探究大模型隐式推理是否真能像显式推理一样思考
Do LLMs Really Think Step-by-step In Implicit Reasoning?
- 通过分析隐藏状态，检验模型在隐式推理时是否真的产生中间步骤
- 提示时模型几乎不生成中间步骤，主要依赖经验而非逻辑推导
- 训练时模型才真正计算中间步骤，但效果受问题格式影响大
链式思维（Chain-of-Thought）能显著提升大模型在复杂任务上的表现，但因其推理慢、成本高，研究者尝试使用无需显式生成中间步骤的隐式链式思维。然而，这种不可见的推理过程引发疑问：隐式链式思维是否真等同于显式链式思维？本研究通过实验探查模型在隐式链式思维训练或提示时的隐藏状态信息。结果发现，当仅通过提示使用隐式链式思维时，模型几乎不进行中间步骤的思考，可能仅依赖经验；而经过训练后，模型确实会计算中间步骤。此外，两种情况下隐式链式思维的效果均受问题格式影响，表明其当前仍存在明显局限。
原文摘要 · Abstract (English)
It has been well-known that Chain-of-Thought can remarkably enhance LLMs' performance on complex tasks. However, because it also introduces slower inference speeds and higher computational costs, many researches have attempted to use implicit CoT, which does not need LLMs to explicitly generate the intermediate steps. However, the invisible reasoning process leaves us a doubt that, can implicit CoT really be equal to explicit CoT? Therefore, in this study, we address this question through experiments. We probe the information of intermediate steps from the model's hidden states when it is either trained or prompted to perform implicit CoT. The results surprisingly indicate that when prompted, LLMs hardly think about intermediate steps, suggesting they may just rely on experience rather than strict step-by-step reasoning. But when trained, they indeed calculate intermediate steps. Moreover, in both situations, we find the effect of using implicit CoT is susceptible to the format of the problem, reaffirming the current deficiency of implicit CoT.
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。