[PROPOSED EVALUATIONS]
Cash proof evaluation protocol
Account coverage, period alignment, reconciliations, and unexplained differences.
Proposed evaluation only. No public dataset, completed benchmark run, ranked models, or measured results are available for these tracks.
Proposed evaluation
Use constructed cases with known timing differences, transfers, missing statements, and duplicate imports. Separately report reconciliation arithmetic, coverage detection, and classification agreement against a documented reviewer reference. State tolerance and units in advance. A zero difference is not a completeness metric. There are no empirical results in this paper.
Before reporting results
Publish the task definition, data permissions, source coverage, reference-label process, metrics, denominators, and limitations. Separate calculation correctness from professional judgment and disclose unresolved disagreement.