Substitution Speech Editing on Ming-Freeform-Audio-Edit-Benchmark basic
4.13WER (%)Ours
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| OursGRPO=true2026.01 | 4.13 | 0.78 | 3.15 | 4.05 | |
| OursGRPO=false2026.01 | 4.4 | 0.78 | 3.12 | 3.94 | |
| FluentSpeech2026.01 | 4.66 | 0.51 | 2.92 | 3.48 | |
| Ming-UniAudio2026.01 | 8.99 | 0.78 | — | — | |
| VoiceCraft2026.01 | 11.98 | 0.58 | 3.01 | 3.6 |