List your company

Multimodal Speech Understanding on MMSU All (test)

75.2AccuracyStep-Audio-R1
9.295226.405143.51560.6249May 26, 2026

Evaluation results

MethodLinks
Step-Audio-R12026.0575.2
Gemini 2.52026.0571.88
Qwen3-Omni2026.0571.28
GPT-4o Audio2026.0564.16
Kimi-Audio2026.0562.59
MiMo-Audio2026.0559.92
Qwen2.5-Omni2026.0557.16
Qwen2-Audio + PCLM + DPOPCLM=true, Training=DPO2026.0555.43
Qwen2-Audio + PCLMPCLM=true2026.0553.3
AF32026.0551.43
Qwen2-Audio2026.0550.82
Qwen2-Audio + SFT w/o PCLMTraining=SFT, PCLM=false2026.0550.74
AF3 + PCLM + DPOPCLM=true, Training=DPO2026.0550.62
Qwen2-Audio + SFT + DPO w/o PCLMTraining=SFT + DPO, PCLM=false2026.0550.36
AF3 + PCLMPCLM=true2026.0550.04
AF3 + SFT w/o PCLMTraining=SFT, PCLM=false2026.0549.98
VITA-Audio2026.0547.71
AF3 + SFT + DPO w/o PCLMTraining=SFT + DPO, PCLM=false2026.0547.1
AF22026.0541.35
SALMONN2026.0511.83