Metraipolis › archive › dc_463c8f1cb75c

Dynamic Lagging using Stable-Prefix Training for Simultaneous Translation — arXiv:2609.05799

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.05799
License (as archived)
CC0-1.0
Archived
2026-10-16T12:34:45.357Z
SHA-256 of the archived text
a54d19bf9b63b2e8c3338253c96f60c5cd10e7f568163a528d3463563b45d4dc
Stable URL of this page
https://metraipolis.com/documento/dc_463c8f1cb75c

How to cite

Dynamic Lagging using Stable-Prefix Training for Simultaneous Translation — arXiv:2609.05799 — arXiv (en), https://arxiv.org/abs/2609.05799 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_463c8f1cb75c (Metraipolis archive dc_463c8f1cb75c, sha256 a54d19bf9b63b2e8c3338253c96f60c5cd10e7f568163a528d3463563b45d4dc)
BibTeX
@misc{dc463c8f1cb75c,
  title = {Dynamic Lagging using Stable-Prefix Training for Simultaneous Translation — arXiv:2609.05799},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.05799},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_463c8f1cb75c; sha256 a54d19bf9b63b2e8c3338253c96f60c5cd10e7f568163a528d3463563b45d4dc},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.05799. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

In streaming simultaneous speech translation, the speech translation system is trained to learn a read-write policy that alternates between consuming source words and generating target ones. In a cascaded setting, the output from the speech recognizer is passed to a separate machine translation component, making it more difficult to learn such a policy. Approximations such as fixed wait-k strategies or target-suffix deletion can be employed, but these approaches do not provide the model with a streaming system's flexibility to make contextual read-write decisions. This paper presents a training strategy for a cascaded machine translation system that enables it to dynamically decide how much of the growing source prefix to translate. We achieve this by fine-tuning a large language model (Qwen3-8B) on stable prefixes of the training data, which are produced by pairing every source sentence prefix in the training data with the longest translation of that prefix that is shared with the full source sentence translation. We fine-tune variants of the model on different subsets of the prefixes and compare against wait-k and target-suffix deletion. We also investigate the effect of fine-tuning the target-token generation confidence. Our experiments show that stable prefixes improve the quality-latency tradeoff when translating from English into German, Japanese, and Chinese across a range of test sets.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2609.05799 — 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 2026-10-16T12:30:24.984Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).