| EventLens: Event-Structure Reinforcement Learning for Video Understanding | ✓ Link | 77.17 | EventLens-8B | 2026-10-12 |
| Kwai Keye-VL 1.5 Technical Report | ✓ Link | 75.5 | Keye-VL-1.5-8B | 2025-09-01 |
| Puzzle Curriculum GRPO for Vision-Centric Reasoning | ✓ Link | 75.4 | Qwen2.5-VL-7B (Jigsaw) | 2025-12-16 |
| Qwen2.5-VL Technical Report | ✓ Link | 74.8 | Qwen2.5-VL-72B | 2025-02-19 |
| PointThinker: Point-Incentivized Parallel Thinking for Multimodal Large Language Model | | 74.7 | PointThinker-7B (extra data) | 2026-06-03 |
| VAST: Video Ability-Stratified Taxonomy for Data-Efficient Video Reasoning | | 74.2 | VideoVAST-7B (extra data) | 2026-06-03 |
| MotionAtlas: A High-Quality Dataset and Benchmark for Dense Motion Captioning | ✓ Link | 74.2 | MotionAtlas-4B | 2026-09-08 |
| InternVideo3: Agentify Foundation Models with Multimodal Contextual Reasoning | ✓ Link | 74.0 | InternVideo3-8B | 2026-06-10 |
| Taming Hallucinations: Boosting MLLMs' Video Understanding via Counterfactual Video Generation | ✓ Link | 73.5 | DNA-Train-7B | 2025-12-30 |
| Ovis2.5 Technical Report | ✓ Link | 72.6 | Ovis2.5-9B | 2025-08-15 |
| OmniJigsaw: Enhancing Omni-Modal Reasoning via Modality-Orchestrated Reordering | ✓ Link | 72.28 | VideoJigsaw | 2026-04-09 |
| Kwai Keye-VL Technical Report | ✓ Link | 71.5 | Keye-VL-8B | 2025-07-02 |
| Structured Causal Video Reasoning via Multi-Objective Alignment | | 70.64 | STEER-4B | 2026-04-06 |
| Reinforced Attention Learning | | 70.3 | On-Policy Distillation + Attention Distillation | 2026-02-04 |
| Unified Spatio-Temporal Token Scoring for Efficient Video VLMs | ✓ Link | 69.9 | Molmo2 (SigLIP2 So400M + Qwen3-4B) baseline | 2026-03-18 |
| Dynin-Omni: Omnimodal Unified Large Diffusion Language Model | ✓ Link | 68.4 | Dynin-Omni (8B) | 2026-03-09 |
| VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding | ✓ Link | 68.1 | VideoLLaMA3 7B | 2025-01-22 |
| Let VLMs Grade Their Own Thoughts: A Self-Quantification Approach to Reasoning-Aware Reward Modeling | | 65.9 | Video-RAISE-7B | 2026-06-03 |
| AnchorPrune: Relevance-Anchored Contextual Expansion for Visual Token Pruning | ✓ Link | 65.4 | LLaVA-Video-7B (AnchorPrune, 1,024 tokens) | 2026-07-08 |
| Mage-VL: An Efficient Codec-Native Streaming Multimodal Foundation Model | ✓ Link | 62.4 | Mage-VL-4B (tc16) | 2026-07-27 |
| SmolVLM: Redefining small and efficient multimodal models | ✓ Link | 53.7 | SmolVLM 2.2B | 2025-04-07 |