I2-Med：视觉引导动态Logits校准的可解释医疗推理
原标题：I$^2$-Med: Interpretable Medical Inference through Visual-Guided Dynamic Logits Calibration
AI 摘要
该论文提出 I2-Med 框架，用于提升医疗视觉语言模型的可解释性与可靠性。I2-Med 结合基于 GRPO 的结构化推理训练，以及无需额外训练的推理阶段视觉引导机制 MedCalib-IF：它利用 BiomedCLIP 评估候选词与医学图像的一致性，并动态校准 logits，从而抑制视觉幻觉。在 OmniMedVQA 基准（覆盖八种成像模态和五类临床问题）上的实验表明，该方法能生成更符合临床、更忠于图像的推理链，并减少视觉幻觉。
正文节选
11email: {phd2101241003, me230003071}@iiti.ac.in, akshatm430@gmail.com, {phd2301241003, mtanveer}@iiti.ac.in I2-Med: Interpretable Medical Inference through Visual-Guided Dynamic Logits Calibration Abstract Medical vision language models have demonstrated strong potential for medical image understanding and reasoning. However, their interpretability and reliability remain limited because medical vision language tasks require precise visual grounding and clinically coherent explanations. This is