Instruction Following on Speech-IFEval
93.89IF RateDeSTA2.5-Audio
Evaluation results
| Method | Links | ||
|---|---|---|---|
| DeSTA2.5-Audio2025.07 | 93.89 | 0.4 | |
| ASR+LLM2025.07 | 93.52 | — | |
| DeSTA22025.07 | 89.23 | -3.57 | |
| VIRBABackbone=Qwen2.5-Omni2025.08 | 68.2 | — | |
| VIRBABackbone=Qwen2-Audio2025.08 | 64.5 | — | |
| Step-Audio-R12025.08 | 62.1 | — | |
| Qwen2.5-Omni2025.08 | 60.7 | — | |
| Kimi-Audio2025.08 | 59.4 | — | |
| AF3Full Name=Audio Flamingo 32025.08 | 57.8 | — | |
| Single-view RL2025.08 | 55.8 | — | |
| GLM-4-Voice2025.08 | 52.3 | — | |
| Group-DPO2025.08 | 50.2 | — | |
| Qwen2-Audio-Instruct2025.07 | 47.11 | — | |
| TTS-SFT2025.08 | 44.7 | — | |
| Qwen2-Audio base2025.08 | 38.5 | — | |
| SALMONN2025.07 | 36.89 | -50.2 | |
| Qwen-Audio2025.07 | 32.98 | — | |
| LTU-AS2025.07 | 29.19 | -54.9 |