Multimodal Reasoning on Text-Audio-Vision Benchmark Level 4
61Pass@1 AccuracyOmni-AutoThink
Evaluation results
| Method | Links | ||
|---|---|---|---|
| Omni-AutoThinkBackbone=Qwen2.5-Omni-7B, Training=SFT + RL2025.12 | 61 | 30 | |
| Qwen3-OmniParameters=30B2025.12 | 51 | 0 | |
| Qwen2.5-OmniParameters=7B2025.12 | 32 | 0 |