Speaker-aware ASR on AISHELL-1 & AISHELL-2 augmented with VoxCeleb & MUSAN (15 dB SNR)
1.3WERUAF-30B-A3B
Evaluation results
| Method | Links | |
|---|---|---|
| UAF-30B-A3BReference prompt=true, Enrollment duration=5-second2026.04 | 1.3 | |
| Kimi-Audio-7BReference prompt=true, Enrollment duration=5-second2026.04 | 2.07 | |
| Qwen3-Omni-30B-A3BReference prompt=true, Enrollment duration=5-second2026.04 | 2.16 | |
| Qwen2.5-Omni-7BReference prompt=true, Enrollment duration=5-second2026.04 | 67.79 |