Answers are grounded in this paper and cite their sources.
Historical analysis reuse requires checking the original computation, current compatibility, method status and permission to use. This study evaluates whether adding nomue verification reduces the resources needed by the same large language model to make these decisions from the same historical evidence and ordinary Python tools. Forty-eight template-derived Welch-analysis tasks covered resolved, rollback and conflicting current bindings, with three scheduled sessions per configuration using gpt-5.6-sol. Across all 144 scheduled pairs, nomue reduced usage-priced API cost by 72.9% and summed task time by 55.9%. On the 103 matched pairs where the ordinary-tool comparator also reached a correct six-field decision with execution evidence under the frozen task contract, nomue reduced cost by 78.1% and time by 62.9%. A sensitivity analysis that additionally credited four disputed comparator answers retained reductions of 78.1% and 63.0%. All 48 tasks completed both configurations in the first two scheduled repetitions; this budget-uninterrupted comparison showed reductions of 76.7% and 61.2%. Thirty-three comparator sessions were stopped by conservative cumulative-budget reservations, so the overall completion difference does not measure unrestricted decision ability. Scoring covers six decision fields and execution evidence, not the accuracy or completeness of every diagnostic explanation. The findings support a bounded efficiency claim for assigned nomue use on intact synthetic records with one model and provider. Version 1 is a preprint and has not undergone journal peer review. The accompanying supplement contains all 288 observation-level records, final answers and analysis code for reproducing aggregate statistics. Full execution traces and production source remain private; the supplement does not provide public end-to-end replay of the experiment.
No takes yet. Share an insight, caveat, or question.
Tasuku Kobayashi (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: