Proactive Sound Response on Proactive-Sound-Bench Multi-round
64.9Success Rate (Human)Audio-Interaction
Evaluation Results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Audio-InteractionCategory=Streaming Audio Language Models2026.06 | 64.9 | 65.8 | 55.7 | 69 | 61.8 | 60 | 62.8 | |
| Gemini-3-FlashCategory=Omni & Audio Language Models2026.06 | 59.6 | 47.5 | 40 | 53.1 | 56.3 | 53.3 | 50.8 | |
| MiniCPM-o-4.5Category=Omni & Audio Language Models2026.06 | 53.2 | 75.4 | 52.9 | 55.1 | 47.6 | 53.3 | 58.9 | |
| Qwen2.5-Omni-7BCategory=Omni & Audio Language Models2026.06 | 34.6 | 40.2 | 19.3 | 24.5 | 31.1 | 60 | 32.1 | |
| Qwen2.5-Omni-3BCategory=Omni & Audio Language Models2026.06 | 28.9 | 42.5 | 17.9 | 36.7 | 17.5 | 40 | 29.3 | |
| Kimi-Audio-InstructCategory=Omni & Audio Language Models2026.06 | 26.3 | 38.6 | 22.1 | 16.3 | 28.2 | 26.7 | 28.4 | |
| Step-Audio 2Category=Omni & Audio Language Models2026.06 | 5.8 | 3.4 | 0 | 6.1 | 1 | 0 | 3 |