Multimodal large language models (MLLMs) are rapidly advancing clinical diagnosis, yet their adaptation pipelines remain anchored to accuracy-based objective…
机构:哈佛
来源:arXiv 2609.40361 | AI4Papers 论文推荐平台