Temporal Sentence Grounding on Charades-STA (R@.3, R@.5, R@.7, mIoU)
85.1R@.3DynFrame-8B
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| DynFrame-8BSize=8B2026.05 | 85.1 | 72.5 | 49.4 | 61.7 | |
| DynFrame-4BSize=4B2026.05 | 83.5 | 71 | 46.5 | 60 | |
| Video-o3Size=7B2026.05 | 83.3 | 71.9 | 49 | 60.7 | |
| VITALSize=7B2026.05 | 83.1 | 72 | 46.7 | 59.9 | |
| VideoChat-R1.5Size=7B2026.05 | 82.8 | 71.6 | 48.3 | 60.6 | |
| Qwen3-VL (Thinking)Size=8B2026.05 | 81.6 | 70.8 | 45.8 | 59.9 | |
| Qwen3-VL (Thinking)Size=4B2026.05 | 80.5 | 69.2 | 44.6 | 59 | |
| Temporal-RLTSize=7B2026.05 | 79.6 | 67.9 | 44.1 | 57 | |
| LOVE-R1Size=7B2026.05 | 74 | 41 | 14 | 44.8 | |
| Qwen2.5-VLSize=7B2026.05 | 64.7 | 43.1 | 22.8 | 43.6 | |
| Video-R1-7BSize=7B2026.05 | 63.5 | 44 | 22.5 | 43.8 | |
| LongVTSize=7B2026.05 | 41 | 25.8 | 11.7 | 27.2 |