Leveraging Graph Structures to Detect Hallucinations in Large Language Models

TL;DR

This paper introduces a graph-based approach using Graph Attention Networks to detect hallucinations in large language models by analyzing the structure of their latent space, improving trustworthiness of generated content.

Contribution

The paper presents a novel method leveraging graph structures and attention mechanisms to identify hallucinations in language model outputs, demonstrating effectiveness without search-based methods.

Findings

Latent space structure differentiates hallucinated from non-hallucinated outputs

Graph Attention Networks can learn and generalize this structure

Contrastive learning enhances robustness of hallucination detection

Abstract

Large language models are extensively applied across a wide range of tasks, such as customer support, content creation, educational tutoring, and providing financial guidance. However, a well-known drawback is their predisposition to generate hallucinations. This damages the trustworthiness of the information these models provide, impacting decision-making and user confidence. We propose a method to detect hallucinations by looking at the structure of the latent space and finding associations within hallucinated and non-hallucinated generations. We create a graph structure that connects generations that lie closely in the embedding space. Moreover, we employ a Graph Attention Network which utilizes message passing to aggregate information from neighboring nodes and assigns varying degrees of importance to each neighbor based on their relevance. Our findings show that 1) there exists a structure in the latent space that differentiates between hallucinated and non-hallucinated generations, 2) Graph Attention Networks can learn this structure and generalize it to unseen generations, and 3) the robustness of our method is enhanced when incorporating contrastive learning. When evaluated against evidence-based benchmarks, our model performs similarly without access to search-based methods.