Can I read REFIND at SemEval-2025 Task 3 - Retrieval-Augmented Factuality Hallucination Detection in Large Language Models on EtoBox?

REFIND at SemEval-2025 Task 3 - Retrieval-Augmented Factuality Hallucination Detection in Large Language Models by v56g4rpd45 is a document available to read on EtoBox.

What is REFIND at SemEval-2025 Task 3 - Retrieval-Augmented Factuality Hallucination Detection in Large Language Models about?

The document introduces REFIND, a novel framework for detecting hallucinated spans in outputs from large language models (LLMs) by leveraging retrieved documents and calculating a Context Sensitivity Ratio (CSR). REFIND significantly outperforms existing methods, achieving superior Intersection-over-Union (IoU) scores across nine languages, including low-resource settings. The framework addresses the critical challenge of hallucination detection, enhancing the reliability of LLM applications.

Author

v56g4rpd45

Language

EN