A new arXiv paper introduces KoViDoRe, a benchmark for Korean visual document retrieval. The work responds to a gap in the field: while multimodal retrieval has advanced, most existing benchmarks focus on English-language content. KoViDoRe aims to support evaluation for Korean documents that are visually rich, such as PDFs and reports.

The abstract notes that recent progress in multimodal retrieval has improved the ability to pull information from such documents, but that the lack of diverse-language benchmarks limits broader progress. By providing a Korean-language benchmark, the authors intend to help researchers develop and test systems that work beyond English-centric settings.

Because the abstract is truncated, the full details of the benchmark's construction, size, and evaluation methodology are not available in the source. The paper's contribution, as described, is the benchmark itself and the motivation behind it.