Speech Conversation on URO-Bench Chinese Basic
93.6Understanding ScoreMPS-thkfirst
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MPS-thkfirstStrategy=Think-First, Language=Chinese2025.10 | 93.6 | 84 | 94.8 | 89.1 | |
| MPS-tbsStrategy=Think-Before-Speaking, Language=Chinese2025.10 | 92.6 | 82.4 | 93.8 | 87.8 | |
| MPS-spkfirstStrategy=Speak-First, Language=Chinese2025.10 | 92.5 | 82.5 | 93.1 | 87.6 | |
| MPS-wo/thkStrategy=without a thinking process, Language=Chinese2025.10 | 91.6 | 77.3 | 87.7 | 83.4 | |
| Step-Audio 2Language=Chinese2025.10 | 91.1 | 75.5 | 86.1 | 83.3 | |
| GPT-4o AudioLanguage=Chinese2025.10 | 89.4 | 65.5 | 85.2 | 78.6 | |
| GPT-RealtimeLanguage=Chinese2025.10 | 88.8 | 72.9 | 90.8 | 80.6 | |
| Kimi-AudioLanguage=Chinese2025.10 | 79.3 | 64.7 | 79.8 | 73.6 | |
| Qwen-OmniLanguage=Chinese2025.10 | 59.7 | 69.7 | 77.3 | 69 |