Fact-Controlled Diagnosis of Hallucinations in Medical Text Summarization

Hallucinations in large language models (LLMs) during summarization of patient-clinician dialogues pose significant risks to patient care and clinical decision-making. However, the phenomenon remains understudied in the clinical ___domain, with uncertainty surrounding the applicability of general-___domain hallucination detectors. The rarity and randomness of hallucinations further complicate their investigation. In this paper, we conduct an evaluation of hallucination detection methods in the medical ___domain, and construct two datasets for the purpose: A fact-controlled Leave-N-out dataset -- generated by systematically removing facts from source dialogues to induce hallucinated content in summaries; and a natural hallucination dataset -- arising organically during LLM-based medical summarization. We show that general-___domain detectors struggle to detect clinical hallucinations, and that performance on fact-controlled hallucinations does not reliably predict effectiveness on natural hallucinations. We then develop fact-based approaches that count hallucinations, offering explainability not available with existing methods. Notably, our LLM-based detectors, which we developed using fact-controlled hallucinations, generalize well to detecting real-world clinical hallucinations. This research contributes a suite of specialized metrics supported by expert-annotated datasets to advance faithful clinical summarization systems.