LCoT-GV: Graph Attention Networks for Verifying Long Reasoning Chains in Large Language Models
2026-08-31 • Computation and Language
Computation and LanguageArtificial Intelligence
AI summaryⓘ
The authors noticed that big AI models solve problems by breaking them into many small steps, but some steps can be wrong or confusing even if the final answer is right. They created a method called Long Chain-of-Thought Graph Verifier (LCoT-GV) that turns these steps into a graph, with each step as a point connected by lines showing how they relate. Then, they use a special AI called a Graph Attention Network to check if the whole chain of reasoning makes sense. They also made a new dataset to test their method, and their results are about as good as other similar techniques.
Large Reasoning ModelsChain-of-ThoughtGraph Attention NetworkReasoning GraphQuestion AnsweringSemantic RelationsLogical RelationsVerification Dataset
Authors
Bérénice Jaulmes, Mehwish Alam
Abstract
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.