论文提出 commitment-verdict 框架识别推理模型的有害过度思考
Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking
做推理模型部署的人值得看这篇:它证明只预测最终对错无法决定何时停止推理。文中给出可量化的 commitment-verdict 框架,并测出继续思考平均掉 3.6 分。想自己写停止规则的话,先按这个框架把该测的量测出来。
一篇 arXiv 论文研究推理模型的"有害过度思考"现象:模型已得出正确答案后继续推理反而可能改错。论文指出,现有 final-correctness 探针只估计 verdict(最终答案正确的概率),无法判断该不该停止,因为 verdict 作为固定结果的预测没有可漂移的规律。
作者提出 commitment-verdict gap,即"现在停止会给的答案"与"推理完会给的答案"之间的差,等于继续推理预期带来的答案质量变化。他们还在 MATH-500 上把准确率-预算曲线分解为链内答案变化和仍在运行的链的构成变化两部分。
实验结果:DeepSeek-R1-Distill-Qwen-1.5B 在 379 个已持有正确答案的点上继续推理平均损失 3.6 分(95% CI [2.2, 5.2])。在来自三个家族的六个推理模型、MATH-500 和 OlympiadBench、预算最高 16384 token 的设定下,准确率下降主要来自哪些链还在跑,而仍在推理的链会保持或改善答案。最优停止规则在 1.5B 和 7B 规模上最多分别增益 +1.8 和 +2.5 分。
Conserved Verdicts, Drifting Commitments: Identifying Harmful Overthinking
A reasoning model that has reached a correct answer can lose it by continuing to think. Deciding whether to stop means comparing two probabilities at the same point of a chain: that the answer the model would give if stopped now is correct (its commitment), and that the answer it will finish with is correct (its verdict). We show that the verdict alone, which is what final-correctness probes estimate, cannot make this comparison. As a forecast of a fixed outcome it has no predictable drift, whether continuing helps or hurts. The commitment-verdict gap equals the expected remaining change in answer quality and the accuracy gained by stopping. A second identity splits any accuracy-versus-budget curve into answer change within chains and change in which chains are still running. We measure both on MATH-500. At all 379 points where DeepSeek-R1-Distill-Qwen-1.5B holds a correct answer, selected without regard to how its chain ends, continuing loses 3.6 points on average (95% CI [2.2, 5.2]), and the loss concentrates in a few problems. In six reasoning models from three families, on MATH-500 and OlympiadBench and with budgets up to 16,384 tokens, declining accuracy curves come from which chains remain, while the chains still reasoning keep or improve their answers. An oracle stopping rule gains at most +1.8 and +2.5 points at 1.5B and 7B. The framework says what to measure before building a stopping rule.
This report is categorized under AITOP's "Papers" desk, focusing on US frontier AI labs and Silicon Valley technology leaders. Within our cross-regional analytical framework, it directly maps to AI developer toolchains, coding agents, and engineering workflows. Key entities and technical topics under continuous tracking include DeepSeek-R1-Distill-Qwen-1.5B, reasoning models, MATH-500, OlympiadBench.
Primary source verification: published directly by arXiv: DeepSeek with first-party authority.