Spoken Instruction Following on OOD Spoken Instruction Suite (test)
96.6ConsistencyVIRBA
Evaluation Results
| Method | Links | |
|---|---|---|
| VIRBABackbone=Qwen2.5-Omni2025.08 | 96.6 | |
| VIRBABackbone=Qwen2-Audio2025.08 | 95.8 | |
| Step-Audio-R12025.08 | 93.9 | |
| Qwen2.5-Omni2025.08 | 93.2 | |
| Single-view RL2025.08 | 90.6 | |
| TTS-SFT2025.08 | 88.4 | |
| Qwen2-Audio base2025.08 | 81.2 |