COMPASS benchmark standardizes speech-to-speech translation evaluation

Arxiv CS.CL (NLP/language models)

COMPASS provides a unified protocol for evaluating speech-to-speech translation models, addressing inconsistent metric reporting. The benchmark aims to enable direct comparisons across studies.

speech-to-speechbenchmarks

Arxiv CS.CL (NLP/language models)·Alkis KoudounasHayato FutamiQuentin JodeletOsamu TakeShinji WatanabeEmiru Tsunoo

8 days ago