List your company

Speech-to-text Translation on CoVoST-2 high-resource X-to-En (test)

37.6Quality Score (Fr)XLS-R (2B) + mBART-ML50N1
30.11232.0563435.944May 2, 2022

Evaluation results

MethodLinks
XLS-R (2B) + mBART-ML50N1Decoder=mBART-ML50N1, Encoder Size=2B2022.0537.633.639.233.836.1—
XMEF-X + mBART-ML50N1Decoder=mBART-ML50N12022.0536.130.638.131.834.2—
XMEF-En + mBART-ML50N1Decoder=mBART-ML50N12022.053528.235.231.132.4—
Wav2SeqBackbone=HuBERT-large2022.0533.226.93429.731—
Wav2SeqBackbone=XLS-R (0.3B)2022.05332833.829.931.1—
XLS-R (0.3B) + mBART-ML50N1Decoder=mBART-ML50N12022.0532.926.734.128.730.6—
XLSR-53 + mBART-ML50N1Decoder=mBART-ML50N12022.0532.326.933.328.630.3—
VP-100K + mBART-ML50N1Decoder=mBART-ML50N12022.0530.423.431.125.727.7—
Maestro2022.12—————38.2
mSLAM-CTCParameters=2B2022.12—————37.8
WhisperEvaluation protocol=Zero-Shot2022.12—————36.2
XLS-RParameters=2B2022.12—————36.1
XMEF-X2022.12—————34.2