List your company

Paralinguistic Reasoning on SAKURA

92.2Animal Accuracy (Single-hop)Qwen2.5-Omni
28.34444.92261.578.078Jul 7, 2026

Evaluation results

MethodLinks
Qwen2.5-OmniParam=3B, Audio (hrs)=∼1.7M–3.3M est.2026.0792.270.664.822.627.415.4802966.134.4
QWEN2-AUDIOParam=7B, Audio (hrs)=∼520k2026.079060.68943.462.240.283.446.481.247.7
Audio Flamingo-3Param=8B, Audio (hrs)=∼100k2026.078861.666.449.470.841.293.843.679.849
ORCAParam=4B, Audio (hrs)=7,0002026.077676.287.685.863.857.294.481.680.575.2
DESTA2.5-AUDIO (8B)Param=8B, Audio (hrs)=7,0002026.0758.85889.28462.257.496.48076.769.9
DESTA2.5-AUDIO 4B BaselineParam=4B, Audio (hrs)=7,0002026.0730.834.244.2385650.89372.25648.8