mvbalaji's picture
Typed Decisions submission artifacts (predictions, metrics, overlap, evaluator)
3e203a7 verified
Raw History Blame Contribute Delete
3.21 kB
{
"model": "mvbalaji/od1-typed-decisions",
"dataset": "LocalLLaMA/typed-decisions@c76749ec58bd8c3d2ea706b31c333a9059c38f90",
"split": "test (400 cases / 2,000 decisions)",
"mode": "specialist (fitted on the train split)",
"card_metrics": {
"acc": 0.7965,
"soft_acc": 0.521672958018039,
"kl": 0.08194745417901107,
"tv": 0.13020994535529862,
"brier": 0.04549292501917222,
"ece": 0.15645043126221042,
"macro_f1": 0.672337649306128,
"score_mae": 0.20133834253665778,
"within_1": 0.985,
"acc_by_type": {
"choice": 0.7683333333333333,
"noul": 0.8666666666666667,
"score": 0.765
},
"n": 2000,
"kind": "specialist",
"mean_na_mass_removed": 0.00012872512336616762
},
"formula_check_vs_card_reference_rows": {
"Uniform": {
"acc": {
"ours": 0.274,
"card": 0.308
},
"soft_acc": {
"ours": 0.318,
"card": 0.311
},
"macro_f1": {
"ours": 0.141,
"card": 0.152
},
"kl": {
"ours": 0.444,
"card": 0.444
},
"tv": {
"ours": 0.381,
"card": 0.381
},
"brier": {
"ours": 0.238,
"card": 0.238
},
"ece": {
"ours": 0.048,
"card": 0.169
}
},
"Prior": {
"acc": {
"ours": 0.483,
"card": 0.47
},
"soft_acc": {
"ours": 0.39,
"card": 0.43
},
"macro_f1": {
"ours": 0.212,
"card": 0.207
},
"kl": {
"ours": 0.394,
"card": 0.347
},
"tv": {
"ours": 0.316,
"card": 0.317
},
"brier": {
"ours": 0.197,
"card": 0.189
},
"ece": {
"ours": 0.039,
"card": 0.088
}
}
},
"overlap_subsets": {
"full": {
"correct": 1593,
"total": 2000,
"accuracy": 0.7965,
"states": 400
},
"template_overlap": {
"correct": 336,
"total": 470,
"accuracy": 0.7148936170212766,
"states": 94
},
"lower_overlap": {
"correct": 1257,
"total": 1530,
"accuracy": 0.8215686274509804,
"states": 306
},
"near_duplicate": {
"correct": 95,
"total": 130,
"accuracy": 0.7307692307692307,
"states": 26
}
},
"corrected_input_variant_A1_not_released": {
"test": {
"correct": 1606,
"total": 2000,
"accuracy": 0.803
},
"paired_diff_vs_released": {
"delta": 0.0065,
"ci95": [
-0.0045,
0.0165
],
"states": 400
},
"fresh_teacher_labelled_set": {
"released": {
"correct": 917,
"total": 1180,
"accuracy": 0.7771186440677966
},
"a1": {
"correct": 902,
"total": 1180,
"accuracy": 0.764406779661017
},
"paired_diff": {
"delta": -0.012711864406779662,
"ci95": [
-0.026271186440677965,
0.0
],
"states": 236
}
}
}
}