Reproduce “MInfoVisQA: A Multilingual Benchmark for Information-Rich Visual Question Answering” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Changjiang Jiang, Jiang Zhou, Xinbin Yuan, Fan Zhang, Tao Feng, Zhuohan Xie
The authors build a paired benchmark to test whether models recover the same chart and table facts across languages. It links 128 seed questions to 8,960 configurations in 24 languages and evaluates aligned and fixed-English or Chinese question-answer settings. All 25 models score lower with non-English visuals and English questions than with English visuals; the comparison does not identify a cause.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.