Multi-modal Understanding on SEED-Bench (test)
52.38TU (Text Understanding)Position
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| PositionCategory=SSL4RL, Base Model=Qwen2.5-VL-3B2025.10 | 52.38 | 70.69 | 73.56 | 72.75 | 62.89 | 72.51 | 64.62 | 55.25 | 64.2 | 69.77 | |
| JigsawCategory=SSL4RL, Base Model=Qwen2.5-VL-3B2025.10 | 48.81 | 69.79 | 70.3 | 71.65 | 63.92 | 70.19 | 62.68 | 53.12 | 62.93 | 67.67 | |
| RotationCategory=SSL4RL, Base Model=Qwen2.5-VL-3B2025.10 | 45.24 | 73.41 | 73.65 | 72.8 | 67.01 | 71.03 | 61.76 | 54.03 | 64.12 | 69.1 | |
| Qwen2.5-VL-3BCategory=Base2025.10 | 41.67 | 53.78 | 60.35 | 63.24 | 64.95 | 62.87 | 58.79 | 51.6 | 60.52 | 60.83 | |
| ContrastiveCategory=SSL4RL, Base Model=Qwen2.5-VL-3B2025.10 | 28.57 | 67.07 | 67.1 | 68.38 | 64.95 | 61.22 | 63.7 | 51.75 | 54.03 | 61.9 |