Reproduce “DocAttriBench: Benchmarking Answer Grounding in Document 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
Luca De Grandis, Silvia Cappelletti, William Raccagni, Marcella Cornia, Lorenzo Baraldi, Rita Cucchiara
DocAttriBench links question-answer examples to document layout regions, making visual answer grounding measurable. The authors assemble 237,418 images and 295,554 examples, and use MAPPET, an automatic masking-based method, to select supporting regions. In grounded answer generation across seven test datasets, DAB fine-tuning raises Qwen3-VL-8B’s average overall answer correctness from 9.4 to 37.0; results cover mostly single-page answers grounded in one region.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.