HuggingFace Daily Papers·· 1 天前
Open-o3-Video-v2：以可验证时空证据进行寻找、定位与推理
Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence
AI 导读
Open-o3-Video-v2 将视频推理重构为智能体式证据落地流程，让模型主动识别缺失信息、获取额外观测并定位支持推理的时刻与目标，再写入轨迹后继续推理或作答。
正文
Abstract:Reliable video reasoning requires models to produce correct answers and expose supporting visual evidence. Open-o3-Video enables verifiable reasoning by embedding timestamps and object-level bounding boxes into reasoning traces, but is restricted to frames from an initial sparse sampler. We introduce Open-o3-Video-v2, which formulates video reasoning as a unified agentic evidencegrounded process. The model identifies missing information, acquires additional observations, grounds the moments and objects supporting its reasoning, and writes this evidence into the trajectory before continuing or answering. To support this behavior, we construct STGR-v2-CoT135K, including 52K multi-turn agentic grounded reasoning trajectories, and STGR-v2-RL-13K for reinforcement learning. We further propose Clue-Seek GRPO (CS-GRPO), which tracks how tool calls change accumulated evidence, rewards clue coverage and step-wise gains, and penalizes excessive, repetitive, or uninformative retrieval. On V-STaR, Open-o3-Video-v2-4B achieves 40.4% mAM and 56.3% mLGM, outperforming its Qwen3-VL-4B backbone by 8.5% and 12.6%, respectively. It also delivers consistent gains across general video understanding, long-video reasoning, and temporal grounding benchmarks. These results show that active evidence acquisition and verifiable object-level grounding are complementary for reliable video reasoning. The code and models are released at this https URL.
| Comments: | Extension of Open-o3-Video (https://arxiv.org/abs/2510.20579). V2 extends single-pass grounded reasoning to agentic multi-turn evidence seeking and grounding, enabling models to actively acquire and reason over missing visual evidence. We further introduce STGR-v2-CoT-135K and CS-GRPO to improve evidence acquisition and grounded reasoning |
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2610.17656 [cs.CV] |
| (or arXiv:2610.17656v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2610.17656 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Jiahao Meng [view email]
[v1]
Tue, 13 Oct 2026 03:36:33 UTC (19,097 KB)
来源：HuggingFace Daily Papers · arxiv.org