Metraipolis › archive › dc_ef669d30fa6a

Which one is banana man? Evaluating vision-language models in multi-turn pragmatic interpretation — arXiv:2608.29571

Source: arXiv (en)
Original: https://arxiv.org/abs/2608.29571
License (as archived): CC0-1.0
Archived: 2026-10-17T08:07:01.886Z
SHA-256 of the archived text: f5f41bd17b77b698cf880da6ed7b23661607f286af72aa4555d1e2b2f6b2387b
Stable URL of this page: https://metraipolis.com/documento/dc_ef669d30fa6a

How to cite

Which one is banana man? Evaluating vision-language models in multi-turn pragmatic interpretation — arXiv:2608.29571 — arXiv (en), https://arxiv.org/abs/2608.29571 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ef669d30fa6a (Metraipolis archive dc_ef669d30fa6a, sha256 f5f41bd17b77b698cf880da6ed7b23661607f286af72aa4555d1e2b2f6b2387b)

BibTeX

@misc{dcef669d30fa6a,
  title = {Which one is banana man? Evaluating vision-language models in multi-turn pragmatic interpretation — arXiv:2608.29571},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.29571},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ef669d30fa6a; sha256 f5f41bd17b77b698cf880da6ed7b23661607f286af72aa4555d1e2b2f6b2387b},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.29571. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Flexible adaptation to context and shared pragmatic intuitions contribute to smooth human conversation. Iterated reference games---in which players repeatedly pick out novel referents using language---present a test case for agents' ability to perform context-sensitive pragmatic reasoning in multi-turn linguistic environments. We tested humans and vision--language models on their ability to identify the intended meaning of descriptions produced in iterated reference games, varying the provided context in terms of amount, order, and relevance. While humans performed well consistently, the models we evaluated could make use of prior context to interpret humans' referring expressions, but they struggled to build up the relevant context to interpret those expressions effectively. Our results suggest that the models we evaluated lack core skills needed for efficient linguistic collaboration.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-01 in cs.CL).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2608.29571 — 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 2026-10-17T07:47:52.710Z dalla pagina https://arxiv.org/catchup/cs.CL/2026-09-01?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).