This dataset is part of the KoViDoRe Benchmark for Korean visual document retrieval evaluation. Specifically, it is focused on Visual Question Answering (VQA) tasks for Korean document images.
Domain: Korean structured document images (e.g., 환경기상, 공공행정 등 다양한 카테고리)
Task: Visual document retrieval (text question → relevant document image)
Format: BEIR-compatible (corpus, queries, qrels)
Number of Documents (Pages): 1,100+… See the full description on the dataset page:
https://huggingface.co/datasets/whybe-choi/kovidore-vqa-v1.0-beir.
