ResearchResearch paperMultimodal Models · Reasoning & Planning · Agents & Tool Use2 sources · Oct 13, 2026

Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence

We introduce Open-o3-Video-v2, which formulates video reasoning as a unified agentic evidencegrounded process.

Key points

  • 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.
  • 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.
  • These results show that active evidence acquisition and verifiable object-level grounding are complementary for reliable video reasoning.

Sources (2)

Extractive summary: sentences quoted from the sources.

Before this

  1. Oct 12, 2026AUGUstus: Autonomous Driving VLA Model with Unified Generation and Understanding
  2. Oct 12, 2026X-Ragent: Tool-Augmented Reinforcement Learning for Evidence-Grounded Chest X-Ray Reasoning
  3. Oct 12, 2026EventLens: Event-Structure Reinforcement Learning for Video Understanding
  4. Oct 11, 2026Pessimistic Latent Task-aware Optimization for Robust Offline Meta-Reinforcement Learning
  5. Oct 11, 2026RSIPO: Recursive Self-Improving Policy Optimization for Agentic Reinforcement Learning via Skill-Policy Co-Evolution
  6. Oct 11, 2026ANTE: Anticipate, Then Execute -- Learning Internal World Models for Coding Agents from Verified Beliefs

Related