Metraipolis › archive › dc_7445752e8108

DIAL-S2ST: Learning Latency Control for Simultaneous Speech-to-Speech Translation — arXiv:2610.32214

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.32214
License (as archived)
CC0-1.0
Archived
2026-10-26T19:53:33.092Z
SHA-256 of the archived text
7d77bcf5e3d9186c6b900cb204c62188b26f5e990a123f7164558fd433da2f1b
Stable URL of this page
https://metraipolis.com/documento/dc_7445752e8108

How to cite

DIAL-S2ST: Learning Latency Control for Simultaneous Speech-to-Speech Translation — arXiv:2610.32214 — arXiv (en), https://arxiv.org/abs/2610.32214 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7445752e8108 (Metraipolis archive dc_7445752e8108, sha256 7d77bcf5e3d9186c6b900cb204c62188b26f5e990a123f7164558fd433da2f1b)
BibTeX
@misc{dc7445752e8108,
  title = {DIAL-S2ST: Learning Latency Control for Simultaneous Speech-to-Speech Translation — arXiv:2610.32214},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.32214},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7445752e8108; sha256 7d77bcf5e3d9186c6b900cb204c62188b26f5e990a123f7164558fd433da2f1b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.32214. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We propose DIAL-S2ST, a post-training method for learning inference-time latency control in simultaneous speech-to-speech translation models. Our approach reuses pretrained translation capabilities through self-distillation from lookahead predictions, while limiting the influence of large text distillation losses to account for the imitation gap. This provides supervision for earlier emission without paired target data or reward computation. To make the learned timing adjustable at inference, we jointly train a single LoRA adapter across scales with matched lookahead guidance. Experiments on French-to-English translation with Hibiki demonstrate that only 10 hours of source speech enable this adaptation, providing flexible latency control with a favorable quality-latency trade-off.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Sound (cs.SD); Audio and Speech Processing (eess.AS).
Pagina del lavoro: https://arxiv.org/abs/2610.32214 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-26T19:45:48.353Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).