Vision-Only Question Answering on VisOnlyQA
79.3AccuracyQwen 2.5VL
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen 2.5VLBest alpha=0.44, Strategy=Best Pairwise2026.05 | 79.3 | |
| Gemma 3Best alpha=0.47, Strategy=Best Pairwise2026.05 | 77.44 | |
| Qwen3-VL-8BTraining Strategy=Staged2026.05 | 64 | |
| GLM 4.1VBest alpha=0.22, Strategy=Best Pairwise2026.05 | 63.72 | |
| Qwen3-VL-8BTraining Strategy=Base2026.05 | 61.78 | |
| Qwen3-VL-8BTraining Strategy=Merged2026.05 | 61.56 | |
| InternVL3.5-8BTraining Strategy=Staged2026.05 | 52 | |
| InternVL3.5-8BTraining Strategy=Merged2026.05 | 50.89 | |
| Qwen2.5-VL-7BTraining Strategy=Base2026.05 | 49.56 | |
| Qwen2.5-VL-7BTraining Strategy=Staged2026.05 | 48.67 | |
| Qwen2.5-VL-7BTraining Strategy=Merged2026.05 | 47.95 | |
| InternVL3-8BTraining Strategy=Staged2026.05 | 46.22 | |
| InternVL3-8BTraining Strategy=Base2026.05 | 41.56 | |
| InternVL3-8BTraining Strategy=Merged2026.05 | 40.22 | |
| InternVL3.5-8BTraining Strategy=Base2026.05 | 36.89 |