用自评估树搜索提升大模型复杂推理能力，不依赖外部奖励。
SELT: Self-Evaluation Tree Search for LLMs with Task Decomposition
- 通过自评估重定义树搜索评分机制，结合任务分解与语义聚类。
- 在MMLU和Seal-Tools上准确率显著提升，减少幻觉与冗余推理。
- 无需微调，适用于多种复杂推理任务，通用性强。
尽管大语言模型在众多应用中取得显著成果，其在复杂推理任务中的表现常有下降。本文提出SELT（自评估树搜索），一种基于改进蒙特卡洛树搜索的框架，无需外部奖励模型即可增强大模型推理能力。通过重新定义上限置信区间评分以匹配大模型内在自评估能力，并将推理过程分解为原子子任务，同时在每个节点引入语义聚类，SELT有效平衡探索与利用，减少冗余推理路径并缓解幻觉问题。我们在知识密集型的MMLU及工具学习数据集Seal-Tools上验证该方法，结果表明相较于基线方法，SELT在答案准确率和推理鲁棒性方面均有显著提升。值得注意的是，该框架无需任务特定微调，展现出跨多样化推理任务的强大泛化能力。相关代码与结果见https://github.com/fairyshine/SELT。
原文摘要 · Abstract (English)
While Large Language Models (LLMs) have achieved remarkable success in a wide range of applications, their performance often degrades in complex reasoning tasks. In this work, we introduce SELT (Self-Evaluation LLM Tree Search), a novel framework that leverages a modified Monte Carlo Tree Search (MCTS) to enhance LLM reasoning without relying on external reward models. By redefining the Upper Confidence Bound scoring to align with intrinsic self-evaluation capabilities of LLMs and decomposing the inference process into atomic subtasks augmented with semantic clustering at each node, SELT effectively balances exploration and exploitation, reduces redundant reasoning paths, and mitigates hallucination. We validate our approach on challenging benchmarks, including the knowledge-based MMLU and the Tool Learning dataset Seal-Tools, where SELT achieves significant improvements in answer accuracy and reasoning robustness compared to baseline methods. Notably, our framework operates without task-specific fine-tuning, demonstrating strong generalizability across diverse reasoning tasks. Relevant results and code are available at https://github.com/fairyshine/SELT .
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。