arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2511.09381cs.CLcs.AI
收藏

自我校正大型语言模型:生成任务 vs 多项选择任务

Self-Correcting Large Language Models: Generation vs. Multiple Choice

University College London(伦敦大学学院);Amazon AGI(亚马逊人工智能实验室);University of Sheffield(谢菲尔德大学);University of Southern California(南加州大学)

Hossein A. Rahmani, Satyapriya Krishna, Xi Wang, Mohammadmehdi Naghiaei, Emine Yilmaz

AI总结:

本文对比研究了大型语言模型在开放式生成和多项选择任务中的自我校正机制,发现两者改进模式与失败模式存在差异,强调自我校正机制设计需考虑任务结构与输出空间的相互作用,对智能体应用有启示。

摘要:

大型语言模型(LLMs)最近展现出通过迭代优化进行自我校正响应的卓越能力,这通常被称为自我一致性或自我反思。然而,这种自我校正机制的动态可能因模型任务是开放式文本生成还是从多个预定义选项中选择最合适的响应而存在显著差异。在本文中,我们通过比较各种自然语言理解和推理任务中的性能趋势与纠错行为,对这两种范式进行了系统性研究,覆盖不同规模和系列的语言模型。我们的实验结果揭示了不同的改进模式和失败模式:开放式生成任务通常受益于重新解释和组合优化的灵活性,而多项选择任务可利用更清晰的解决方案边界,但可能受限于提供的选项。这种对比也反映了新兴智能体LLM应用面临的双重需求:高效智能体不仅要生成和优化开放式计划或解释,还要在受限的行动空间中做出可靠的离散选择。因此,我们的研究结果强调,自我校正机制的设计应考虑任务结构与输出空间的相互作用,这对LLMs的知识密集型推理和决策导向型应用均具有启示意义。

英文摘要:

Large language models have recently demonstrated remarkable abilities to self-correct their responses through iterative refinement, often referred to as self-consistency or self-reflection. However, the dynamics of this self-correction mechanism may differ substantially depending on whether the model is tasked with open-ended text generation or with selecting the most appropriate response from multiple predefined options. In this paper, we conduct a systematic investigation of these two paradigms by comparing performance trends and error-correction behaviors across various natural language understanding and reasoning tasks, covering language models of different scales and families. Our experimental results reveal distinct patterns of improvement and failure modes: \textit{While open-ended generation often benefits from the flexibility of re-interpretation and compositional refinement, multiple-choice selection can leverage clearer solution boundaries but may be limited by the provided options}. This contrast also reflects the dual demands faced by emerging agentic LLM applications: effective agents must not only generate and refine open-ended plans or explanations, but also make reliable discrete choices when operating within constrained action spaces. Our findings, therefore, highlight that the design of self-correction mechanisms should take into account the interaction between task structure and output space, with implications for both knowledge-intensive reasoning and decision-oriented applications of LLMs.

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。

↑