Video Understanding on Visually Grounded (VG) question
48.2AccuracyTW-GRPO
Evaluation Results
| Method | Links | |
|---|---|---|
| TW-GRPOFrames=642026.04 | 48.2 | |
| VidGroundFrames=642026.04 | 47.9 | |
| VidGroundFrames=322026.04 | 47.6 | |
| Video-RTSFrames=642026.04 | 46.4 | |
| Video-RTSFrames=322026.04 | 46.3 | |
| Qwen2.5-VL-7BFrames=642026.04 | 46.3 | |
| TW-GRPOFrames=322026.04 | 45.9 | |
| Qwen2.5-VL-7B-SFTFrames=642026.04 | 45.8 | |
| VidGroundFrames=162026.04 | 45.2 | |
| Qwen2.5-VL-7B-SFTFrames=322026.04 | 44.6 | |
| Qwen2.5-VL-7BFrames=322026.04 | 44.4 | |
| TW-GRPOFrames=162026.04 | 44.1 | |
| Video-RTSFrames=162026.04 | 43.5 | |
| Video-R1Frames=322026.04 | 43.1 | |
| Qwen2.5-VL-7BFrames=162026.04 | 42.9 | |
| LongVILA-R1-7BFrames=322026.04 | 42.9 | |
| Video-R1Frames=642026.04 | 42.9 | |
| LongVILA-R1-7BFrames=642026.04 | 42 | |
| Video-R1Frames=162026.04 | 41.7 | |
| Qwen2.5-VL-7B-SFTFrames=162026.04 | 41.1 | |
| LongVILA-R1-7BFrames=162026.04 | 39.7 |