SurgCoT：通过思维链基准推进手术视频中的时空推理
SurgCoT: Advancing Spatiotemporal Reasoning in Surgical Videos through a Chain-of-Thought Benchmark
提供了一个带结构化思维链标注（问题-选项-知识-线索-答案）的手术视频时空推理基准，可用于诊断多模态大模型在因果动作排序、线索-动作对齐、可供性映射等细粒度时序环节上的失效点，并作为逐步推理评测协议的模板迁移到其他视频理解任务。
原文依据 we introduce SurgCoT, a unified benchmark for evaluating chain-of-thought (CoT) reasoning in MLLMs across 7 surgical specialties and 35 diverse procedures
摘要译文
由模型翻译 · 术语统一 · 请对照原文核对关键结论手术视频上的细粒度时空推理至关重要，然而多模态大语言模型（MLLMs）在该领域的能力仍基本未被探索。为填补这一空白，我们提出 SurgCoT，一个统一的基准，用于评估 MLLMs 在 7 个手术专科、35 种不同术式中的思维链（CoT）推理能力。SurgCoT 通过一个结构化 CoT 框架与一套密集标注协议（问题-选项-知识-线索-答案）评估五个核心推理维度：因果动作排序、线索-动作对齐、可供性映射、微过渡定位与异常起始追踪，其中知识字段提供必要的背景上下文，线索字段提供确定的时空证据。对 10 个领先 MLLMs 的评估表明：1）商业模型优于开源模型与医学专用变体；2）手术 CoT 推理存在显著差距；3）SurgCoT 能够实现有效评估并增强渐进式时空推理。SurgCoT 提供了一个可复现的测试平台，以缩小 MLLM 能力与临床推理需求之间的差距。代码与数据：https://github.com/CVI-SZU/SurgCoT。
英文原文摘要 1186 字符
Fine-grained spatiotemporal reasoning on surgical videos is critical, yet the capabilities of Multi-modal Large Language Models (MLLMs) in this domain remain largely unexplored. To bridge this gap, we introduce SurgCoT, a unified benchmark for evaluating chain-of-thought (CoT) reasoning in MLLMs across 7 surgical specialties and 35 diverse procedures. SurgCoT assesses five core reasoning dimensions: Causal Action Ordering, Cue-Action Alignment, Affordance Mapping, Micro-Transition Localization, and Anomaly Onset Tracking, through a structured CoT framework with an intensive annotation protocol (Question-Option-Knowledge-Clue-Answer), where the Knowledge field provides essential background context and Clue provides definitive spatiotemporal evidence. Evaluation of 10 leading MLLMs shows: 1) commercial models outperform open-source and medical-specialized variants; 2) significant gaps exist in surgical CoT reasoning; 3) SurgCoT enables effective evaluation and enhances progressive spatiotemporal reasoning. SurgCoT provides a reproducible testbed to narrow the gap between MLLM capabilities and clinical reasoning demands. Code and data: https://github.com/CVI-SZU/SurgCoT.