REFIND at SemEval-2025 Task 3: Retrieval-Augmented Factuality Hallucination Detection in Large Language Models

TL;DR

REFIND is a retrieval-augmented framework that detects hallucinated spans in large language model outputs by leveraging retrieved documents and a novel sensitivity metric, improving reliability across multiple languages.

Contribution

The paper introduces REFIND, a new retrieval-augmented method with the Context Sensitivity Ratio for effective hallucination detection in LLM outputs across diverse languages.

Findings

Outperformed baseline models in hallucination detection accuracy.

Demonstrated robustness across nine languages, including low-resource settings.

Achieved superior IoU scores in identifying hallucinated spans.

Abstract

Hallucinations in large language model (LLM) outputs severely limit their reliability in knowledge-intensive tasks such as question answering. To address this challenge, we introduce REFIND (Retrieval-augmented Factuality hallucINation Detection), a novel framework that detects hallucinated spans within LLM outputs by directly leveraging retrieved documents. As part of the REFIND, we propose the Context Sensitivity Ratio (CSR), a novel metric that quantifies the sensitivity of LLM outputs to retrieved evidence. This innovative approach enables REFIND to efficiently and accurately detect hallucinations, setting it apart from existing methods. In the evaluation, REFIND demonstrated robustness across nine languages, including low-resource settings, and significantly outperformed baseline models, achieving superior IoU scores in identifying hallucinated spans. This work highlights the effectiveness of quantifying context sensitivity for hallucination detection, thereby paving the way for more reliable and trustworthy LLM applications across diverse languages. Our code is available at https://github.com/oneonlee/REFIND.