# Expert-level Multimodal Understanding on MMMU (Accuracy) Source: [Expert-level Multimodal Understanding on MMMU (Accuracy)]() Collection index: [llms.txt]() Data updated: 2026-03-31T09:15:34.044Z Evaluation of model performance on the MMMU dataset focusing on expert-level multimodal understanding capabilities, 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 | Expert-level Multimodal Understanding | | Dataset (including any stated split) | MMMU | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 15 of 15 | | 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 | | --- | --- | --- | --- | | Qwen3-VL-8B | 60.9 | Not recorded | 691bfe976750be83fe5cf516 | | PAPO | 60.8 | Not recorded | 68fa7bd11e209131cea8ab14 | | VAPO | 60.2 | Not recorded | 68fa7bd11e209131cea8ab14 | | VL-Rethinker | 57.9 | Not recorded | 68fa7bd11e209131cea8ab14 | | SenseNova-SI | 57.6 | Not recorded | 691bfe976750be83fe5cf516 | | MM-Eureka | 55.7 | Not recorded | 68fa7bd11e209131cea8ab14 | | InternVL3-8B | 55.6 | Not recorded | 691bfe976750be83fe5cf516 | | Base Model | 52.7 | Not recorded | 68fa7bd11e209131cea8ab14 | | VST-7B-SFT | 50.9 | Not recorded | 691bfe976750be83fe5cf516 | | Bagel-7B-MoT | 50.4 | Not recorded | 691bfe976750be83fe5cf516 | | SenseNova-SI | 50.2 | Not recorded | 691bfe976750be83fe5cf516 | | SenseNova-SI | 49.4 | Not recorded | 691bfe976750be83fe5cf516 | | Cambrian-S-7B | 47.1 | Not recorded | 691bfe976750be83fe5cf516 | | SenseNova-SI | 44.4 | Not recorded | 691bfe976750be83fe5cf516 | | InternVL3-2B | 43.2 | Not recorded | 691bfe976750be83fe5cf516 | ### Qwen3-VL-8B Accuracy: 60.9. Result ID: 69cb761647f34ce924b35e32. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-source General Models"}. This record does not assert independent reproduction. ### PAPO Accuracy: 60.8. Result ID: 69cb909647f34ce924b41ba7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"7B","Decoding Strategy":"greedy decoding"}. This record does not assert independent reproduction. ### VAPO Accuracy: 60.2. Result ID: 69cb909647f34ce924b41ba9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"7B"}. This record does not assert independent reproduction. ### VL-Rethinker Accuracy: 57.9. Result ID: 69cb909647f34ce924b41ba8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"7B"}. This record does not assert independent reproduction. ### SenseNova-SI Accuracy: 57.6. Result ID: 69cb761647f34ce924b35e38. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"Qwen3-VL-8B","Model Category":"Ours"}. This record does not assert independent reproduction. ### MM-Eureka Accuracy: 55.7. Result ID: 69cb909647f34ce924b41ba6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"7B"}. This record does not assert independent reproduction. ### InternVL3-8B Accuracy: 55.6. Result ID: 69cb761647f34ce924b35e34. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-source General Models"}. This record does not assert independent reproduction. ### Base Model Accuracy: 52.7. Result ID: 69cb909647f34ce924b41ba5. [Reporting paper]() Recorded evaluation configuration: {"Model Scale":"7B"}. This record does not assert independent reproduction. ### VST-7B-SFT Accuracy: 50.9. Result ID: 69cb761647f34ce924b35e35. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Open-source SI Models"}. This record does not assert independent reproduction. ### Bagel-7B-MoT Accuracy: 50.4. Result ID: 69cb761647f34ce924b35e31. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Open-source General Models"}. This record does not assert independent reproduction. ### SenseNova-SI Accuracy: 50.2. Result ID: 69cb761647f34ce924b35e37. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"Bagel-7B-MoT","Model Category":"Ours"}. This record does not assert independent reproduction. ### SenseNova-SI Accuracy: 49.4. Result ID: 69cb761647f34ce924b35e3a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"InternVL3-8B","Model Category":"Ours"}. This record does not assert independent reproduction. ### Cambrian-S-7B Accuracy: 47.1. Result ID: 69cb761647f34ce924b35e36. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-source SI Models"}. This record does not assert independent reproduction. ### SenseNova-SI Accuracy: 44.4. Result ID: 69cb761647f34ce924b35e39. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Base Architecture":"InternVL3-2B","Model Category":"Ours"}. This record does not assert independent reproduction. ### InternVL3-2B Accuracy: 43.2. Result ID: 69cb761647f34ce924b35e33. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Open-source General Models"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.