Complex Scene Reasoning on EMMA mini
25.25ScoreSceneAlign
Evaluation Results
| Method | Links | |
|---|---|---|
| SceneAlignModel=Qwen3-VL-4B2026.01 | 25.25 | |
| SFTModel=Qwen3-VL-4B2026.01 | 24 | |
| SceneAlignModel=Qwen2.5-VL-3B2026.01 | 23.25 | |
| SceneAlignModel=InternVL3-8B2026.01 | 23.25 | |
| SFTModel=Qwen2.5-VL-3B2026.01 | 22.75 | |
| BaseModel=Qwen3-VL-4B2026.01 | 22.75 | |
| SceneAlignModel=Qwen2.5-VL-7B2026.01 | 22.75 | |
| SFTModel=InternVL3-8B2026.01 | 22.25 | |
| SFTModel=Qwen2.5-VL-7B2026.01 | 21 | |
| BaseModel=InternVL3-8B2026.01 | 21 | |
| BaseModel=Qwen2.5-VL-3B2026.01 | 20 | |
| SceneAlignModel=LLaVA-Next-8B2026.01 | 19.5 | |
| BaseModel=Qwen2.5-VL-7B2026.01 | 17.75 | |
| SFTModel=LLaVA-Next-8B2026.01 | 17 | |
| AoTModel=LLaVA-Next-8B2026.01 | 16.75 | |
| BaseModel=LLaVA-Next-8B2026.01 | 15.75 | |
| LLaVA-ReasonerModel=LLaVA-Next-8B2026.01 | 14 |