kev-4b-code-verify-v2 / run_eval.json
jtatman's picture
Upload code-verify-4b-v2
6ac368e verified
Raw History Blame Contribute Delete
6.51 kB
{
"development": {
"baseline kev": {
"auroc": 0.9711489898989899,
"acc": 0.9464285714285714,
"brier": 0.04904470588256766,
"ece": 0.06521459834615231,
"cov@2%": 0.7611607142857143,
"cov@5%": 0.8236607142857143,
"cov@10%": 0.8928571428571429
},
"fine-tuned kev": {
"auroc": 0.9684027777777777,
"acc": 0.9464285714285714,
"brier": 0.04231278466552767,
"ece": 0.05751592743965165,
"cov@2%": 0.65625,
"cov@5%": 0.8415178571428571,
"cov@10%": 0.8928571428571429
},
"evidence score": {
"auroc": 0.9590277777777778,
"acc": 0.9553571428571429,
"brier": 0.04669189453125,
"ece": 0.07882254464285716,
"cov@2%": 0.6852678571428571,
"cov@5%": 0.8058035714285714,
"cov@10%": 0.8794642857142857
},
"mean(evidence, baseline)": {
"auroc": 0.9763257575757576,
"acc": 0.9464285714285714,
"brier": 0.039230741564232695,
"ece": 0.07375138990908783,
"cov@2%": 0.7834821428571429,
"cov@5%": 0.84375,
"cov@10%": 0.8928571428571429
},
"mean(evidence, fine-tuned)": {
"auroc": 0.9768623737373737,
"acc": 0.9575892857142857,
"brier": 0.03790862505206988,
"ece": 0.06543454258088069,
"cov@2%": 0.7321428571428571,
"cov@5%": 0.84375,
"cov@10%": 0.8928571428571429
}
},
"test_unseen_coder": {
"baseline kev": {
"auroc": 0.9638752052545156,
"acc": 0.9295774647887324,
"brier": 0.06336158381334493,
"ece": 0.039717450620457166,
"cov@2%": 0.28169014084507044,
"cov@5%": 0.4788732394366197,
"cov@10%": 0.647887323943662
},
"fine-tuned kev": {
"auroc": 0.9893267651888341,
"acc": 0.9154929577464789,
"brier": 0.055145570909923364,
"ece": 0.05822282925003647,
"cov@2%": 0.4507042253521127,
"cov@5%": 0.6197183098591549,
"cov@10%": 0.647887323943662
},
"evidence score": {
"auroc": 0.9433497536945813,
"acc": 0.9154929577464789,
"brier": 0.08483714788732394,
"ece": 0.1338028169014084,
"cov@2%": 0.15492957746478872,
"cov@5%": 0.15492957746478872,
"cov@10%": 0.5915492957746479
},
"mean(evidence, baseline)": {
"auroc": 0.9786535303776683,
"acc": 0.9154929577464789,
"brier": 0.06692693831242305,
"ece": 0.08200779432387027,
"cov@2%": 0.4084507042253521,
"cov@5%": 0.5774647887323944,
"cov@10%": 0.6197183098591549
},
"mean(evidence, fine-tuned)": {
"auroc": 0.9811165845648604,
"acc": 0.9154929577464789,
"brier": 0.0623680250089783,
"ece": 0.08442088475882523,
"cov@2%": 0.39436619718309857,
"cov@5%": 0.5774647887323944,
"cov@10%": 0.647887323943662
}
},
"test_unseen_coder_all_tasks": {
"baseline kev": {
"auroc": 0.9705357142857143,
"acc": 0.9262295081967213,
"brier": 0.0600707465937226,
"ece": 0.02946642500461688,
"cov@2%": 0.3114754098360656,
"cov@5%": 0.5122950819672131,
"cov@10%": 0.6229508196721312
},
"fine-tuned kev": {
"auroc": 0.9839285714285714,
"acc": 0.9467213114754098,
"brier": 0.03959559951764304,
"ece": 0.02761496864374707,
"cov@2%": 0.4713114754098361,
"cov@5%": 0.5881147540983607,
"cov@10%": 0.6352459016393442
},
"evidence score": {
"auroc": 0.943543956043956,
