Video Understanding on TempBench
70.2Temporal AccuracyGemini-3.1-Pro
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| Gemini-3.1-Pro2026.04 | 70.2 | 69 | 71.3 | 61 | 62.5 | |
| VTHM-MoE2026.04 | 50.7 | 54.5 | 52.3 | 52.7 | 53.1 | |
| BaselineCoT=true, SFT=true2026.04 | 42.7 | 40 | 44.4 | 33.7 | 39.6 | |
| Qwen3-VL-235B2026.04 | 39.7 | 40.8 | 46.6 | 35.3 | 41.6 | |
| Qwen3-VL-30B2026.04 | 38.5 | 36.7 | 44.7 | 31.5 | 38.3 | |
| Internvl3.5-VL-241B2026.04 | 37.3 | 38.1 | 45.6 | 34.2 | 40.5 | |
| BaselineCoT=true2026.04 | 36 | 32.2 | 37.1 | 28.6 | 32.7 | |
| BaselineSFT=true2026.04 | 35.1 | 33.5 | 30.1 | 28.5 | 34.8 | |
| Baseline2026.04 | 23.9 | 23.6 | 28.5 | 22.4 | 24.2 |