Instruction Following and Open-ended question answering on MMAU-Pro
100AIF ScoreHuman
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| Human2026.05 | 100 | 77.56 | 77.3 | 77.9 | |
| MAPO2026.05 | 95.4 | 62 | 85.32 | 65.29 | |
| Gemini 2.5 Flash2026.05 | 95.1 | 57.39 | 67.5 | 59.2 | |
| Qwen3-Omni-InstructTraining mode=Instruction-tuned2026.05 | 89.66 | 57.84 | 87.38 | 61.84 | |
| GPT-4o Audio2026.05 | 82.5 | 53.2 | 43.2 | 52.5 | |
| Qwen3-Omni-ThinkingTraining mode=Thinking baseline2026.05 | 62.07 | 59.63 | 84.77 | 62.63 | |
| Qwen2.5-OmniModel size=7B2026.05 | 61.3 | 52.01 | 52.3 | 52.2 |