VLR-CVC/DocVQA-2026

DocVQA 2026 | ICDAR2026 Competition on Multimodal Reasoning over Documents in Multiple Domains Building upon previous DocVQA benchmarks, this evaluation dataset introduces challenging reasoning questions over a diverse collection of documents spanning eight domains, including business reports, scientific papers, slides, posters, maps, comics, infographics, and engineering drawings. By expanding coverage to new document domains and… See the full description on the dataset page: https://huggingface.co/datasets/VLR-CVC/DocVQA-2026.