发现o1类模型测试时扩展能力存疑，长思维链反而易出错。
Revisiting the Test-Time Scaling of o1-like Models: Do they Truly Possess Test-Time Scaling Capabilities?
- 通过对比长短思维链，发现正确答案常比错误答案更短。
- 长思维链含更多自我修正，导致性能下降，非计算资源越多越好。
- 提出最短多数投票法，结合并行扩展与思维链长度特征，提升效果。
大型语言模型（LLMs）中测试时扩展（test-time scaling）的兴起，以OpenAI的o1系列为代表，在推理阶段通过动态增加计算资源来提升推理能力。尽管QwQ、Deepseek-R1（R1）和LIMO等后续模型复制了这一进展，但它们是否真正具备测试时扩展能力仍不明确。本研究发现，这些o1类模型的思维链（CoT）越长，准确率并未持续提升；事实上，同一问题的正确解答往往比错误解答更短。进一步分析表明，该现象与模型自修正能力密切相关：更长的思维链包含更多自修正行为，常导致性能退化。我们对比了QwQ、R1和LIMO上顺序与并行扩展策略的表现，发现并行策略在覆盖范围和可扩展性上更优。基于此，提出“最短多数投票”方法，结合并行扩展与思维链长度特性，显著优于传统多数投票，大幅提升了模型的测试时扩展能力。
原文摘要 · Abstract (English)
The advent of test-time scaling in large language models (LLMs), exemplified by OpenAI's o1 series, has advanced reasoning capabilities by scaling computational resource allocation during inference. While successors like QwQ, Deepseek-R1 (R1) and LIMO replicate these advancements, whether these models truly possess test-time scaling capabilities remains underexplored. This study found that longer CoTs of these o1-like models do not consistently enhance accuracy; in fact, correct solutions are often shorter than incorrect ones for the same questions. Further investigation shows this phenomenon is closely related to models' self-revision capabilities - longer CoTs contain more self-revisions, which often lead to performance degradation. We then compare sequential and parallel scaling strategies on QwQ, R1 and LIMO, finding that parallel scaling achieves better coverage and scalability. Based on these insights, we propose Shortest Majority Vote, a method that combines parallel scaling strategies with CoT length characteristics, significantly improving models' test-time scalability compared to conventional majority voting approaches.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。