"acc": 0.9057377049180327,
"brier": 0.09111320255251247,
"ece": 0.09400301997638061,
"cov@2%": 0.0,
"cov@5%": 0.20901639344262296,
"cov@10%": 0.6188524590163934
},
"mean(evidence, baseline)": {
"auroc": 0.9703983516483516,
"acc": 0.9262295081967213,
"brier": 0.06636942416344632,
"ece": 0.06373308985502091,
"cov@2%": 0.3094262295081967,
"cov@5%": 0.5061475409836066,
"cov@10%": 0.6270491803278688
},
"mean(evidence, fine-tuned)": {
"auroc": 0.9801339285714286,
"acc": 0.9426229508196722,
"brier": 0.05355769253857334,
"ece": 0.08502821807951433,
"cov@2%": 0.3442622950819672,
"cov@5%": 0.5799180327868853,
"cov@10%": 0.6352459016393442
}
},
"test_unseen_qwen3_5_9b_defiant_iq2m": {
"baseline kev": {
"auroc": 0.9166666666666666,
"acc": 0.9166666666666666,
"brier": 0.07422605985253933,
"ece": 0.07673742820730627,
"cov@2%": 0.20833333333333334,
"cov@5%": 0.2777777777777778,
"cov@10%": 0.7083333333333334
},
"fine-tuned kev": {
"auroc": 0.9461805555555556,
"acc": 0.9027777777777778,
"brier": 0.06887600741202703,
"ece": 0.029364648847460262,
"cov@2%": 0.3472222222222222,
"cov@5%": 0.4444444444444444,
"cov@10%": 0.6944444444444444
},
"evidence score": {
"auroc": 0.9817708333333334,
"acc": 0.9166666666666666,
"brier": 0.0666232638888889,
"ece": 0.11284722222222222,
"cov@2%": 0.4861111111111111,
"cov@5%": 0.6111111111111112,
"cov@10%": 0.7083333333333334
},
"mean(evidence, baseline)": {
"auroc": 0.9800347222222222,
"acc": 0.9305555555555556,
"brier": 0.06047935935863502,
"ece": 0.10117383649315945,
"cov@2%": 0.5555555555555556,
"cov@5%": 0.6388888888888888,
"cov@10%": 0.7361111111111112
},
"mean(evidence, fine-tuned)": {
"auroc": 0.9782986111111112,
"acc": 0.9166666666666666,
"brier": 0.059680672299262584,
"ece": 0.1075121568818185,
"cov@2%": 0.5833333333333334,
"cov@5%": 0.625,
"cov@10%": 0.7222222222222222
}
},
"test_unseen_qwen3_5_9b_defiant_iq2m_all_tasks": {
"baseline kev": {
"auroc": 0.9662908930150309,
"acc": 0.9382716049382716,
"brier": 0.05267810391166952,
"ece": 0.03597538353446246,
"cov@2%": 0.345679012345679,
"cov@5%": 0.6172839506172839,
"cov@10%": 0.7119341563786008
},
"fine-tuned kev": {
"auroc": 0.9838822575891542,
"acc": 0.9567901234567902,
"brier": 0.035262343000345975,
"ece": 0.02702561098457988,
"cov@2%": 0.4794238683127572,
"cov@5%": 0.6666666666666666,
"cov@10%": 0.7119341563786008
},
"evidence score": {
"auroc": 0.9647159593280283,
"acc": 0.9238683127572016,
"brier": 0.06824839826404823,
"ece": 0.07486938929840165,
"cov@2%": 0.0,
"cov@5%": 0.49176954732510286,
"cov@10%": 0.6934156378600823
},
"mean(evidence, baseline)": {
"auroc": 0.9758325965222517,
"acc": 0.9403292181069959,
"brier": 0.05304097125217489,
"ece": 0.06640969399185599,
"cov@2%": 0.5185185185185185,
"cov@5%": 0.6213991769547325,
"cov@10%": 0.7119341563786008
},
"mean(evidence, fine-tuned)": {
"auroc": 0.9836243737105806,
"acc": 0.9567901234567902,
"brier": 0.04255867127685762,
"ece": 0.07530209304107607,
"cov@2%": 0.5246913580246914,
"cov@5%": 0.668724279835391,
"cov@10%": 0.7119341563786008
}
}
}