Self-Contrast: Better Reflection Through Inconsistent Solving Perspectives
Wenqi Zhang1, Yongliang Shen1, Linjuan Wu1 Qiuying Peng2, Jun Wang2, Yueting Zhuang1, Weiming Lu1† 1College of Computer Science and Technology, Zhejiang University 2OPPO Research Institute, China {zhangwenqi,
Abstract
The reflection capacity of Large Language Model (LLM) has garnered extensive attention. A post-hoc prompting strategy, e.g., reflexion and self-refine, refines LLM’s response based on self-evaluated or external feedback. However, recent research indicates without external feedback, LLM’s intrinsic reflection is unstable. Our investigation unveils that the key bottleneck is the quality of the self-evaluated feedback. We find LLMs often exhibit overconfidence or high randomness when self-evaluate, offering stubborn or inconsistent feedback, which causes poor reflection. To remedy this, we advocate Self-Contrast: It adaptively explores diverse solving perspectives tailored to the request, contrasts the differences, and summarizes these discrepancies into a checklist which could be used to re-examine and eliminate discrepancies. Our method provides LLM with diverse perspectives to alleviate stubborn biases. Moreover, their discrepancies indicate potential errors or inherent uncertainties that LLM often overlooks. Reflecting upon these can prompt more accurate and stable reflection. Experiments conducted on a series of reasoning and translation tasks with different LLMs serve to undersc
中文速览
大语言模型（LLM）在回答问题后往往可以通过"反思"来修正错误，但研究发现，没有外部反馈时这种自我反思效果极差——根源在于模型自我评估质量低劣：要么过度自信、认为原答案没问题，要么多次评估同一答案却给出截然不同的结论。为此，论文提出了"自我对比"（Self-Contrast）方法：让模型先自动生成多个不同视角的解题提示词，从不同角度各自作答，再对比这些答案之间的差异，把差异点归纳成一份"检查清单"，最后依据清单重新审视并修正答案。实验在数学推理和机器翻译任务上验证了该方法，结果显示它显著优于直接反思基线，反思结果也更加稳定。这项工作的价值在于揭示了LLM自我反思失败的真正瓶颈，并提供了一种无需任何外部监督信号就能让模型更可靠地自我纠错的实用路径。
原文 arXiv:2401.02009;中英对照 + 大白话阅读 https://aha.fim.ai/paper/2401.02009v3