Temporal Video Grounding on QVHighlights TimeLens (test)
84.1Recall @ IoU=0.3Gemini-2.5-Pro
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Gemini-2.5-Pro2026.02 | 84.1 | 75.9 | 61.1 | |
| Gemini-2.5-Flash2026.02 | 78.2 | 69.4 | 55 | |
| Gemini-2.0-Flash2026.02 | 76.2 | 66.4 | 48.3 | |
| Video-OPDbackbone=Qwen3-VL-8B2026.02 | 73.8 | 60.3 | 50.4 | |
| GPT-52026.02 | 72.4 | 60.4 | 46.4 | |
| OP-FKDbackbone=Qwen3-VL-8B2026.02 | 71.2 | 58.9 | 46.5 | |
| OP-RKDbackbone=Qwen3-VL-8B2026.02 | 70.7 | 57.4 | 44.5 | |
| GRPObackbone=Qwen3-VL-8B2026.02 | 69.8 | 53 | 41.5 | |
| GPT-4o2026.02 | 69 | 54.8 | 38.5 | |
| Time-R1-7B2026.02 | 65.8 | 51.5 | 36.1 | |
| VideoChat-R1.5-7Breproduced=true2026.02 | 62.2 | 44.5 | 28.3 | |
| MiMo-VL-7B2026.02 | 57.1 | 42.6 | 28.4 | |
| TVG-R1-7Breproduced=true2026.02 | 55.8 | 41.2 | 28 | |
| Qwen2.5-VL-7Breproduced=true2026.02 | 55 | 41.7 | 29.3 | |
| Qwen3-VL-8B-Instructreproduced=true2026.02 | 46.6 | 38.2 | 29.5 | |
| VideoChat-Flash-7B2026.02 | 45.2 | 30.6 | 16.7 | |
| VideoChat-R1-7B2026.02 | 29.3 | 19.1 | 9.4 |