Instructional Temporal Answer Grounding in Video on CrossTask (test)
55.89R@0.3CACR
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| CACRFamily=LVLM, Training Setting=supervised2026.06 | 55.89 | 41.77 | 22.4 | 40.19 | |
| MutualSLFamily=VLP, Training Setting=supervised2026.06 | 49.59 | 32.32 | 17.04 | 35.33 | |
| VTPSLFamily=VLP, Training Setting=supervised2026.06 | 42.94 | 36.6 | 25.48 | 31.69 | |
| Time-R1Family=LVLM, Training Setting=supervised2026.06 | 27.39 | 13.85 | 4.82 | 18.97 | |
| TimeZeroFamily=LVLM, Training Setting=zero-shot2026.06 | 27.15 | 14.75 | 4.46 | 18.57 | |
| TimescopeFamily=LVLM, Training Setting=supervised2026.06 | 26.85 | 15.48 | 5.73 | 19.7 | |
| MoRSFamily=VLP, Training Setting=supervised2026.06 | 23.66 | 18.83 | 10.96 | 16.92 | |
| VTG-R1Family=LVLM, Training Setting=supervised2026.06 | 23.64 | 9.27 | 2.42 | 15.03 | |
| GPT-3.5Family=LLM, Training Setting=zero-shot2026.06 | 22.85 | 10.15 | 3.97 | 14.48 | |
| TFVTGFamily=LVLM, Training Setting=zero-shot2026.06 | 22.16 | 10.97 | 3.45 | 14.14 | |
| VSLNetFamily=VLP, Training Setting=supervised2026.06 | 19.04 | 12.76 | 8.32 | 13.22 | |
| Qwen2.5-VL-7B-InstructFamily=LVLM, Training Setting=zero-shot2026.06 | 6.52 | 3.44 | 1.9 | 6.05 | |
| Random Mode2026.06 | 3.57 | 0.54 | 0.2 | 2.79 |