List your company

Audio Question Answering on MMAU-style Complex QA

85.7AccuracySpeech-Hands
58.857665.826372.79579.7637Jan 14, 2026

Evaluation results

MethodLinks
Speech-Handsmode=parallel agentic, sampling=majority sampling2026.0185.777.37
Speech-Handsmode=parallel agentic, setting=SFT with official training data2026.0183.3475.75
Qwen2.5-Omnisetting=GRPO with external audio data2026.0182.1575.1
AudioFlamingo 3 (AF3)2026.0181.2674.49
Qwen2.5-Omnisetting=GRPO with official training data2026.0179.1265.54
Qwen2.5-Omnisetting=GER: => AF3 (cascaded agentic)2026.0177.4868.93
Qwen2.5-Omnisetting=SFT with official training data2026.0176.6163.13
Gemini-2-Flash2026.0159.8956.61
Qwen2.5-Omni2026.0159.8957.87