667541,
"confidence_thresholds": {
"0.65": {
"n": 147,
"coverage": 0.49,
"accuracy": 0.8367346938775511
},
"0.8": {
"n": 93,
"coverage": 0.31,
"accuracy": 0.8494623655913979
},
"0.9": {
"n": 33,
"coverage": 0.11,
"accuracy": 0.9393939393939394
}
},
"reliability": [
{
"n": 4,
"confidence": 0.2544439619109396,
"accuracy": 0.5
},
{
"n": 21,
"confidence": 0.30978139390048054,
"accuracy": 0.3333333333333333
},
{
"n": 31,
"confidence": 0.3617844818117397,
"accuracy": 0.25806451612903225
},
{
"n": 21,
"confidence": 0.4331388408231027,
"accuracy": 0.42857142857142855
},
{
"n": 35,
"confidence": 0.5037856980053429,
"accuracy": 0.45714285714285713
},
{
"n": 29,
"confidence": 0.5615874261913189,
"accuracy": 0.5862068965517241
},
{
"n": 18,
"confidence": 0.6337426574404175,
"accuracy": 0.5555555555555556
},
{
"n": 23,
"confidence": 0.6971773777962333,
"accuracy": 0.8695652173913043
},
{
"n": 25,
"confidence": 0.7676468601541085,
"accuracy": 0.84
},
{
"n": 42,
"confidence": 0.8360288136827827,
"accuracy": 0.7857142857142857
},
{
"n": 27,
"confidence": 0.8914703703703704,
"accuracy": 0.8148148148148148
},
{
"n": 24,
"confidence": 0.9689162707922456,
"accuracy": 1.0
}
]
},
"typed_decisions": {
"n": 2000,
"accuracy": 0.7425,
"accuracy_ci95": [
0.7225,
0.762
],
"nll": 0.7436633069434331,
"brier_onehot": 0.4235755392133792,
"ece15": 0.2097095930770926,
"mean_confidence": 0.5356001013082533,
"uniform_accuracy": 0.3175,
"confidence_thresholds": {
"0.65": {
"n": 392,
"coverage": 0.196,
"accuracy": 0.9617346938775511
},
"0.8": {
"n": 129,
"coverage": 0.0645,
"accuracy": 0.9844961240310077
},
"0.9": {
"n": 0,
"coverage": 0.0,
"accuracy": null
}
},
"reliability": [
{
"n": 1,
"confidence": 0.2657,
"accuracy": 1.0
},
{
"n": 34,
"confidence": 0.31793218780429244,
"accuracy": 0.23529411764705882
},
{
"n": 289,
"confidence": 0.3751108461445502,
"accuracy": 0.5155709342560554
},
{
"n": 387,
"confidence": 0.4333358872661366,
"accuracy": 0.6382428940568475
},
{
"n": 454,
"confidence": 0.4990018080263838,
"accuracy": 0.7400881057268722
},
{
"n": 269,
"confidence": 0.5663064150458634,
"accuracy": 0.7843866171003717
},
{
"n": 206,
"confidence": 0.6269887756403439,
"accuracy": 0.9077669902912622
},
{
"n": 129,
"confidence": 0.6995473861386212,
"accuracy": 0.9534883720930233
},
{
"n": 102,
"confidence": 0.7637360536255958,
"accuracy": 0.9411764705882353
},
{
"n": 122,
"confidence": 0.8339088022933011,
"accuracy": 0.9836065573770492
},
{
"n": 7,
"confidence": 0.8706124126698384,
"accuracy": 1.0
}
]
}
},
"jevlike": {
"newsfin_short": {
"n": 140,
"accuracy": 0.6357142857142857,
"accuracy_ci95": [
0.5571428571428572,
0.7142857142857143
],
"nll": 0.8810642049018346,
"brier_onehot": 0.4594364556189868,
"ece15": 0.09626927118430638,
"mean_confidence": 0.7003433504563511,
"uniform_accuracy": 0.2794763973450739,
"confidence_thresholds": {
"0.65": {
"n": 74,
"coverage": 0.5285714285714286,
"accuracy": 0.8243243243243243
},
"0.8": {
"n": 51,
"coverage": 0.36428571428571427,
"accuracy": 0.9019607843137255
},
"0.9": {
"n": 36,
"coverage": 0.2571428571428571,
"accuracy": 0.9166666666666666
}
},
"reliability": [
{
"n": 4,
"confidence": 0.31435249081859906,
"accuracy": 0.0
},
{
"n": 4,
"confidence": 0.3683047456063695,
