Video-Language Modeling on 5 Video benchmark suite macro-average
77.4PASS@1 AccuracyQwen3.5-27B
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Qwen3.5-27BRole=Teacher2026.06 | 77.4 | — | |
| Qwen3.5-9BTraining Stage=ZPPO2026.06 | 74.2 | 0.4 | |
| Qwen3.5-9BTraining Stage=Base2026.06 | 73.8 | — | |
| Qwen3.5-4BTraining Stage=ZPPO2026.06 | 71.8 | 0.3 | |
| Qwen3.5-4BTraining Stage=Base2026.06 | 71.5 | — | |
| Qwen3.5-2BTraining Stage=ZPPO2026.06 | 63.2 | 2.6 | |
| Qwen3.5-2BTraining Stage=Base2026.06 | 60.6 | — | |
| Qwen3.5-0.8BTraining Stage=ZPPO2026.06 | 52.8 | 4.5 | |
| Qwen3.5-0.8BTraining Stage=Base2026.06 | 48.3 | — |