List your company

Multiple-choice audio understanding on MMAU mini (test)

78.4Average AccuracySpectCount (Audio Flamingo 3)
14.12830.81447.564.186Apr 16, 2026Apr 24, 2026May 2, 2026May 11, 2026May 19, 2026May 27, 2026Jun 5, 2026

Evaluation results

MethodLinks
SpectCount (Audio Flamingo 3)Size=8.3B, Trained Parameters=26.2M, Setting=SpectCount2026.0678.483.1877.5474.47
Audio Flamingo 3Size=8.3B, Setting=Base (reproduced)2026.0673.981.0871.8668.77
Audio Flamingo 3Size=8.3B, Setting=Base (reported)2026.0673.379.5873.9566.37
Qwen2.5-OmniDecoding=TCD2026.0473.2797169.7
Qwen2.5-OmniDecoding=Greedy2026.0471.578.165.970.6
Qwen2.5-OmniDecoding=AAD, alpha=0.52026.047178.16867
Gemini 2.0 Flash2026.0470.571.265.375.1
Qwen2.5-OmniDecoding=AAD, alpha=1.02026.0470.475.168.667.6
Qwen2-Audio-Inst.Decoding=TCD2026.0463.866.16362.5
SpectCount (Qwen2-Audio-Instruct)Size=8.4B, Trained Parameters=25.9M, Setting=SpectCount2026.0663.670.5758.3861.86
GPT-4o Audio2026.0462.564.656.366.7
Qwen2-Audio-Inst.Decoding=Greedy2026.0462.365.161.760
Qwen2-Audio-InstructSize=8.4B, Setting=Base (reported)2026.0659.667.2756.2955.26
Qwen2-Audio-InstructSize=8.4B, Setting=Base (reproduced)2026.0658.266.6757.1950.75
Global MergeKeep Ratio=0.752026.0556.4———
LTBM (local)Keep Ratio=0.752026.0555.6———
BaselineKeep Ratio=0.752026.0555.4———
BaselineKeep Ratio=0.52026.0555.4———
BaselineKeep Ratio=0.252026.0555.4———
GPT-4o mini Audio2026.045350.839.269.1
Global MergeKeep Ratio=0.52026.0553———
LTBM (local)Keep Ratio=0.52026.0552.3———
Segmentwise Top-KKeep Ratio=0.752026.0551———
UniAvgKeep Ratio=0.52026.0550.2———
Global Top-KKeep Ratio=0.752026.0549.3———
Global MergeKeep Ratio=0.252026.0547.7———
LTBM (local)Keep Ratio=0.252026.0547.2———
Segmentwise Top-KKeep Ratio=0.52026.0547.2———
Global Top-KKeep Ratio=0.52026.0546.6———
Global Top-KKeep Ratio=0.252026.0543.3———
Mini-OmniDecoding=TCD2026.0442.848.734.445.4
Segmentwise Top-KKeep Ratio=0.252026.0542.7———
Mini-OmniDecoding=Greedy2026.0441.246.633.843.5
UniAvgKeep Ratio=0.252026.0539.5———
SALMONN2026.0434.941.137.126.4
GAMA-IT2026.0422.830.926.710.8
GAMA2026.0420.831.817.712.9
LTU2026.0417.420.41615.9
Audio Flamingo Chat2026.0416.625.217.76.9