Audio Reasoning (Generation) on SpeechR Gen
64.2AccuracyStep-Audio-R1
Evaluation results
| Method | Links | |
|---|---|---|
| Step-Audio-R12025.08 | 64.2 | |
| VIRBABackbone=Qwen2.5-Omni2025.08 | 61.3 | |
| AF3Full name=Audio Flamingo 32025.08 | 58 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 56.9 | |
| Kimi-Audio2025.08 | 55 | |
| Qwen2.5-Omni2025.08 | 50.1 | |
| Single-view RL2025.08 | 46.2 | |
| GLM-4-Voice2025.08 | 44.2 | |
| Qwen2-Audio base2025.08 | 39 |