Reproduce “PEAR: Progressive Evidence-Based AutoResearch for Industrial Search Systems” 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
Yifan Wang, Shipeng Zhu, Fei Xiong, Yuqin Yang, Yonghui Huang, Kunyao Wu, Yue Wang, Weichao Meng and 1 more
PEAR links hypotheses, candidate changes, results, and experimental context for each search strategy, then advances candidates through confidence-gated evaluations from shadow traffic to randomized online tests. It updates hypotheses from supportive, conflicting, or inconclusive evidence. In two real-world A/B experiments, Main Order/DAU increased 2.7336% and 3.2957%, respectively, both with p < 0.05, in one e-commerce general-search environment.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.