Complex Scene Reasoning on SEEDBench
77.03Accuracy (All)SceneAlign
Evaluation results
| Method | Links | ||
|---|---|---|---|
| SceneAlignModel=Qwen3-VL-4B2026.01 | 77.03 | 79.89 | |
| SceneAlignModel=Qwen2.5-VL-7B2026.01 | 76.54 | 79.12 | |
| SFTModel=Qwen2.5-VL-7B2026.01 | 75.17 | 78.46 | |
| BaseModel=Qwen3-VL-4B2026.01 | 74.8 | 76.93 | |
| SFTModel=Qwen3-VL-4B2026.01 | 74.55 | 77.12 | |
| BaseModel=Qwen2.5-VL-7B2026.01 | 74.07 | 77.43 | |
| SceneAlignModel=InternVL3-8B2026.01 | 73.45 | 76.22 | |
| SFTModel=InternVL3-8B2026.01 | 72.1 | 75.05 | |
| SceneAlignModel=Qwen2.5-VL-3B2026.01 | 72.09 | 75.6 | |
| BaseModel=Qwen2.5-VL-3B2026.01 | 70.94 | 74.88 | |
| BaseModel=InternVL3-8B2026.01 | 70.72 | 73.67 | |
| SFTModel=Qwen2.5-VL-3B2026.01 | 70.3 | 74.83 | |
| SceneAlignModel=LLaVA-Next-8B2026.01 | 59.8 | 74.6 | |
| LLaVA-ReasonerModel=LLaVA-Next-8B2026.01 | 58.12 | 73.52 | |
| SFTModel=LLaVA-Next-8B2026.01 | 57.71 | 72.8 | |
| BaseModel=LLaVA-Next-8B2026.01 | 57.58 | 72.73 | |
| AoTModel=LLaVA-Next-8B2026.01 | 57.58 | 72.74 |