Speech Conversation on URO-Bench Chinese Pro
77.2Understanding (U)MPS-spkfirst
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MPS-spkfirstStrategy=Speak-First, Language=Chinese2025.10 | 77.2 | 84.8 | 79 | 79.9 | |
| MPS-tbsStrategy=Think-Before-Speaking, Language=Chinese2025.10 | 75.3 | 84.2 | 79.5 | 79 | |
| MPS-thkfirstStrategy=Think-First, Language=Chinese2025.10 | 75.2 | 84.2 | 85.2 | 80.5 | |
| MPS-wo/thkStrategy=without a thinking process, Language=Chinese2025.10 | 75.1 | 74.7 | 72.9 | 74.4 | |
| Step-Audio 2Language=Chinese2025.10 | 74.8 | 63.2 | 65.1 | 68.3 | |
| GPT-RealtimeLanguage=Chinese2025.10 | 72.3 | 62.6 | 74.2 | 70.6 | |
| GPT-4o AudioLanguage=Chinese2025.10 | 70.6 | 57.2 | 70.2 | 67.1 | |
| Kimi-AudioLanguage=Chinese2025.10 | 60.4 | 59.3 | 76.2 | 66 | |
| Qwen-OmniLanguage=Chinese2025.10 | 59 | 59.8 | 58.7 | 59.1 |