Metraipolis › archive › dc_1052e90dff4f

Polish Medical Visual Question Answering: Vision-Language Models Underutilize Visual Evidence — arXiv:2608.12928

Source
arXiv (en)
Original
https://arxiv.org/abs/2608.12928
License (as archived)
CC0-1.0
Archived
2026-10-19T04:17:25.501Z
SHA-256 of the archived text
3dba190b85cc1138ebd5ef69b7179806e00d91351448fbc5fb0445ff9d7a8204
Stable URL of this page
https://metraipolis.com/documento/dc_1052e90dff4f

How to cite

Polish Medical Visual Question Answering: Vision-Language Models Underutilize Visual Evidence — arXiv:2608.12928 — arXiv (en), https://arxiv.org/abs/2608.12928 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_1052e90dff4f (Metraipolis archive dc_1052e90dff4f, sha256 3dba190b85cc1138ebd5ef69b7179806e00d91351448fbc5fb0445ff9d7a8204)
BibTeX
@misc{dc1052e90dff4f,
  title = {Polish Medical Visual Question Answering: Vision-Language Models Underutilize Visual Evidence — arXiv:2608.12928},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2608.12928},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_1052e90dff4f; sha256 3dba190b85cc1138ebd5ef69b7179806e00d91351448fbc5fb0445ff9d7a8204},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2608.12928. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

We introduce a Polish-language medical visual question answering (VQA) benchmark, built from Polish Board Certification Examination questions for licensed physicians and dentists pursuing specialist certification. The benchmark comprises image-containing questions spanning diverse medical specialties and visual domains, together with a text-only question answering (QA) control set. We evaluate Polish-oriented, general-purpose open-weight, and commercial vision-language models. The task remains challenging: the best model achieves 79.0\% accuracy on the full VQA set, and only GPT-5.6 surpasses the approximate human reference on the subset with available candidate responses; all other evaluated models perform worse than humans. To assess visual grounding, we compare complete inputs with configurations omitting the image, the question, or both, and categorize questions by image importance. Models derive more useful information from the question text than from the image and perform worse on image-dominant questions. Across both QA and VQA, they nevertheless achieve above-chance accuracy from the answer choices alone, showing that non-trivial performance can persist even when key task components are missing.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-08-14 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2608.12928 — 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 2026-10-19T04:13:06.559Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-08-14?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).