# Video Multimodal Understanding on MMVU Source: [Video Multimodal Understanding on MMVU]() Collection index: [llms.txt]() Data updated: 2026-06-12T08:04:16.235Z Evaluation of model performance on the MMVU dataset for Video Multimodal Understanding tasks, measured using Accuracy. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Video Multimodal Understanding | | Dataset (including any stated split) | MMVU | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 18 of 18 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | SDRL | 68.6 | Not recorded | 69ca039e0636e5bf4afc942a | | VideoRFT | 67.3 | Not recorded | 69ca039e0636e5bf4afc942a | | TW-GRPO | 65.8 | Not recorded | 69ca039e0636e5bf4afc942a | | Qwen2.5-VL-7B | 65.4 | Not recorded | 69ca039e0636e5bf4afc942a | | SDRL | 64.8 | Not recorded | 69ca039e0636e5bf4afc942a | | Video-R1 | 64.2 | Not recorded | 69ca039e0636e5bf4afc942a | | VideoChat-R1 | 64.2 | Not recorded | 69ca039e0636e5bf4afc942a | | Video-R1 | 63.8 | Not recorded | 69ca039e0636e5bf4afc942a | | VideoRFT | 63.5 | Not recorded | 69ca039e0636e5bf4afc942a | | Qwen2.5-VL-7B | 63.2 | Not recorded | 69ca039e0636e5bf4afc942a | | VideoRFT | 60.5 | Not recorded | 69ca039e0636e5bf4afc942a | | Qwen2.5-VL-7B | 59.2 | Not recorded | 69ca039e0636e5bf4afc942a | | Video-R1 | 51.3 | Not recorded | 69ca039e0636e5bf4afc942a | | LLaVA-OneVision-7B | 49.2 | Not recorded | 69ca039e0636e5bf4afc942a | | MSA-PT | 47.5 | Not recorded | 6a2b9299d33dfb23e783b069 | | Full | 45.8 | Not recorded | 6a2b9299d33dfb23e783b069 | | MSA-CPT | 45.8 | Not recorded | 6a2b9299d33dfb23e783b069 | | VideoLLaMA2 | 44.8 | Not recorded | 69ca039e0636e5bf4afc942a | ### SDRL Accuracy: 68.6. Result ID: 69ca1d9b47f34ce924b22f31. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"RL"}. This record does not assert independent reproduction. ### VideoRFT Accuracy: 67.3. Result ID: 69ca1d9b47f34ce924b22f2b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"SFT+ RL","Input frames":"16-frame"}. This record does not assert independent reproduction. ### TW-GRPO Accuracy: 65.8. Result ID: 69ca1d9b47f34ce924b22f2f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"RL"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 65.4. Result ID: 69ca1d9b47f34ce924b22f25. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"None"}. This record does not assert independent reproduction. ### SDRL Accuracy: 64.8. Result ID: 69ca1d9b47f34ce924b22f30. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"RL","Training dataset":"EventFlow"}. This record does not assert independent reproduction. ### Video-R1 Accuracy: 64.2. Result ID: 69ca1d9b47f34ce924b22f2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"SFT+ RL"}. This record does not assert independent reproduction. ### VideoChat-R1 Accuracy: 64.2. Result ID: 69ca1d9b47f34ce924b22f2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"RL"}. This record does not assert independent reproduction. ### Video-R1 Accuracy: 63.8. Result ID: 69ca1d9b47f34ce924b22f2c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"RL"}. This record does not assert independent reproduction. ### VideoRFT Accuracy: 63.5. Result ID: 69ca1d9b47f34ce924b22f2e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"RL","Input frames":"16-frame"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 63.2. Result ID: 69ca1d9b47f34ce924b22f27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"None","Chain-of-Thought (CoT)":"ours CoT"}. This record does not assert independent reproduction. ### VideoRFT Accuracy: 60.5. Result ID: 69ca1d9b47f34ce924b22f29. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"SFT","Input frames":"16-frame"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Accuracy: 59.2. Result ID: 69ca1d9b47f34ce924b22f26. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"None","Chain-of-Thought (CoT)":"video-r1 CoT"}. This record does not assert independent reproduction. ### Video-R1 Accuracy: 51.3. Result ID: 69ca1d9b47f34ce924b22f28. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Training":"SFT"}. This record does not assert independent reproduction. ### LLaVA-OneVision-7B Accuracy: 49.2. Result ID: 69ca1d9b47f34ce924b22f24. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Training":"None"}. This record does not assert independent reproduction. ### MSA-PT Accuracy: 47.5. Result ID: 6a2bbd6ffbc373592c2a06f7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Training Strategy":"from-scratch sparse pretraining"}. This record does not assert independent reproduction. ### Full Accuracy: 45.8. Result ID: 6a2bbd6ffbc373592c2a06f6. [Reporting paper]() Recorded evaluation configuration: {"Training Strategy":"Full-Attention baseline"}. This record does not assert independent reproduction. ### MSA-CPT Accuracy: 45.8. Result ID: 6a2bbd6ffbc373592c2a06f8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Training Strategy":"sparse continued pretraining"}. This record does not assert independent reproduction. ### VideoLLaMA2 Accuracy: 44.8. Result ID: 69ca1d9b47f34ce924b22f23. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Training":"None"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.