Audio Synthesis on URO-Bench
4.23NMOSSLAM-Omni
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| SLAM-OmniSize=0.5B2025.09 | 4.23 | 4.44 | |
| Mini-OmniSize=0.5B2025.09 | 4.15 | 4.42 | |
| LLaMA-OmniSize=8B2025.09 | 4.09 | 4 | |
| SpeechGPTSize=7B2025.09 | 4.04 | 3.92 | |
| Qwen2.5-3B (AR)Size=3B2025.09 | 3.96 | 4.16 | |
| VITA-AudioSize=7B2025.09 | 3.95 | 4.24 | |
| Pretrain+TtTSize=3B2025.09 | 3.9 | 4.23 | |
| TtTSize=3B2025.09 | 3.89 | 4.25 | |
| GLM-4-VoiceSize=9B2025.09 | 3.86 | 4.15 | |
| Kimi-AudioSize=7B2025.09 | 3.52 | 2.93 | |
| Qwen2.5-3B (NAR)Size=3B2025.09 | 3.47 | 2.35 | |
| MoshiSize=7B2025.09 | 3.1 | 3.05 |