Reproduce “Question Directed Graph Attention Network for Numerical Reasoning over Text” 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
Kunlong Chen, Weidi Xu, Xingyi Cheng, Zou Xiaochuan, Yuyu Zhang, Le Song, Taifeng Wang, Yuan Qi and 1 more
QDGAT helps answer numerical reading-comprehension questions by using question-guided message passing over a graph linking typed numbers with entities mentioned in the same sentence. On DROP’s test set, it reports 83.23% EM and 86.38% F1, versus NumNet+’s leaderboard-sourced 81.52% EM and 84.84% F1. This comparison indicates stronger performance in this benchmark setting, but does not establish general numerical reasoning ability.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.