| Benchmark | Model | Rank | Results |
|---|---|---|---|
| multimodal-reasoning-on-mathvista | Qwen3-VL-8B (Learning from Primitive) (extra data) | – | Accuracy: 67.9 |
| multimodal-reasoning-on-mmmu | Qwen3-VL-8B (Learning from Primitive) (extra data) | – | Accuracy: 60.74 |
| visual-question-answering-vqa-on-realworldqa | Qwen3-VL-8B (Learning from Primitive) (extra data) | – | Accuracy: 71.83 |