Speech-to-Speech Question-Answering on Average (LlamaQ, TriviaQA, WebQ)
50.6AccuracyDIFFUSPEECH
Evaluation Results
| Method | Links | |
|---|---|---|
| DIFFUSPEECHType=Diff., S→S capability=true2026.01 | 50.6 | |
| MoshiType=AR, S→S capability=true2026.01 | 45.1 | |
| MinMoType=AR, S→S capability=true2026.01 | 43.1 | |
| Llama-Omni2Type=AR, S→S capability=true2026.01 | 40.6 | |
| SpiritLMType=AR, S→S capability=true2026.01 | 17.5 | |
| SpeechGPTType=AR, S→S capability=true2026.01 | 6.4 |