Instruction Following on CommonEval (ComE)
4.08GPT-scoreBaichuan-Audio
Evaluation Results
| Method | Links | |
|---|---|---|
| Baichuan-AudioFull Duplex=false2026.07 | 4.08 | |
| Kimi-AudioFull Duplex=false2026.07 | 3.97 | |
| DuplexPOFull Duplex=true2026.07 | 3.74 | |
| SFT BaselineFull Duplex=true2026.07 | 3.48 | |
| Freeze-OmniFull Duplex=true2026.07 | 3.46 | |
| Qwen2-AudioFull Duplex=false2026.07 | 3.43 | |
| GLM-4-VoiceFull Duplex=false2026.07 | 3.42 | |
| SALM-DuplexFull Duplex=true2026.07 | 2.5 | |
| MoshiFull Duplex=true2026.07 | 1.6 |