Hallucination Detection in Large Language Models Using Diversion Decoding

2026-07-11

AI Summary

This work addresses the challenge of hallucination in large language models, which undermines their reliability in text generation. The authors propose a novel "divergence decoding" approach that introduces adversarial perturbations during decoding to actively probe the model’s resistance to alternative answers. From these perturbations, uncertainty features are extracted and used to train a lightweight classifier for hallucination detection. By integrating uncertainty quantification directly into the generation process, the method significantly reduces computational overhead while enhancing both detection performance and robustness. Experimental results demonstrate that the proposed technique outperforms existing methods across multiple benchmarks, offering a scalable and efficient solution for mitigating hallucinations in generative language models.

Abstract

Large language models (LLMs) have emerged as a powerful tool for retrieving knowledge through seamless, human-like interactions. Despite their advanced text generation capabilities, LLMs exhibit hallucination tendencies, where they generate factually incorrect statements and fabricate knowledge, undermining their reliability and trustworthiness. Multiple studies have explored methods to evaluate LLM uncertainty and detect hallucinations. However, existing approaches are often probabilistic and computationally expensive, limiting their practical applicability. In this paper, we introduce diversion decoding, a novel method for developing an LLM uncertainty heuristic by actively challenging model-generated responses during the decoding phase. Through diversion decoding, we extract features that capture the LLM's resistance to produce alternative answers and utilize these features to train a machine-learning model to develop a heuristic measure of the LLM's uncertainty. Our experimental results demonstrate that diversion decoding outperforms existing methods with significantly lower computational complexity, making it an efficient and robust solution for evaluating hallucination detection.

Problem

• hallucination
• large language models
• uncertainty
• factuality
• reliability

Innovation

• diversion decoding
• hallucination detection
• large language models
• uncertainty estimation
• efficient inference