Audio Question Answering on Bio-acoustic QA
81.25AccuracySpeech-Hands
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Speech-Handsmode=parallel agentic, sampling=majority sampling2026.01 | 81.25 | 77.37 | |
| Qwen2.5-Omnisetting=SFT with official training data2026.01 | 78.13 | 63.13 | |
| Qwen2.5-Omnisetting=GRPO with official training data2026.01 | 78.09 | 65.54 | |
| Qwen2.5-Omnisetting=GER: => AF3 (cascaded agentic)2026.01 | 76.29 | 68.93 | |
| AudioFlamingo 3 (AF3)2026.01 | 71.88 | 74.49 | |
| Speech-Handsmode=parallel agentic, setting=SFT with official training data2026.01 | 67.86 | 75.75 | |
| Qwen2.5-Omnisetting=GRPO with external audio data2026.01 | 62.32 | 75.1 | |
| Qwen2.5-Omni2026.01 | 47.32 | 57.87 | |
| Gemini-2-Flash2026.01 | 42.03 | 56.61 |