"accuracy": 0.5
},
{
"n": 13,
"confidence": 0.43412605262834336,
"accuracy": 0.3076923076923077
},
{
"n": 14,
"confidence": 0.5018749093199865,
"accuracy": 0.42857142857142855
},
{
"n": 17,
"confidence": 0.5625792631911706,
"accuracy": 0.5294117647058824
},
{
"n": 15,
"confidence": 0.624756897657645,
"accuracy": 0.5333333333333333
},
{
"n": 6,
"confidence": 0.6991859315065998,
"accuracy": 0.8333333333333334
},
{
"n": 16,
"confidence": 0.765267463088596,
"accuracy": 0.5625
},
{
"n": 11,
"confidence": 0.8288046343252776,
"accuracy": 0.9090909090909091
},
{
"n": 11,
"confidence": 0.9037337492339783,
"accuracy": 0.7272727272727273
},
{
"n": 29,
"confidence": 0.9729322257523271,
"accuracy": 0.9655172413793104
}
]
},
"newsfin_long": {
"n": 56,
"accuracy": 0.7142857142857143,
"accuracy_ci95": [
0.5892857142857143,
0.8214285714285714
],
"nll": 0.7747561322704637,
"brier_onehot": 0.3982446040756354,
"ece15": 0.1139029663017493,
"mean_confidence": 0.7618069910792377,
"uniform_accuracy": 0.24032936018230133,
"confidence_thresholds": {
"0.65": {
"n": 38,
"coverage": 0.6785714285714286,
"accuracy": 0.8157894736842105
},
"0.8": {
"n": 33,
"coverage": 0.5892857142857143,
"accuracy": 0.8484848484848485
},
"0.9": {
"n": 26,
"coverage": 0.4642857142857143,
"accuracy": 0.9230769230769231
}
},
"reliability": [
{
"n": 1,
"confidence": 0.25787782972278783,
"accuracy": 0.0
},
{
"n": 2,
"confidence": 0.30044849334548607,
"accuracy": 0.0
},
{
"n": 1,
"confidence": 0.33464981710949976,
"accuracy": 1.0
},
{
"n": 5,
"confidence": 0.4328088221147037,
"accuracy": 0.6
},
{
"n": 2,
"confidence": 0.5085217603938686,
"accuracy": 0.5
},
{
"n": 6,
"confidence": 0.5693763805464213,
"accuracy": 0.5
},
{
"n": 4,
"confidence": 0.6606546915216647,
"accuracy": 0.75
},
{
"n": 2,
"confidence": 0.689887285232544,
"accuracy": 0.5
},
{
"n": 4,
"confidence": 0.8263626768434467,
"accuracy": 0.5
},
{
"n": 10,
"confidence": 0.9067653114584996,
"accuracy": 0.9
},
{
"n": 19,
"confidence": 0.9723644101980917,
"accuracy": 0.8947368421052632
}
]
},
"heldout_tasks": {
"n": 300,
"accuracy": 0.58,
"accuracy_ci95": [
0.5266666666666666,
0.6366666666666667
],
"nll": 0.8911869889628542,
"brier_onehot": 0.5134231622791497,
"ece15": 0.08311700140649199,
"mean_confidence": 0.6243801023474437,
"uniform_accuracy": 0.3491236190667541,
"confidence_thresholds": {
"0.65": {
"n": 137,
"coverage": 0.45666666666666667,
"accuracy": 0.7445255474452555
},
"0.8": {
"n": 64,
"coverage": 0.21333333333333335,
"accuracy": 0.8125
},
"0.9": {
"n": 37,
"coverage": 0.12333333333333334,
"accuracy": 0.8648648648648649
}
},
"reliability": [
{
"n": 1,
"confidence": 0.23789036628440569,
"accuracy": 1.0
},
{
"n": 21,
"confidence": 0.29895297878815114,
"accuracy": 0.23809523809523808
},
{
"n": 28,
"confidence": 0.3638488940938339,
"accuracy": 0.5
},
{
"n": 28,
"confidence": 0.4302174204297628,
"accuracy": 0.25
},
{
"n": 26,
"confidence": 0.5017499667404475,
"accuracy": 0.4230769230769231
},
{
"n": 33,
"confidence": 0.5623847003664137,
"accuracy": 0.5151515151515151
},
{
"n": 36,
"confidence": 0.6323279372286387,
"accuracy": 0.6666666666666666
},
{
"n": 30,
"confidence": 0.700325317035411,
"accuracy": 0.6666666666666666
},
{
"n": 33,
"confidence": 0.7652276990151226,
