MedCBR：在能力边界对齐医学视觉语言模型的熵自适应评分标准方法
Aligning Medical Vision Language Models at the Capability Boundary with Entropy Adaptive Rubric Grounding
AI 导读
一项研究提出 MedCBR，一个面向医学视觉语言模型的两阶段对齐框架，先用基于难度的边界偏好优化（DBPO）聚焦模型能力边界，再用熵门控的评分标准强化学习（EGPO）奖励答案正确性、解剖学定位和推理保真度。在六个医学基准上平均准确率达 61.2%，在鉴别诊断等困难任务上提升最大，消融显示两阶段组合收益超过各自改进之和。该论文已被 ACM Multimedia Asia 2026 接收。
正文 · 原文
Abstract:Medical diagnostic reasoning requires verifiable links between visual evidence and clinical knowledge, yet current medical vision language models are aligned with objectives that reward the wrong behavior. Uniform supervised fine tuning lets easy cases dominate updates while clinically critical boundary cases saturate prematurely. Sparse rewards based only on final-answer correctness cannot assess anatomical grounding or clinical logic, inviting reward hacking through answer shortcuts. We propose MedCBR, a two-stage alignment framework coupling boundary-aware preference optimization with rubric-grounded reinforcement learning. First, we construct counterfactual boundary preferences at the model's capability frontier using empirical pass rates and preference pairs that share reasoning prefixes but diverge at decisive diagnostic branches. Difficulty Weighted Boundary Preference Optimization (DBPO) concentrates learning on this calibrated boundary region. Second, we distill reusable clinical rubrics from boundary trajectories and activate them according to prediction entropy. Entropy Gated Grounded Policy Optimization (EGPO) rewards answer correctness, anatomical grounding, and reasoning fidelity while suppressing rubric noise on reliable samples. A reference policy anchored to the boundary-aligned model couples population-level boundary priors with instance-level uncertainty, making diagnostic reasoning more accurate and clinically grounded. Experiments on six medical benchmarks show consistent improvements over strong general, specialized, and reinforcement learning baselines, reaching 61.2% average accuracy with the largest gains on challenging differential diagnosis tasks. Ablations show super-additive gains: combining the two stages exceeds the sum of their individual improvements.
| Comments: | Accepted at ACM Multimedia Asia 2026 (MMAsia 2026). 10 pages, 4 figures |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2610.39835 [cs.CV] |
| (or arXiv:2610.39835v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.39835 arXiv-issued DOI via DataCite |
Submission history
From: Chunzheng Zhu [view email]
[v1]
Thu, 29 Oct 2026 11:15:12 UTC (2,078 KB)
来源：arXiv cs.AI · arxiv.org