List your company

Audio Understanding / Audio Reasoning on MMAR

62.2AccuracyQwen2.5-Omni-7B + CoAT
44.20848.87953.5558.221Jan 26, 2026Feb 16, 2026Mar 10, 2026Apr 1, 2026Apr 22, 2026May 14, 2026Jun 5, 2026

Evaluation results

MethodLinks
Qwen2.5-Omni-7B + CoATBackbone=Qwen2.5-Omni-7B, Method=CoAT2026.0662.2
MiMo-Audio + SICL-AT3few-shot=true2026.0161.4
MiMo-Audio + SICL-AT2few-shot=true2026.0161
Audio Flamingo 3 + CoATBackbone=Audio Flamingo 3, Method=CoAT2026.0659.6
MiMo-Audio + Vanilla SICLfew-shot=true2026.0158.2
MiMo-Audio + SICL-AT1few-shot=true2026.0157.7
Qwen2.5-Omni-7BBackbone=Qwen2.5-Omni-7B, Method=Baseline2026.0656.7
Audio Flamingo 3Backbone=Audio Flamingo 3, Method=Baseline2026.0655.7
MiMo-Audiofew-shot=false2026.0154.7
Qwen2.5-Omni + SICL-AT3few-shot=true2026.0154.5
Qwen2.5-Omni + SICL-AT2few-shot=true2026.0154.4
Qwen2.5-Omni + SICL-AT1few-shot=true2026.0154.2
Qwen2.5-Omni + Vanilla SICLfew-shot=true2026.0153.8
Qwen2-Audio + CoATBackbone=Qwen2-Audio, Method=CoAT2026.0652.6
Qwen2.5-Omni (Fine-tuned on CV-en)few-shot=false2026.0150.5
Qwen2.5-Omni (Fine-tuned on CV-en) + Vanilla SICLfew-shot=true2026.0150.4
Qwen2.5-Omnifew-shot=false2026.0149.2
Qwen2-AudioBackbone=Qwen2-Audio, Method=Baseline2026.0647.1
Qwen2.5-Omni (Fine-tuned on RSR)few-shot=false2026.0144.9