当思考失效时：大语言模型中推理对指令遵循的陷阱
When Thinking Fails: The Pitfalls of Reasoning for Instruction-Following in LLMs
AI总结：
该研究发现显式CoT推理会降低大语言模型的指令遵循准确率，提出约束注意力指标量化注意力偏移，并验证分类器选择推理等策略可有效恢复性能。
摘要：
推理增强型大语言模型（RLLMs），无论是经过显式推理训练还是通过思维链（CoT）提示，都已在许多复杂推理任务上取得了最先进的性能。然而，我们发现了一个令人惊讶且此前被忽视的现象：显式CoT推理会显著降低指令遵循准确率。我们在两个基准上评估了15个模型：IFEval（包含简单的、可通过规则验证的约束）和ComplexBench（包含复杂的组合式约束），一致观察到应用CoT提示时性能出现下降。通过大规模案例研究和基于注意力的分析，我们识别出推理有帮助（例如在格式或词汇精度方面）或有害（例如因忽略简单约束或引入不必要内容）的常见模式。我们提出了“约束注意力”这一指标，用于量化生成过程中的模型关注度，并表明CoT推理常会将注意力从与指令相关的token上转移开。为缓解这些影响，我们提出并评估了四种策略：上下文学习、自反思、自选择推理和分类器选择推理。我们的结果表明，选择性推理策略，尤其是分类器选择推理，可大幅恢复损失的性能。据我们所知，这是首项系统揭示推理引发的指令遵循失败并提出实用缓解策略的研究。
英文摘要：
Reasoning-enhanced large language models (RLLMs), whether explicitly trained for reasoning or prompted via chain-of-thought (CoT), have achieved state-of-the-art performance on many complex reasoning tasks. However, we uncover a surprising and previously overlooked phenomenon: explicit CoT reasoning can significantly degrade instruction-following accuracy. Evaluating 15 models on two benchmarks: IFEval (with simple, rule-verifiable constraints) and ComplexBench (with complex, compositional constraints), we consistently observe performance drops when CoT prompting is applied. Through large-scale case studies and an attention-based analysis, we identify common patterns where reasoning either helps (e.g., with formatting or lexical precision) or hurts (e.g., by neglecting simple constraints or introducing unnecessary content). We propose a metric, constraint attention, to quantify model focus during generation and show that CoT reasoning often diverts attention away from instruction-relevant tokens. To mitigate these effects, we introduce and evaluate four strategies: in-context learning, self-reflection, self-selective reasoning, and classifier-selective reasoning. Our results demonstrate that selective reasoning strategies, particularly classifier-selective reasoning, can substantially recover lost performance. To our knowledge, this is the first work to systematically expose reasoning-induced failures in instruction-following and offer practical mitigation strategies.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
