Common Voice 15 EN Evaluation

[PREPONDERANCE]

I tested MOSS-Audio under common voice 15 en subset and found some of the results confusing.

I do not specify any generation configurations though. What generation configuration do you recommend to avoid this situaion?

[Benioh]

Try temperature=1.0, top_k=50, and simplify the prompt to just Transcribe the audio.