"accuracy": 0.696969696969697
},
{
"n": 18,
"confidence": 0.8330853245824539,
"accuracy": 0.7777777777777778
},
{
"n": 23,
"confidence": 0.9067225341857484,
"accuracy": 0.7391304347826086
},
{
"n": 23,
"confidence": 0.960167099210642,
"accuracy": 0.9130434782608695
}
]
},
"typed_decisions": {
"n": 2000,
"accuracy": 0.4595,
"accuracy_ci95": [
0.433,
0.485
],
"nll": 1.3444350869309427,
"brier_onehot": 0.7480423473427633,
"ece15": 0.23062323931638637,
"mean_confidence": 0.683465700725639,
"uniform_accuracy": 0.3175,
"confidence_thresholds": {
"0.65": {
"n": 1103,
"coverage": 0.5515,
"accuracy": 0.5058930190389845
},
"0.8": {
"n": 870,
"coverage": 0.435,
"accuracy": 0.5149425287356322
},
"0.9": {
"n": 506,
"coverage": 0.253,
"accuracy": 0.5731225296442688
}
},
"reliability": [
{
"n": 2,
"confidence": 0.2636414220883421,
"accuracy": 0.0
},
{
"n": 112,
"confidence": 0.31072671180670275,
"accuracy": 0.3392857142857143
},
{
"n": 220,
"confidence": 0.36609577130409987,
"accuracy": 0.38181818181818183
},
{
"n": 183,
"confidence": 0.43270138063245134,
"accuracy": 0.366120218579235
},
{
"n": 161,
"confidence": 0.4997503451031316,
"accuracy": 0.4472049689440994
},
{
"n": 126,
"confidence": 0.5668673187578344,
"accuracy": 0.46825396825396826
},
{
"n": 119,
"confidence": 0.6321133013126154,
"accuracy": 0.4369747899159664
},
{
"n": 111,
"confidence": 0.7002285366333068,
"accuracy": 0.4864864864864865
},
{
"n": 96,
"confidence": 0.7659850098182762,
"accuracy": 0.46875
},
{
"n": 203,
"confidence": 0.8370369211558824,
"accuracy": 0.4729064039408867
},
{
"n": 347,
"confidence": 0.9023447644701712,
"accuracy": 0.3919308357348703
},
{
"n": 320,
"confidence": 0.970245866206789,
"accuracy": 0.675
}
]
}
},
"omni": {
"newsfin_short": {
"n": 140,
"accuracy": 0.6642857142857143,
"accuracy_ci95": [
0.5857142857142857,
0.7428571428571429
],
"nll": 1.0239277480868136,
"brier_onehot": 0.4833709544993766,
"ece15": 0.086806222512248,
"mean_confidence": 0.6900754779093492,
"uniform_accuracy": 0.2794763973450739,
"confidence_thresholds": {
"0.65": {
"n": 86,
"coverage": 0.6142857142857143,
"accuracy": 0.7906976744186046
},
"0.8": {
"n": 56,
"coverage": 0.4,
"accuracy": 0.8392857142857143
},
"0.9": {
"n": 36,
"coverage": 0.2571428571428571,
"accuracy": 0.8611111111111112
}
},
"reliability": [
{
"n": 2,
"confidence": 0.16850247623873293,
"accuracy": 0.0
},
{
"n": 4,
"confidence": 0.25145491106610773,
"accuracy": 0.5
},
{
"n": 3,
"confidence": 0.30153553466505506,
"accuracy": 1.0
},
{
"n": 11,
"confidence": 0.36396977029923866,
"accuracy": 0.36363636363636365
},
{
"n": 10,
"confidence": 0.4277883482274337,
"accuracy": 0.4
},
{
"n": 7,
"confidence": 0.5038134981024585,
"accuracy": 0.42857142857142855
},
{
"n": 10,
"confidence": 0.5570586830396977,
"accuracy": 0.6
},
{
"n": 18,
"confidence": 0.6415175415281524,
"accuracy": 0.5555555555555556
},
{
"n": 9,
"confidence": 0.6942038665711675,
"accuracy": 0.7777777777777778
},
{
"n": 10,
"confidence": 0.7485815357473065,
"accuracy": 0.7
},
{
"n": 11,
"confidence": 0.825007934222517,
"accuracy": 0.8181818181818182
},
{
"n": 19,
"confidence": 0.8981023578758742,
"accuracy": 0.7894736842105263
},
{
"n": 26,
"confidence": 0.98324251744931,
"accuracy": 0.8846153846153846
}
]
},
