Maintaining Informative Coherence: Migrating Hallucinations in Large Language Models via Absorbing Markov Chains

TL;DR

This paper introduces a decoding method using absorbing Markov chains to reduce hallucinations in large language models, improving the fidelity and coherence of generated text without extra training.

Contribution

It presents a novel decoding strategy that quantifies information significance and loss, enhancing LLM reliability by considering all possible token paths during generation.

Findings

Outperforms existing methods in hallucination mitigation on multiple datasets

Improves information fidelity and coherence in generated texts

Does not require additional training or external data

Abstract

Large Language Models (LLMs) are powerful tools for text generation, translation, and summarization, but they often suffer from hallucinations-instances where they fail to maintain the fidelity and coherence of contextual information during decoding, sometimes overlooking critical details due to their sampling strategies and inherent biases from training data and fine-tuning discrepancies. These hallucinations can propagate through the web, affecting the trustworthiness of information disseminated online. To address this issue, we propose a novel decoding strategy that leverages absorbing Markov chains to quantify the significance of contextual information and measure the extent of information loss during generation. By considering all possible paths from the first to the last token, our approach enhances the reliability of model outputs without requiring additional training or external data. Evaluations on datasets including TruthfulQA, FACTOR, and HaluEval highlight the superior performance of our method in mitigating hallucinations, underscoring the necessity of ensuring accurate information flow in web-based applications.