Video understanding on VideoZeroBench
21Level 3 Accuracy (pass@1)Codex-GPT-5.4 Low
Evaluation results
| Method | Links | |
|---|---|---|
| Codex-GPT-5.4 LowSetting=Strong Baseline2026.05 | 21 | |
| Code-X (SFT)Model=Qwen3.6-27B, Thinking budget=4K2026.05 | 18.4 | |
| Code-X (RL)Model=Qwen3.5-9B, Thinking budget=4K2026.05 | 15.8 | |
| Baseline (direct image)Model=Qwen3.6-27B, Thinking budget=4K2026.05 | 13 | |
| Code-X (SFT)Model=Qwen3.5-9B, Thinking budget=4K2026.05 | 9.5 | |
| Direct MLLMModel=Qwen3.6-27B, Thinking budget=4K2026.05 | 7.4 | |
| Direct MLLMModel=Qwen3.5-9B, Thinking budget=4K2026.05 | 7.2 | |
| OmniAtlas-Omni-30B-A3BSetting=Strong Baseline2026.05 | 6 | |
| Baseline (direct image)Model=Qwen3.5-9B, Thinking budget=4K2026.05 | 5.9 |