"newsfin_long": {
"n": 56,
"accuracy": 0.5535714285714286,
"accuracy_ci95": [
0.4107142857142857,
0.6785714285714286
],
"nll": 1.0747286532578382,
"brier_onehot": 0.5316587234647131,
"ece15": 0.20183825996408222,
"mean_confidence": 0.6893829250347027,
"uniform_accuracy": 0.24032936018230133,
"confidence_thresholds": {
"0.65": {
"n": 32,
"coverage": 0.5714285714285714,
"accuracy": 0.78125
},
"0.8": {
"n": 23,
"coverage": 0.4107142857142857,
"accuracy": 0.7391304347826086
},
"0.9": {
"n": 13,
"coverage": 0.23214285714285715,
"accuracy": 0.9230769230769231
}
},
"reliability": [
{
"n": 1,
"confidence": 0.17508739445476842,
"accuracy": 0.0
},
{
"n": 3,
"confidence": 0.30233710435052524,
"accuracy": 0.3333333333333333
},
{
"n": 6,
"confidence": 0.3677328588099678,
"accuracy": 0.16666666666666666
},
{
"n": 2,
"confidence": 0.44699478884245114,
"accuracy": 0.0
},
{
"n": 4,
"confidence": 0.4922646316138709,
"accuracy": 0.0
},
{
"n": 5,
"confidence": 0.5582380460237041,
"accuracy": 0.4
},
{
"n": 4,
"confidence": 0.622083597856155,
"accuracy": 0.75
},
{
"n": 4,
"confidence": 0.6889762293752948,
"accuracy": 1.0
},
{
"n": 4,
"confidence": 0.7732806664163974,
"accuracy": 0.75
},
{
"n": 6,
"confidence": 0.8408541777219157,
"accuracy": 0.5
},
{
"n": 4,
"confidence": 0.8780159919507924,
"accuracy": 0.5
},
{
"n": 13,
"confidence": 0.9821660460455572,
"accuracy": 0.9230769230769231
}
]
},
"heldout_tasks": {
"n": 300,
"accuracy": 0.6333333333333333,
"accuracy_ci95": [
0.58,
0.69
],
"nll": 0.8258969806486628,
"brier_onehot": 0.4569257718777433,
"ece15": 0.09554911465107167,
"mean_confidence": 0.6315795431338078,
"uniform_accuracy": 0.3491236190667541,
"confidence_thresholds": {
"0.65": {
"n": 148,
"coverage": 0.49333333333333335,
"accuracy": 0.7972972972972973
},
"0.8": {
"n": 61,
"coverage": 0.20333333333333334,
"accuracy": 0.8688524590163934
},
"0.9": {
"n": 26,
"coverage": 0.08666666666666667,
"accuracy": 0.9615384615384616
}
},
"reliability": [
{
"n": 1,
"confidence": 0.1046529915301365,
"accuracy": 0.0
},
{
"n": 4,
"confidence": 0.2577561216819676,
"accuracy": 0.75
},
{
"n": 11,
"confidence": 0.302752906130526,
"accuracy": 0.2727272727272727
},
{
"n": 27,
"confidence": 0.3719780866552427,
"accuracy": 0.4444444444444444
},
{
"n": 17,
"confidence": 0.43146238975131995,
"accuracy": 0.4117647058823529
},
{
"n": 37,
"confidence": 0.49454230114799697,
"accuracy": 0.35135135135135137
},
{
"n": 36,
"confidence": 0.5579886897891916,
"accuracy": 0.7222222222222222
},
{
"n": 29,
"confidence": 0.6272471234022386,
"accuracy": 0.41379310344827586
},
{
"n": 29,
"confidence": 0.6959158351866087,
"accuracy": 0.8275862068965517
},
{
"n": 48,
"confidence": 0.7644648380498819,
"accuracy": 0.7708333333333334
},
{
"n": 27,
"confidence": 0.8256353707236136,
"accuracy": 0.7777777777777778
},
{
"n": 17,
"confidence": 0.9039215980455362,
"accuracy": 0.9411764705882353
},
{
"n": 17,
"confidence": 0.9717127266067751,
"accuracy": 0.9411764705882353
}
]
},
"typed_decisions": {
"n": 2000,
"accuracy": 0.4765,
"accuracy_ci95": [
0.4545,
0.499
],
"nll": 1.0863653374222986,
"brier_onehot": 0.6166074547470112,
"ece15": 0.11012610270598122,
"mean_confidence": 0.5241069191428419,
"uniform_accuracy": 0.3175,
"confidence_thresholds": {
