Metraipolis › archive › dc_d79cfcbcbef2

Familiar letters, not related languages, let unseen languages into a shared space of meaning — arXiv:2611.06006

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.06006
License (as archived): CC0-1.0
Archived: 2026-11-05T18:18:41.347Z
SHA-256 of the archived text: 53a566d6a31b191993dee4d38592f98499ea51e9f5475f3b3732eb09d3850b33
Stable URL of this page: https://metraipolis.com/documento/dc_d79cfcbcbef2

How to cite

Familiar letters, not related languages, let unseen languages into a shared space of meaning — arXiv:2611.06006 — arXiv (en), https://arxiv.org/abs/2611.06006 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_d79cfcbcbef2 (Metraipolis archive dc_d79cfcbcbef2, sha256 53a566d6a31b191993dee4d38592f98499ea51e9f5475f3b3732eb09d3850b33)

BibTeX

@misc{dcd79cfcbcbef2,
  title = {Familiar letters, not related languages, let unseen languages into a shared space of meaning — arXiv:2611.06006},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.06006},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_d79cfcbcbef2; sha256 53a566d6a31b191993dee4d38592f98499ea51e9f5475f3b3732eb09d3850b33},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.06006. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Hundreds of languages have written text but almost no translations, and whether a model learning one from untranslated text builds on a related language or familiar letters is unclear: they usually coincide. Here we show that familiar letters, not a related language, let an unseen language into a shared space of meaning. GlyphMind, a tokenizer-free vision transformer that reads text as pixels and lets any sentence be shown in any script, was pretrained to align translations across 27 writing systems. Adapted on untranslated Lao, a language and script it had never seen, it retrieved English translations seven times as often as the strongest simple baseline. Rewritten in unfamiliar letters, Lao did not enter although its close relative Thai was in pretraining; Santali, which had none, entered in Devanagari, familiar from Hindi, but not in its own script. Transliteration into familiar scripts may thus open multilingual models to languages without translations.

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/2611.06006 — 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 2026-11-05T18:13:03.941Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).