PathoJev：共享视觉上下文的高效病理决策模型
原标题：PathoJev: Efficient Pathology Decisions from Shared Visual Context
AI 摘要
研究团队提出病理模型 PathoJev，基于 System One 范式，对整张切片和 ROI 直接预测答案概率，而非逐题生成文本。模型对每张切片只编码一次，多个问题共享视觉上下文并批量评估，同时复用混合骨干网络的循环状态以降低计算量。在三个全切片基准上，9B 和 27B 版本平均准确率分别为 77.2% 和 78.1%，超过 SlideChat 5.4 和 6.3 个百分点；WSI-Bench 上 9B 模型比 SlideChat 快 9.1 倍，八题工作负载下共享上下文带来 6.9 倍和 5.9 倍加速。
正文节选
PathoJev: Efficient Pathology Decisions from Shared Visual Context Abstract Pathology workflows require multiple visual judgments about the same slide, from identifying tumor type to assessing grade and invasion. Generating and parsing a separate textual response for each structured question adds latency between visual evidence and usable decisions. Building on the System One paradigm, we introduce PathoJev, a pathology model that directly predicts answer probabilities for whole-slide and region