发表机构
Télécom Paris; Institut Polytechnique de Paris; BNP Paribas（巴黎电信学院; 巴黎理工学院; 法国巴黎银行）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
针对大语言模型长推理链存在的步骤缺陷问题，提出LCoT-GV图框架结合图注意力网络验证推理链，构建新数据集且方法具竞争力。
AI 中文摘要
大型推理模型会生成长推理链（Long Chains-of-Thought，LCoTs），这类模型会先将问题拆解为多个较小的推理步骤，再得出结论。不过，即便最终答案正确，这些步骤也常存在矛盾、无依据的推断或无关步骤。我们提出长推理链图验证器（Long Chain-of-Thought Graph Verifier，LCoT-GV），这是一种基于图的框架，可将LCoTs表示为推理图：图中每个节点代表一个推理步骤，边编码语义与逻辑关系。随后，我们训练图注意力网络，从推理图中预测推理链的正确性。我们从多个推理基准构建了一个面向图的验证数据集，用于各领域问答任务。实验结果表明，我们的方法与最相似的方法相比具有竞争力。
英文摘要
Large Reasoning Models produce Long Chains-of-Thought (LCoTs) which involve breaking down the problem into smaller reasoning steps before reaching the conclusion. However, these steps often contain contradictions, unsupported inferences, or irrelevant steps, even when the final answer is correct. We propose Long Chain-of-Thought Graph Verifier (LCoT-GV), a graph-based framework that represents LCoTs as reasoning graphs. Each node in the graph represents a reasoning step and the edges encode semantic and logical relations. A Graph Attention Network is then trained to predict chain-of-thought correctness from the reasoning graph. We construct a new graph-oriented verification dataset from multiple reasoning benchmarks for question answering in various domains. The results show that our method is competitive with the most similar approaches.
Comments6 pages without references, 2 tables, 1 algorithm, 1 figure
