| VideoLoop: Looped Working Memory Against Semantic Thrashing in Long-Form Video Agents | ✓ Link | 88.8 | VideoLoop (Gemini 3.1 Pro) | 2026-09-29 |
| Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity | | 86.9 | Seed2.0 Pro | 2026-06-30 |
| HopChain: Multi-Hop Data Synthesis for Generalizable Vision-Language Reasoning | | 80.0 | Qwen3.5-397B-A17B | 2026-03-17 |
| Less from More: Reinforcing Sparse Video Reasoning from Dense References | ✓ Link | 72.7 | SAVER-4B (0.1 fps) (extra data) | 2026-10-07 |
| SAGE: Training Smart Any-Horizon Agents for Long Video Reasoning with Reinforcement Learning | ✓ Link | 68.1 | SAGE-Flash (SAGE-MM: Qwen3-VL-8B-Instruct [+SFT] [+RL]) | 2025-12-15 |
| OneThinker: All-in-one Reasoning Model for Image and Video | ✓ Link | 66.2 | OneThinker-8B | 2025-12-02 |
| VideoAuto-R1: Video Auto Reasoning via Thinking Once, Answering Twice | ✓ Link | 65.0 | VideoAuto-R1 (Qwen3-VL-8B) | 2026-01-08 |
| VideoKR: Towards Knowledge- and Reasoning-Intensive Video Understanding | ✓ Link | 63.0 | VideoKR (SFT + RL) on Qwen3-VL-8B-Instruct (Frames=128) | 2026-06-03 |
| Open-o3-Video-v2: Seek, Ground, and Reason with Verifiable Spatio-Temporal Evidence | ✓ Link | 63.0 | Open-o3-Video-v2-8B | 2026-10-13 |
| Enhancing Video Vision Language Model with Hippocampal Sensing | | 62.7 | HippoVLM-8B | 2026-06-03 |
| ResAdapt: Adaptive Resolution for Efficient Multimodal Reasoning | ✓ Link | 59.6 | ResAdapt (Qwen3-VL-8B, 128 frames) | 2026-03-30 |
| Insight-V++: Towards Advanced Long-Chain Visual Reasoning with Multimodal Large Language Models | | 57.6 | Insight-V++ (Self-Evolving, Qwen2.5-VL, 7B) | 2026-03-18 |
| AdaTooler-V: Adaptive Tool-Use for Images and Videos | ✓ Link | 56.8 | AdaTooler-V-7B | 2025-12-18 |
| VisionCoach: Reinforcing Grounded Video Reasoning via Visual-Perception Prompting | ✓ Link | 54.4 | VisionCoach-7B | 2026-03-15 |
| Let VLMs Grade Their Own Thoughts: A Self-Quantification Approach to Reasoning-Aware Reward Modeling | | 53.6 | Video-RAISE-7B | 2026-06-03 |
| PointThinker: Point-Incentivized Parallel Thinking for Multimodal Large Language Model | | 53.0 | PointThinker-7B (extra data) | 2026-06-03 |
| From Pixels to Words -- Towards Native One-Vision Models at Scale | ✓ Link | 51.6 | NEO-ov (8B) | 2026-05-27 |
| VAST: Video Ability-Stratified Taxonomy for Data-Efficient Video Reasoning | | 50.6 | VideoVAST-7B (extra data) | 2026-06-03 |
| Watch Before You Answer: Learning from Visually Grounded Post-Training | ✓ Link | 49.4 | VidGround (64 frames) | 2026-04-06 |
| Puzzle Curriculum GRPO for Vision-Centric Reasoning | ✓ Link | 42.66 | Qwen2.5-VL-7B (Mix) | 2025-12-16 |
| OmniStream: Mastering Perception, Reconstruction and Action in Continuous Streams | ✓ Link | 40.0 | OmniStream-7B | 2026-03-12 |