强模型纠错能力反而更弱，因为错得更深。
Decomposing LLM Self-Correction: The Accuracy-Correction Paradox and Error Depth Hypothesis
- 拆解自纠正为检测、定位、修正三步，发现错误深度影响纠错效果。
- 弱模型纠正率是强模型的1.6倍（26.8% vs 16.7%），因错得浅。
- 提示错误位置反而降低纠错率，适合设计不依赖外部反馈的自修正系统者关注。
大型语言模型（LLMs）常被认为具备自纠正能力，但近期研究指出其内在自纠正——在无外部反馈下修正自身输出——仍基本无效。本文系统分解自纠正为三个子能力：错误检测、错误定位与错误修正。在GSM8K-Complex数据集上对三个主流模型进行跨模型实验（每模型500样本，共346个错误），发现显著的准确率-纠正悖论：较弱模型（GPT-3.5，准确率66%）的内在纠正率比更强模型（DeepSeek，准确率94%）高1.6倍（26.8% vs 16.7%）。提出错误深度假说：更强模型虽出错少，但错误更深，难以自纠正。不同架构的错误检测率差异巨大（10%至82%），但检测能力不能预测纠正成功率——Claude仅检测10%错误，却能实现29%的内在纠正。意外发现：提供错误位置提示反而损害所有模型表现。研究挑战了模型能力与自提升之间的线性假设，对自精炼流水线设计具有重要启示。
原文摘要 · Abstract (English)
Large Language Models (LLMs) are widely believed to possess self-correction capabilities, yet recent studies suggest that intrinsic self-correction--where models correct their own outputs without external feedback--remains largely ineffective. In this work, we systematically decompose self-correction into three distinct sub-capabilities: error detection, error localization, and error correction. Through cross-model experiments on GSM8K-Complex (n=500 per model, 346 total errors) with three major LLMs, we uncover a striking Accuracy-Correction Paradox: weaker models (GPT-3.5, 66% accuracy) achieve 1.6x higher intrinsic correction rates than stronger models (DeepSeek, 94% accuracy)--26.8% vs 16.7%. We propose the Error Depth Hypothesis: stronger models make fewer but deeper errors that resist self-correction. Error detection rates vary dramatically across architectures (10% to 82%), yet detection capability does not predict correction success--Claude detects only 10% of errors but corrects 29% intrinsically. Surprisingly, providing error location hints hurts all models. Our findings challenge linear assumptions about model capability and self-improvement, with important implications for the design of self-refinement pipelines.
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。