Handwriting Recognition Using Large Vision-Language Models: Impact of Segmentation

A journal-length extension of our segmentation-free recognition work: a broader set of models, scripts and collections, and a closer analysis of when segmentation helps a vision-language model and when it throws away the context the model needs.

Citation

T. Hallyburton, L. Berset, O. D'Ancona, A. Scius-Bertrand, G. A. Fink and A. Fischer. "Handwriting Recognition Using Large Vision-Language Models: Impact of Segmentation." Informatica. Submitted June 2026, under review.