Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Analyzing and Improving Fine-grained Preference Optimization in Medical LVLMs

About

Large Vision-Language Models (LVLMs) have achieved strong performance across medical imaging tasks, yet they remain prone to factual inconsistencies, poor visual grounding, and misalignment with clinically meaningful feedback. Existing post-training alignment approaches, including Direct Preference Optimization (DPO) and its variants, face three critical limitations in the medical domain: (1) sequence-level reward signals treat clinically critical tokens identically to generic filler text; (2) reliance on static supervised fine-tuning references as preferred responses introduces an off-policy distribution shift, steering optimization toward stylistic artifacts over clinical correctness; and (3) alignment objectives lack explicit visual grounding constraints, leaving models insensitive to subtle yet diagnostically decisive pathological features. Our method leverages a bidirectional token-wise KL regularizer alongside a visual-contrastive grounding objective that pairs clean and lesion-corrupted images to penalize responses generated without adequate visual evidence. Together, these components form a fine-grained, on-policy alignment framework that constructs preference pairs by minimally editing model-generated outputs, correcting only clinically erroneous spans while preserving the original linguistic style. Extensive experiments across medical imaging tasks and clinical text generation benchmarks validate the effectiveness of our approach.

Shayan Mohammadizadehsamakosh, Pritam Sarkar, Leonid Sigal, Ali Etemad, Elham Dolatabadi• 2026

Related benchmarks

TaskDatasetResultRank
Medical Visual Question AnsweringSLAKE (test)
Closed Accuracy81.32
82
Medical Visual Question AnsweringVQA-RAD (test)
Closed Accuracy71.26
65
Showing 2 of 2 rows

Other info

Follow for update