Detecting Hallucination in LLMs: Tracing the Topological Signatures of Impaired Context Sharing

2026-09-17

AI Summary

研究通过分析注意力图的拓扑结构和Forman-Ricci曲率，提出了一种有效区分大语言模型中幻觉与非幻觉响应的方法，提高了幻觉检测的准确性。

Abstract

In this work, we examine the topology of information flow patterns within attention graphs to effectively distinguish hallucinated from non-hallucinated responses. We analyze the Forman-Ricci curvature to identify structural patterns indicating information bottlenecks in attention graphs. We then introduce a method that captures both semi-local and global information-flow characteristics of attention heads associated with hallucinated responses. We evaluate our approach extensively across several LLMs and established benchmarks. Empirical results demonstrate that our proposed single-pass approach provides consistent improvements over existing attention-based and multi-response baselines across two hallucination-detection benchmarks, while achieving competitive performance across diverse LLM architectures. Further analysis reveals that impaired context sharing among tokens during causal generation is strongly associated with hallucination occurrences in LLMs. In particular, hallucinated responses are consistently characterized by an over-reliance on self-attention, diffused context retrieval from earlier tokens, or information over-squashing, especially in the final transformer layer.

Problem

• Hallucination
• LLMs
• Attention Graphs
• Context Sharing
• Forman-Ricci Curvature

Innovation

• Forman-Ricci curvature
• information flow patterns
• attention graphs
• hallucination detection
• context sharing