Spoken Question Answering on AlpacaEval
3.88WERLLaMA-Omni2
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LLaMA-Omni2Parameters=7B2026.03 | 3.88 | 4.43 | — | — | |
| Qwen2.5-Omni-7BParameters=7B2026.03 | 5.54 | 4.2 | — | — | |
| VocalNetParameters=8B2026.03 | 7.43 | 4.49 | — | — | |
| SPEECH-OMNI-LITEParameters=8B2026.03 | 23.77 | 3.73 | — | — | |
| GLM-4-VoiceParameters=9B2026.03 | 29.44 | 3.86 | — | — | |
| Freeze-OmniParameters=7B2026.03 | 49.47 | 4.14 | — | — | |
| Baichuan-Omni-1.5Parameters=7B2026.03 | 86.53 | 4.3 | — | — | |
| CF-DuplexBackbone LLM=Qwen3-1.7B, Full-Duplex=true2026.05 | — | — | 3.94 | 4.16 | |
| Freeze-OmniBackbone LLM=Qwen2-7B-Instruct, Full-Duplex=true2026.05 | — | — | 2.46 | 3.9 | |
| GLM-4-VoiceBackbone LLM=GLM-4-9B-Base, Full-Duplex=false2026.05 | — | — | 3.58 | 3.82 | |
| MoshiBackbone LLM=Helium (7B), Full-Duplex=true2026.05 | — | — | 1.76 | 1.84 | |
| SALM-DuplexBackbone LLM=TinyLlama-1.1B-chat, Full-Duplex=true2026.05 | — | — | 2.99 | — | |
| XA-DuplexBackbone LLM=Qwen3-1.7B, Full-Duplex=true2026.05 | — | — | 3.87 | 4.04 |