Speech-to-Speech Question-Answering on TriviaQA
57.3AccuracyVocalNet
Evaluation results
| Method | Links | |
|---|---|---|
| VocalNetModel Parameters=8B2026.03 | 57.3 | |
| Qwen2.5-Omni-7BModel Parameters=7B2026.03 | 56.3 | |
| Kimi-AudioModality=S2T2026.04 | 50.49 | |
| SPEECH-OMNI-LITEModel Parameters=8B2026.03 | 50.4 | |
| FreezeEmpathModality=S2T2026.04 | 49.71 | |
| Fun-Audio-Chat-8BModality=S2T2026.04 | 49.02 | |
| Baichuan-Omni-1.5Model Parameters=7B2026.03 | 47.28 | |
| DIFFUSPEECHType=Diff., S→S capability=true2026.01 | 45.4 | |
| GLM-4-VoiceModel Parameters=9B2026.03 | 44.84 | |
| LLaMA-Omni2Model Parameters=7B2026.03 | 44.84 | |
| MoshiType=AR, S→S capability=true2026.01 | 42.1 | |
| ELLSAmode=S2S2025.10 | 41.7 | |
| MinMoType=AR, S→S capability=true2026.01 | 40.1 | |
| LLaMA-Omni2-7BModality=S2T2026.04 | 39.94 | |
| Step-Audio2-MiniModality=S2T2026.04 | 38.87 | |
| OpenS2SModality=S2T2026.04 | 36.82 | |
| Llama-Omni2Type=AR, S→S capability=true2026.01 | 35.3 | |
| Freeze-OmniModel Parameters=7B2026.03 | 32.36 | |
| Freeze-Omnimode=S2S2025.10 | 28.5 | |
| Moshimode=S2S2025.10 | 16.7 | |
| SpeechGPTType=AR, S→S capability=true2026.01 | 14.8 | |
| SpiritLMType=AR, S→S capability=true2026.01 | 6.5 |