Structure-dependent Inference on MMMU Reasoning
66.67AccuracySceneAlign
Evaluation results
| Method | Links | |
|---|---|---|
| SceneAlignModel=Qwen3-VL-4B2026.01 | 66.67 | |
| SFTModel=Qwen3-VL-4B2026.01 | 66.33 | |
| BaseModel=Qwen3-VL-4B2026.01 | 64.26 | |
| SFTModel=InternVL3-8B2026.01 | 60.6 | |
| BaseModel=InternVL3-8B2026.01 | 60.51 | |
| SceneAlignModel=InternVL3-8B2026.01 | 60.37 | |
| SceneAlignModel=Qwen2.5-VL-7B2026.01 | 52.83 | |
| SFTModel=Qwen2.5-VL-7B2026.01 | 51.53 | |
| BaseModel=Qwen2.5-VL-7B2026.01 | 51.11 | |
| SceneAlignModel=Qwen2.5-VL-3B2026.01 | 50.7 | |
| SFTModel=Qwen2.5-VL-3B2026.01 | 48.09 | |
| BaseModel=Qwen2.5-VL-3B2026.01 | 47.11 | |
| SceneAlignModel=LLaVA-Next-8B2026.01 | 42.85 | |
| LLaVA-ReasonerModel=LLaVA-Next-8B2026.01 | 42.44 | |
| SFTModel=LLaVA-Next-8B2026.01 | 41.2 | |
| AoTModel=LLaVA-Next-8B2026.01 | 40.89 | |
| BaseModel=LLaVA-Next-8B2026.01 | 40.67 |