Compositional Question Answering on GQA (Exact)
73.4Exact AccuracySceneAlign
Evaluation Results
| Method | Links | |
|---|---|---|
| SceneAlignModel=LLaVA-Next-8B2026.01 | 73.4 | |
| BaseModel=LLaVA-Next-8B2026.01 | 73 | |
| SFTModel=LLaVA-Next-8B2026.01 | 73 | |
| AoTModel=LLaVA-Next-8B2026.01 | 73 | |
| LLaVA-ReasonerModel=LLaVA-Next-8B2026.01 | 72.8 | |
| BaseModel=Qwen2.5-VL-7B2026.01 | 71.8 | |
| SFTModel=Qwen2.5-VL-7B2026.01 | 71.4 | |
| SceneAlignModel=Qwen2.5-VL-7B2026.01 | 71.4 | |
| SFTModel=Qwen2.5-VL-3B2026.01 | 69.6 | |
| SceneAlignModel=Qwen2.5-VL-3B2026.01 | 69.6 | |
| BaseModel=Qwen2.5-VL-3B2026.01 | 69.4 | |
| SFTModel=Qwen3-VL-4B2026.01 | 63 | |
| SceneAlignModel=Qwen3-VL-4B2026.01 | 63 | |
| BaseModel=Qwen3-VL-4B2026.01 | 62.4 | |
| SceneAlignModel=InternVL3-8B2026.01 | 60.8 | |
| SFTModel=InternVL3-8B2026.01 | 60.4 | |
| BaseModel=InternVL3-8B2026.01 | 60.2 |