{
  "id": 1134,
  "url": "https://arxiv.org/abs/2606.12590v1",
  "title": "Analyzing and Improving Fine-grained Preference Optimization in Medical LVLMs",
  "summary": "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) r",
  "authors": "Shayan Mohammadizadehsamakosh, Pritam Sarkar, Leonid Sigal, Ali Etemad, Elham Dolatabadi",
  "category": "research",
  "topics": "safety-alignment,healthcare",
  "orgs": null,
  "regions": null,
  "published_at": "2026-06-10T18:35:36.000Z",
  "fetched_at": "2026-07-14T14:15:03.615Z",
  "source_slug": "arxiv-ethics",
  "source_name": "arXiv",
  "source_homepage": "https://arxiv.org",
  "ethics_ai_record_url": "https://ethics.ai/record/1134",
  "original_url": "https://arxiv.org/abs/2606.12590v1",
  "evidence_status": "source-only",
  "attribution": "via ethics.ai"
}