360度思考：全景视频推理的基准测试与自举
Think in 360{\deg}: Benchmarking and Bootstrapping Panoramic Video Reasoning
Lidong Lu · Lidong Zou · Xinyi Mao · Jiaxin Lv · Zhenquan Li · Liangrui Dong · et al.
中文摘要
全景视频捕捉相机周围所有方向，但相关证据可能出现在不同方向和不同时间。理解它需要模型跨观察球面和时间连接观察。我们推出 PanoVideoBench，首个手动标注的全景视频理解综合基准。其 1,000 个问题涵盖五个类别和 15 种问题类型，并带有目标搜索和全景能力的诊断标注。对 11 个专有、开源和全景专用模型的评估表明，跨视角跟踪和球面-时间证据整合是其最弱的能力。为增强这些能力，我们构建 PanoVideoCoT，首个基于真实全景视频的推理数据集，通过球面-时间接地合成，将验证过的全景观察和成功的智能体探索转换为 13,465 个 QA-CoT 对。从通用视频 MLLM 出发，PanoVideoReasoner 通过全景 CoT SFT 和证据特权在线策略自蒸馏从该语料库学习，这是 Vision-OPD 的 OPSD 的全景实例化，其中教师使用验证过的透视视图引导学生生成的轨迹。它在 PanoVideoBench 上达到 55.7%，比骨干模型提高 12.5 分，比之前 SOTA 提高 2.6。增益集中在基准所暴露的相同空间、跨视角和时间能力上，表明接地球面-时间监督可以使通用视频 MLLM 具备全景推理能力。
关键要点
- 01PanoVideoBench 是首个手动标注的全景视频理解基准，含 1,000 个问题、5 类 15 种题型。
- 02评估 11 个模型发现跨视角跟踪与球面-时间证据整合是最弱能力。
- 03PanoVideoCoT 通过球面-时间接地合成构建 13,465 个 QA-CoT 对。
- 04PanoVideoReasoner 结合全景 CoT SFT 与证据特权在线策略自蒸馏，达 55.7%。
- 05性能较骨干提升 12.5 分、较前 SOTA 提升 2.6 分，增益集中于空间、跨视角和时间能力。
解读
尚无解读。
原始英文摘要
arXiv:2610.42082v1 Announce Type: new Abstract: A panoramic video captures all directions around the camera, but relevant evidence can appear in different directions and at different times. Understanding it requires models to connect observations across the viewing sphere and time. We introduce PanoVideoBench, the first manually annotated benchmark for comprehensive panoramic video understanding. Its 1,000 questions span five categories and 15 question types, with diagnostic annotations for target search and panoramic capabilities. Evaluation of 11 proprietary, open-source, and panoramic-specialized models identifies cross-view tracking and sphere-time evidence integration as their weakest capabilities. To strengthen these capabilities, we construct PanoVideoCoT, the first reasoning dataset grounded in real panoramic video, through sphere-time-grounded synthesis that converts verified panoramic observations and successful agentic explorations into 13,465 QA-CoT pairs. Starting from a general video MLLM, PanoVideoReasoner learns from this corpus through Panoramic CoT SFT and Evidence-Privileged On-Policy Self-Distillation, a panoramic instantiation of Vision-OPD's OPSD in which the teacher uses verified perspective views to guide student-generated trajectories. It reaches 55.7% on PanoVideoBench, improving the backbone by 12.5 points and the previous SOTA by 2.6. The gains concentrate on the same spatial, cross-view, and temporal capabilities exposed by the benchmark, showing that grounded sphere-time supervision can equip a general video MLLM with panoramic reasoning.