Same-Number Citation Swaps: Stress-Testing Jev as a Financial Evidence Judge

Excerpt

arXiv:2610.08675v1 Announce Type: new Abstract: Financial reports repeat values across periods, metrics and accounting lines, allowing an LLM-generated calculation to be numerically correct while citing the wrong financial role. We evaluate what probabilistic evidence verification adds beyond number matching using Jev as a source-support verifier for GPT-4.1-mini calculation traces. A signed-number-at-pointer baseline explains most recovery over exact quotation checks. To isolate the remaining r