Video Temporal Grounding on DiDeMo (test)
69.2Recall@1 (IoU=0.3)EVIDENT
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| EVIDENTSource dataset=Charades-STA, LLM size=7B2026.05 | 69.2 | 55.4 | 32.3 | 48.5 | |
| Qwen2.5-VL†Source dataset=Charades-STA, LLM size=7B2026.05 | 68.2 | 53.8 | 30.3 | 47.4 | |
| EVIDENTSource dataset=QVHighlights, LLM size=7B2026.05 | 62.4 | 36 | 17.4 | 40.6 | |
| Qwen2.5-VL†Source dataset=QVHighlights, LLM size=7B2026.05 | 59.4 | 34.1 | 16.9 | 38.2 | |
| Chrono-BLIPSource dataset=Charades-STA, LLM size=4B2026.05 | 56 | 43.2 | 23.1 | 37.7 | |
| Chrono-BLIPSource dataset=QVHighlights, LLM size=4B2026.05 | 43 | 24.4 | 11.8 | 33.7 | |
| HawkEyeSource dataset=Zero-Shot, LLM size=7B2026.05 | 37.8 | 18.4 | 7.8 | 24.7 | |
| CG-DETRSource dataset=QVHighlights2026.05 | 36.6 | 21.2 | 8.9 | 24.6 | |
| EaTRSource dataset=QVHighlights2026.05 | 36 | 17.8 | 6.5 | 24.1 | |
| EaTRSource dataset=Charades-STA2026.05 | 32.7 | 20.1 | 7.8 | 21.2 | |
| CG-DETRSource dataset=Charades-STA2026.05 | 30.3 | 21.8 | 11 | 20.7 |