用图结构分析推理路径，提升大模型判断自身答案可信度的能力。
All Roads Lead to Rome: Graph-Based Confidence Estimation for Large Language Model Reasoning
- 将推理过程建模为有向图，利用节点中心性等图属性评估置信度。
- 在三个推理数据集上验证，显著提升置信度估计准确性。
- 无需训练，适用于多种大模型，适合需要可靠判断的场景。
置信度估计对大语言模型的可靠部署至关重要。现有方法多针对事实性问答任务，难以泛化至推理任务。为此，我们提出一套无需训练、基于图结构的置信度估计方法，专为推理任务设计。该方法将推理路径建模为有向图，通过节点中心性、路径汇聚度和路径权重等图属性来估计置信度。在两个大模型与三个推理数据集上的实验表明，该方法不仅提升了置信度估计性能，还在两个下游任务中实现了更好的表现。
原文摘要 · Abstract (English)
Confidence estimation is essential for the reliable deployment of large language models (LLMs). Existing methods are primarily designed for factual QA tasks and often fail to generalize to reasoning tasks. To address this gap, we propose a set of training-free, graph-based confidence estimation methods tailored to reasoning tasks. Our approach models reasoning paths as directed graphs and estimates confidence by exploiting graph properties such as centrality, path convergence, and path weighting. Experiments with two LLMs on three reasoning datasets demonstrate improved confidence estimation and enhanced performance on two downstream tasks.
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。