Reproduce “Searching for What Matters: Open-Ended Discovery of Multiple Hypotheses with Agentic Data Exploration” 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
Sanchit Kabra, Nurendra Choudhary, Jiangshu Du, Eddie Huang, Danai Koutra, Chandan K Reddy
HypAgent searches open-ended tabular data for multiple mechanisms without a supplied question. It proposes directions, uses observational residuals to redirect later searches, and refines candidates into claims about form, direction, magnitude, and conditions. On OpenHypBench, it scores 0.47 precision, 0.51 recall, and 0.49 F1, outperforming the strongest baseline by F1; this synthetic benchmark does not establish real-world causal discovery.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.