Hallucination Detection-Guided Preference Optimization for Clinical Summarization
For healthcare applications requiring reliable summarization, this work provides an automated method to improve factual faithfulness without sacrificing quality.
The paper introduces hallucination detection-guided iterative refinement and preference learning to reduce hallucinations in clinical summarization, achieving 24% and 48% hallucination reduction for Llama-3.1-8B-Instruct while preserving fluency and relevance.
Large language models (LLMs) have shown promise on summarization tasks, but they often produce hallucinations, which are unsupported or incorrect statements that limit their reliability in specialized healthcare applications. We introduce \itermodelfull (\itermodel), an inference-time method that leverages hallucination detectors to guide iterative summary revisions toward factual corrections. Building on this, we propose \itermodel for Preference Learning (\model), which converts detector-guided refinement trajectories into preference pairs for model finetuning. Extensive experiments show that our methods substantially reduce hallucinations for Llama and Gemma models in summarizing real-world clinical notes from \MimicIV. For example, \itermodel reduces 24\% and \model reduces 48\% hallucinations in Llama-3.1-8B-Instruct. Importantly, both methods preserve summary fluency, coherence, and relevance according to human expert and LLM-Jury evaluations. Together, these results demonstrate that detection-informed refinement and preference learning offer an automated solution for improving factual faithfulness in clinical summarization.