Spoken Instruction Following on Accent OOD (test)
0.689Win Rate (LC)VIRBA
Evaluation Results
| Method | Links | |
|---|---|---|
| VIRBABackbone=Qwen2.5-Omni2025.08 | 0.689 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 0.657 | |
| Step-Audio-R12025.08 | 0.627 | |
| Qwen2.5-Omni2025.08 | 0.61 | |
| Single-view RL2025.08 | 0.565 | |
| TTS-SFT2025.08 | 0.528 | |
| Qwen2-Audio base2025.08 | 0.415 |