让医疗影像诊断AI学会自我反思，避免误用工具
MIRA: Medical Image Reflection for Agentic Diagnosis
- 通过动态调用图像处理和网络搜索工具，自主验证证据相关性
- 在9个基准上平均得分64.73，工具使用正确率提升至73.8%
- 适合需要高可靠性医疗AI决策的临床辅助场景
医疗视觉代理可使用工具检查图像并检索外部知识，但盲目使用工具可能引入噪声或误导性证据。可靠诊断不仅需获取额外观察，还需验证工具操作是否必要及所得证据是否支持当前假设。我们提出MIRA（Medical Image Reflection for Agentic Diagnosis），一种用于自主证据搜索与反思验证的医疗视觉诊断框架。MIRA动态调用图像处理操作（如缩放、定位、指认、旋转、测量）及网络搜索，并评估所获证据的相关性与一致性。通过两阶段训练策略构建：首先，基于工具增强的蒙特卡洛树搜索数据引擎探索多样诊断假设，联合验证视觉定位准确性和语义一致性，生成监督微调轨迹；其次，通过强化学习进一步优化决策：将失败案例提炼为候选原则，仅保留能提升保留回滚奖励的原则。在九个医学视觉推理基准上，MIRA平均得分为64.73，比其Qwen3-VL-8B基线提升7.44分。有用工具使用判断率从56.2%提升至73.8%，有害判断率从8.9%降至1.6%。定性分析表明，MIRA可重新审视证据、纠正过早结论，并自适应调整工具使用策略。
原文摘要 · Abstract (English)
Medical visual agents can use tools to inspect images and retrieve external knowledge, but indiscriminate tool use may introduce noisy or misleading evidence. Reliable diagnosis therefore requires not only acquiring additional observations, but also verifying whether tool actions are necessary and whether the resulting evidence supports the current hypothesis. We introduce MIRA (Medical Image Reflection for Agentic Diagnosis), a medical visual diagnostic framework for autonomous evidence search and reflective verification. MIRA dynamically invokes image-processing operations, including zooming, grounding, pointing, rotation, and measurement, as well as web search, while evaluating the relevance and consistency of the acquired evidence. We develop MIRA through a two-stage training strategy. First, a tool-augmented Monte Carlo Tree Search data engine explores diverse diagnostic hypotheses and jointly verifies visual grounding accuracy and semantic consistency to construct supervised fine-tuning trajectories. Second, reinforcement learning further improves decision-making through online reflective principle evolution: failure cases are distilled into candidate principles, and only principles that improve held-out rollout rewards are retained. Across nine medical visual reasoning benchmarks, MIRA achieves an average score of 64.73, improving its Qwen3-VL-8B backbone by 7.44 points. It also increases useful tool-use judgments from 56.2% to 73.8% and reduces harmful judgments from 8.9% to 1.6%. Qualitative analyses show that MIRA can re-examine evidence, correct premature conclusions, and adapt its tool-use strategy. Project page: https://MIRA-VL.github.io/
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。