Speech-to-speech translation on CVSS En-Fr (test)
17.54ASR BLEUCMLM + DIFFNORM + CG
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| CMLM + DIFFNORM + CGModel Category=Non-autoregressive, Decoding iterations=15, Classifier-free guidance (w)=0.52024.05 | 17.54 | — | — | |
| CMLM + DIFFNORMModel Category=Non-autoregressive, Decoding iterations=152024.05 | 17.27 | — | — | |
| ConformerModel Category=Autoregressive2024.05 | 17.07 | — | — | |
| CMLM + BiPModel Category=Non-autoregressive, Decoding iterations=152024.05 | 16.97 | — | — | |
| CMLM + CGModel Category=Non-autoregressive, Decoding iterations=15, Classifier-free guidance (w)=0.52024.05 | 16.89 | — | — | |
| Norm TransformerModel Category=Autoregressive2024.05 | 15.93 | — | — | |
| CMLMModel Category=Non-autoregressive, Decoding iterations=152024.05 | 15.62 | — | — | |
| TransformerModel Category=Autoregressive2024.05 | 15.28 | — | — |