Speech-to-Speech on URO-Bench Pro Task
55.47Understanding ScoreGLM-4-Voice
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| GLM-4-VoiceSize=9B2025.09 | 55.47 | 51.89 | 61.3 | |
| Kimi-AudioSize=7B2025.09 | 53.25 | 41.44 | 50.17 | |
| LLaMA-OmniSize=8B2025.09 | 34.66 | 51.51 | 43.91 | |
| SLAM-OmniSize=0.5B2025.09 | 34.49 | 27.39 | 40.23 | |
| Pretrain+TtTSize=3B2025.09 | 32.38 | 43.76 | 46.1 | |
| VITA-AudioSize=7B2025.09 | 32.36 | 54.77 | 45.81 | |
| MoshiSize=7B2025.09 | 26.38 | 21.06 | 33.93 | |
| Mini-OmniSize=0.5B2025.09 | 23.51 | 33.09 | 33.46 | |
| TtTSize=3B2025.09 | 23.37 | 33.78 | 34.82 | |
| SpeechGPTSize=7B2025.09 | 19.03 | 14.29 | 28.88 | |
| Qwen2.5-3B (AR)Size=3B2025.09 | 16.32 | 34.99 | 25.9 | |
| Qwen2.5-3B (NAR)Size=3B2025.09 | 12.59 | 13.7 | 25.64 |