"0.65": {
"n": 523,
"coverage": 0.2615,
"accuracy": 0.7074569789674953
},
"0.8": {
"n": 251,
"coverage": 0.1255,
"accuracy": 0.6573705179282868
},
"0.9": {
"n": 100,
"coverage": 0.05,
"accuracy": 0.72
}
},
"reliability": [
{
"n": 10,
"confidence": 0.25256444635115816,
"accuracy": 0.3
},
{
"n": 189,
"confidence": 0.30877873898245745,
"accuracy": 0.2698412698412698
},
{
"n": 472,
"confidence": 0.36585330883687267,
"accuracy": 0.2817796610169492
},
{
"n": 302,
"confidence": 0.42887444845445033,
"accuracy": 0.38741721854304634
},
{
"n": 243,
"confidence": 0.49993474017946105,
"accuracy": 0.3868312757201646
},
{
"n": 155,
"confidence": 0.5580824847934738,
"accuracy": 0.6645161290322581
},
{
"n": 218,
"confidence": 0.6316491082827173,
"accuracy": 0.7889908256880734
},
{
"n": 77,
"confidence": 0.6980893665549133,
"accuracy": 0.8441558441558441
},
{
"n": 83,
"confidence": 0.7678636543006925,
"accuracy": 0.6024096385542169
},
{
"n": 112,
"confidence": 0.8273256031580205,
"accuracy": 0.6160714285714286
},
{
"n": 69,
"confidence": 0.8956754913922487,
"accuracy": 0.6086956521739131
},
{
"n": 70,
"confidence": 0.9641727620937258,
"accuracy": 0.7714285714285715
}
]
}
}
},
"speed": {
"short_1q": {
"n": 20,
"p50_ms": 26.517087999991418,
"p95_ms": 29.029050500071207,
"mean_ms": 27.113317650002955,
"min_ms": 24.377534999985073,
"max_ms": 29.310716000054526,
"samples_ms": [
26.25009999997019,
25.432290000026114,
26.320786000042062,
28.980330999957005,
26.646911999932854,
28.12136200009263,
25.81902299993999,
26.38726400004998,
25.392123000074207,
28.523082999981852,
26.326767000000473,
28.44737800000985,
28.463398999974743,
29.310716000054526,
28.692237999848658,
24.377534999985073,
28.48372800008292,
25.53521799995906,
29.014226000072085,
25.741874000004827
],
"warmup_calls": 3,
"first_shape_call_ms": 27.415586999950392,
"questions_per_call": 1,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"short_5q": {
"n": 20,
"p50_ms": 60.07641500002592,
"p95_ms": 65.5484159998764,
"mean_ms": 58.98180429998092,
"min_ms": 51.847296999994796,
"max_ms": 66.96372600003997,
"samples_ms": [
56.865289999905144,
53.42559599989727,
54.647222000085094,
60.657620000029056,
61.10556999988148,
55.25912699999935,
53.47630400001435,
52.484391000007236,
51.847296999994796,
60.60912699990695,
59.04960399993797,
61.89990600000783,
61.19320799984962,
65.47392599986779,
63.0494520000866,
55.42915399996673,
66.96372600003997,
63.058288999854994,
63.59757400014132,
59.54370300014489
],
"warmup_calls": 3,
"first_shape_call_ms": 54.20621600001141,
"questions_per_call": 5,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"short_10q": {
"n": 20,
"p50_ms": 104.81810399994629,
"p95_ms": 119.18314979989191,
"mean_ms": 105.92105819998778,
"min_ms": 88.4976249999454,
"max_ms": 120.43251000000055,
"samples_ms": [
92.77988300004836,
104.6406630000547,
96.89195900000414,
119.11739399988619,
105.96081699986826,
103.6471459999575,
98.7587220001842,
98.4291290001238,
88.4976249999454,
114.96733699982542,
109.29184199994779,
104.99554499983788,
110.70990500002154,
115.22976200012636,
118.5596349998832,
95.96929799999998,
120.43251000000055,
104.30120299997725,
115.14932800014321,
100.09146099991995
],
"warmup_calls": 3,
