Search papers, labs, and topics across Lattice.
The authors introduce MedFabric, a data-centric pipeline for generating realistic, word-level fabrications in medical text, addressing limitations in existing hallucination datasets. They then develop EtHER, a modular word-level fabrication detector that uses Text2Table Decomposition, Word Masking and Filling, and Hybrid Sentence Pair Evaluation to improve factual alignment. Experiments show that MedFabric outperforms existing detectors by over 15% on word-level fabrication benchmarks, demonstrating its effectiveness in detecting subtle factual errors.
Existing hallucination detection methods are missing subtle, word-level medical errors, but a new data-centric pipeline and detector closes the gap by 15%.
Large Language Models exhibit strong reasoning and semantic understanding capabilities but often hallucinate in domains that require expert knowledge, among which fabrications, the generation of factually incorrect yet fluent statements, pose the greatest risk in medical contexts. Existing medical hallucination datasets inadequately capture fabrication phenomena due to limited fabrication coverage, stylistic disparities between human and LLM-authored texts, and distributional drift during hallucinated sample synthesis. To address this, we propose a data-centric pipeline to generate realistic and word-level fabrications that preserve syntactic and stylistic fidelity while introducing subtle factual deviations, resulting in MedFabric. Building upon this dataset, we introduce ETHER, a modular word-level fabrication detector integrating Text2Table Decomposition, Word Masking and Filling and Hybrid Sentence Pair Evaluation to enhance factual alignment. Empirical results demonstrate that MedFabric outperforms state-of-the-art detectors by over 15% on word-level fabrication benchmarks while maintaining consistent performance across structural similarities, offering a comprehensive framework for reliable and domain-specific factuality detection.
