Audio Reasoning on SoundMind
69.8AccuracyStep-Audio-R1
Evaluation Results
| Method | Links | |
|---|---|---|
| Step-Audio-R12025.08 | 69.8 | |
| VIRBABackbone=Qwen2.5-Omni2025.08 | 65.5 | |
| AF3Full name=Audio Flamingo 32025.08 | 62.8 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 60.6 | |
| Kimi-Audio2025.08 | 59.3 | |
| Qwen2.5-Omni2025.08 | 54.8 | |
| Single-view RL2025.08 | 48.9 | |
| GLM-4-Voice2025.08 | 46.5 | |
| Qwen2-Audio base2025.08 | 41.2 |