cfo-hrm / test-best-no-reference.json
aznatkoiny's picture
Publish CFO-HRM Mindweave reconciliation checkpoint
d151de8 verified
Raw
History Blame Contribute Delete
2.68 kB
{
"ablations": [
"reference"
],
"abstention_rate": 0.19943019943019946,
"assignment_accuracy": 0.9544159544159544,
"assignment_brier_score": 0.040028318497643156,
"assignment_ece": 0.052495648772558766,
"auto_match_coverage": 0.7962962962962963,
"bank_fee_recall": 0.3333333333333333,
"checkpoint": "runs/mindweave-robust-20260723/best.pt",
"checkpoint_step": 50,
"confidence_threshold": 0.9,
"dataset": {
"bank_rows": 702,
"cases": 30,
"inferred_bank_fees": 3,
"periods": [
"2025-10",
"2025-11",
"2025-12"
],
"provenance_matches": 699,
"unresolved_rows": 0
},
"deterministic_baseline": {
"abstention_rate": 0.37037037037037035,
"assignment_accuracy": 0.9914529914529915,
"assignment_brier_score": 0.02659615478817943,
"assignment_ece": 0.11705321258430706,
"auto_match_coverage": 0.6253561253561254,
"bank_fee_recall": 0.0,
"confidence_threshold": 0.9,
"dollar_weighted_accuracy": 0.9906683143533712,
"evaluated_bank_rows": 702.0,
"evaluated_batches": 30.0,
"exact_batch_accuracy": 0.9,
"exception_accuracy": 0.9957264957264957,
"exception_macro_recall": 0.5,
"false_auto_match_rate": 0.0,
"match_f1": 0.9914163090128756,
"match_precision": 0.9914163090128756,
"match_recall": 0.9914163090128756,
"matched_exception_recall": 1.0,
"mean_assignment_confidence": 0.8743997788686826,
"selective_assignment_accuracy": 1.0,
"selective_coverage": 0.6296296296296297,
"selective_dollar_accuracy": 1.0,
"selective_dollar_coverage": 0.5729711903110923,
"supervision_coverage": 1.0,
"total_bank_rows": 702.0,
"unmatched_recall": 1.0
},
"device": "cuda",
"dollar_weighted_accuracy": 0.9430017219021095,
"eval_exception_loss": 0.22356658428907394,
"eval_loss": 0.2087109088897705,
"eval_match_loss": 0.13039841502904892,
"eval_one_to_one_loss": 0.0012840180424973369,
"evaluated_bank_rows": 702.0,
"evaluated_batches": 30.0,
"exact_batch_accuracy": 0.6,
"exception_accuracy": 0.9957264957264957,
"exception_macro_recall": 0.6659513590844063,
"false_auto_match_rate": 0.0,
"match_f1": 0.9542203147353362,
"match_precision": 0.9542203147353362,
"match_recall": 0.9542203147353362,
"matched_exception_recall": 0.9985693848354793,
"mean_assignment_confidence": 0.9043555453883532,
"segments": 4,
"selective_assignment_accuracy": 1.0,
"selective_coverage": 0.8005698005698005,
"selective_dollar_accuracy": 1.0,
"selective_dollar_coverage": 0.7870466571741648,
"split": "test",
"supervision_coverage": 1.0,
"total_bank_rows": 702.0,
"unmatched_recall": 1.0
}