Metraipolis › archive › dc_eeec58dad320

Translation Circuit Discovery and Targeted Optimization for Low-Resource Multilingual Translation — arXiv:2610.21456

Source: arXiv (en)
Original: https://arxiv.org/abs/2610.21456
License (as archived): CC0-1.0
Archived: 2026-10-16T09:57:41.813Z
SHA-256 of the archived text: d7d66231408c183b9dbbc4143acf42b2ef7f722d07b9dcd8a3b70e16a30418d6
Stable URL of this page: https://metraipolis.com/documento/dc_eeec58dad320

How to cite

Translation Circuit Discovery and Targeted Optimization for Low-Resource Multilingual Translation — arXiv:2610.21456 — arXiv (en), https://arxiv.org/abs/2610.21456 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_eeec58dad320 (Metraipolis archive dc_eeec58dad320, sha256 d7d66231408c183b9dbbc4143acf42b2ef7f722d07b9dcd8a3b70e16a30418d6)

BibTeX

@misc{dceeec58dad320,
  title = {Translation Circuit Discovery and Targeted Optimization for Low-Resource Multilingual Translation — arXiv:2610.21456},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.21456},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_eeec58dad320; sha256 d7d66231408c183b9dbbc4143acf42b2ef7f722d07b9dcd8a3b70e16a30418d6},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.21456. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Recent multilingual large language models achieve strong translation performance, yet the internal mechanisms responsible for low-resource multilingual translation remain insufficiently understood. In this paper, we study translation circuits in multilingual decoder-only LLMs for Hindi-to-Urdu and Hindi-to-Sindhi (Devanagari). We propose TITAN, a mechanistic framework that combines causal intervention, sparse feature decomposition, and targeted preference optimization to identify and optimize crucial translation components in the transformer. Unlike conventional full-parameter fine-tuning, TITAN restricts optimization to sparse, causally important circuits, enabling scalable and parameter-efficient adaptation while updating fewer than 5% of model parameters. Experimental results show that TITAN achieves BLEU improvements of 33.75% on Hindi-to-Urdu and 45.86% on Hindi-to-Sindhi over the strongest multilingual baselines. Human expert evaluation further confirms substantial gains in adequacy and fluency across both language pairs. Additional analyses reveal that multilingual translation is implemented through sparse, structured circuits operating over a shared semantic subspace rather than through an English-centric pivot representation. Overall, our findings demonstrate that mechanistic circuit discovery can directly improve low-resource multilingual translation while providing interpretable insights into multilingual transformer behavior.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2610.21456 — 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 2026-10-16T08:44:47.655Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).