Video Temporal Grounding on TACoS (R1@0.3, R1@0.5, R1@0.7, mIoU)
41.97R@0.5MASRA
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MASRAFeatures=SF+C2026.05 | 41.97 | 54.61 | 26.17 | 38.84 | |
| RGTRFeatures=SF+C2026.05 | 40.31 | 53.04 | 24.32 | 37.44 | |
| KDAFeatures=SF+C2026.05 | 40.13 | — | 24.34 | — | |
| CG-DETRFeatures=SF+C2026.05 | 39.61 | 52.23 | 22.23 | 36.48 | |
| TaROBase model=Qwen3-VL-8B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 36.5 | 52.3 | 22.5 | — | |
| UVCOMFeatures=SF+C2026.05 | 36.39 | — | 23.32 | — | |
| UniVTGFeatures=SF+C2026.05 | 34.97 | 51.44 | 17.35 | 33.6 | |
| Qwen3-VL-8B-InstructBase model=Qwen3-VL-8B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 30.5 | 38.6 | 20.4 | — | |
| 2D-TANFeatures=SF+C2026.05 | 27.99 | 40.01 | 12.92 | 27.22 | |
| M-DETRFeatures=SF+C2026.05 | 24.67 | 37.97 | 11.97 | 25.49 | |
| VSLNetFeatures=SF+C2026.05 | 23.54 | 35.54 | 13.15 | 24.99 | |
| TaROBase model=Qwen2.5-VL-7B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 22.5 | 37.9 | 8.6 | — | |
| Time-R1Base model=Qwen2.5-VL-7B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 21.5 | 36.8 | 7.9 | — | |
| VideoChat-R1.5Base model=Qwen2.5-VL-7B-Instruct, Evaluation Protocol=Zero-shot2026.06 | 16.2 | 28.6 | 6.17 | — |