arXiv：cs.CL· Yuqi Zhang, Di Zhang·· 16 小时前
RCCM：诊断推理提示词迁移中的收益与损害
RCCM: Diagnosing Gains and Harms in Reasoning-Prompt Transfer
AI 导读
RCCM（Reasoning-Control Capability Mismatch）通过将同一问题的答案与生成结束记录配对，诊断推理提示词迁移中的性能反转：在五个开源权重部署与九种策略上，总体准确率掩盖了大量问题级别的收益与损害。
正文
Abstract:A reasoning prompt can improve one deployment and degrade another. RCCM (Reasoning-Control Capability Mismatch) examines these reversals by pairing answers to the same questions with records of how generation ends. Across five open-weight deployments and nine strategies, aggregate accuracy hides substantial question-level gains and harms; a common-budget MATH500 evaluation examines transfer on a separate sample. On 60 additional questions, concise reasoning improves Phi's automatic answer matching at a 2,048-token cap across three collections, but sharply reduces Qwen's. Author-reviewed final-answer commitments give a smaller, statistically unresolved Phi gain. Reserving an answer stage also improves matching, with additional input and call costs. These results show why a prompt benefit must be interpreted alongside its execution and evaluation: matching a reference answer need not mean delivering it, and an apparent compression benefit may admit another explanation. RCCM records this evidence for reproducible transfer comparisons.
| Comments: | 21 pages, 2 figures. Includes a size-limited reproducibility subset as ancillary files |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2610.19872 [cs.CL] |
| (or arXiv:2610.19872v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2610.19872 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Yuqi Zhang [view email]
[v1]
Wed, 14 Oct 2026 09:31:04 UTC (7,706 KB)
来源：arXiv：cs.CL · arxiv.org