Speaker-attributed Automatic Speech Recognition on AliMeeting
20.09Word Error Rate (WER)GLSC-SDR
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| GLSC-SDR2026.03 | 20.09 | 25.43 | 5.34 | 83.26 | |
| Qwen2.5-omni-sftmode=fine-tuned2026.03 | 20.22 | 26.77 | 6.55 | 81.28 | |
| TagSpeech2026.03 | 25.42 | 33.84 | 8.42 | 81.63 | |
| Gemini-3-pro*source=converted from [18]2026.03 | 26.75 | 32.84 | 6.09 | — | |
| ViebVoice-ASR2026.03 | 27.4 | 29.33 | 1.93 | — | |
| Gemini-2.5-pro*source=converted from [18]2026.03 | 27.43 | 41.64 | 14.2 | — | |
| Qwen2.5-omni2026.03 | 29.21 | 43.64 | 14.43 | 71.28 |