Audio Reasoning on MMAU mini (test)
77.7Average ScoreOmni-R1
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Omni-R12026.04 | 77.7 | — | — | — | |
| OmniJigsawvariant=CMM2026.04 | 76.3 | — | — | — | |
| HumanOmniV22026.04 | 75.9 | — | — | — | |
| OmniJigsawvariant=SMS2026.04 | 75.8 | — | — | — | |
| AudioJigsaw2026.04 | 75.4 | — | — | — | |
| OmniJigsawvariant=JMI2026.04 | 74.5 | — | — | — | |
| Qwen3-Omni-30B2026.04 | 74.4 | — | — | — | |
| AuTAgentBackbone=GPT-4o Audio2026.02 | 67.2 | 66.97 | 61.8 | 72.97 | |
| Tools w/ Desc.Backbone=GPT-4o Audio2026.02 | 62.9 | 62.16 | 57.19 | 69.37 | |
| Vanilla (No-Tool)Backbone=GPT-4o Audio2026.02 | 57.4 | 58.86 | 56.59 | 56.76 | |
| AuTAgentBackbone=Qwen2-Audio-7B-Instruct2026.02 | 54.2 | 61.56 | 53.59 | 47.45 | |
| Random SelectionBackbone=GPT-4o Audio2026.02 | 53.6 | 57.96 | 48.5 | 54.35 | |
| All-Tools (Concat.)Backbone=GPT-4o Audio2026.02 | 51.7 | 54.35 | 46.11 | 54.65 | |
| Tools w/ Desc.Backbone=Qwen2-Audio-7B-Instruct2026.02 | 51.2 | 63.06 | 50.3 | 40.24 | |
| Vanilla (No-Tool)Backbone=Qwen2-Audio-7B-Instruct2026.02 | 50 | 56.46 | 49.4 | 44.14 | |
| Random SelectionBackbone=Qwen2-Audio-7B-Instruct2026.02 | 49.1 | 60.96 | 47.9 | 38.44 | |
| All-Tools (Concat.)Backbone=Qwen2-Audio-7B-Instruct2026.02 | 40.9 | 55.26 | 35.63 | 31.83 |