AIB_IWSLT25_Offline_en-zh_unconstrained_contrastive1
We employ cascade speech translation system consisting of Whisper-large and Qwen-2.5-7B-instruct, we perform sliding window ASR on the input audio then perform segment-level translation based on the transcription from Whisper.
From\To | de | zh |
---|---|---|
en | 0.495 |