"first_shape_call_ms": 93.11735100004626,
"questions_per_call": 10,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"short_50q": {
"n": 20,
"p50_ms": 682.1585460000961,
"p95_ms": 754.3870993999008,
"mean_ms": 689.1575670499719,
"min_ms": 618.0243280000468,
"max_ms": 763.9201289998709,
"samples_ms": [
674.0678559999651,
678.1810669999686,
618.0243280000468,
753.8853609999023,
707.519105000074,
670.8609360000537,
680.0984779999908,
671.1490820000563,
641.9182600000113,
688.8830089999374,
693.8877620000312,
684.2186140002013,
674.6635569998034,
738.8550079999732,
707.0535409998229,
627.7758280000398,
747.2571369999059,
698.2282959997974,
763.9201289998709,
662.7039869999862
],
"warmup_calls": 3,
"first_shape_call_ms": 596.6050259999065,
"questions_per_call": 50,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"long_1q": {
"n": 20,
"p50_ms": 123.90287249991161,
"p95_ms": 131.76431719999755,
"mean_ms": 124.6096985999884,
"min_ms": 119.37196000008043,
"max_ms": 131.96851399993648,
"samples_ms": [
127.72182899993823,
126.77874800010613,
123.88399099995695,
123.92175399986627,
125.15549199997622,
131.75357000000076,
130.19190099998923,
131.96851399993648,
127.03383000007307,
119.37196000008043,
123.77765799988083,
120.07203400003164,
119.416425000054,
119.82972900000277,
120.96001899999465,
120.67090499999722,
120.97485899994354,
120.59792499985633,
128.58475899997757,
129.5280700001058
],
"warmup_calls": 3,
"first_shape_call_ms": 122.09491500016156,
"questions_per_call": 1,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"long_5q": {
"n": 20,
"p50_ms": 685.7518550001487,
"p95_ms": 728.8032338999642,
"mean_ms": 696.9230963500081,
"min_ms": 674.529241000073,
"max_ms": 733.069986000146,
"samples_ms": [
678.4065869999267,
675.5267930000173,
680.8560730000863,
675.0147320001361,
674.6946719999869,
680.0351919998775,
715.0993089999247,
724.1639889998623,
722.2430469998926,
724.2127150000215,
733.069986000146,
728.5786679999546,
719.8392059999605,
713.666480000029,
690.3053800001544,
681.198330000143,
674.529241000073,
675.8001450000393,
680.2929479999875,
690.9284339999431
],
"warmup_calls": 3,
"first_shape_call_ms": 690.0589629999558,
"questions_per_call": 5,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"long_10q": {
"n": 20,
"p50_ms": 1321.0661100000607,
"p95_ms": 1336.965719349837,
"mean_ms": 1321.2527787499992,
"min_ms": 1311.405681999986,
"max_ms": 1341.6646919999948,
"samples_ms": [
1333.539275000021,
1322.840443000132,
1312.4046030000045,
1312.6979950000077,
1326.8468310000117,
1323.0493790001674,
1317.6598679999643,
1312.3101529999985,
1316.093828000021,
1311.405681999986,
1311.9573989999935,
1324.5337289999952,
1341.6646919999948,
1319.2917769999895,
1324.1005910001604,
1312.415898999916,
1336.7184049998286,
1325.0531029998456,
1324.322876999986,
1316.149045999964
],
"warmup_calls": 3,
"first_shape_call_ms": 1341.1980430000767,
"questions_per_call": 10,
"window_counts": [
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1,
1
]
},
"throughput": {
"states": 16,
"questions": 48,
"repetitions": 3,
"batch_size_laya": 8,
"median_ms": 507.1860810000999,
"questions_per_second": 94.63982115863813,
"samples_ms": [
506.4106110000921,
507.1860810000999,
507.4538389999361
]
}
}
}