Reproduce “DrugTargetWorld: A Synthetic Biobank for Training and Benchmarking AI Scientists” 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
Samuel Margolis, Paul Schmiedmayer, Alan Huang, Ethan Chen, Ishan Bhattacharjee, Atman Shah, Ben Viggiano, Fang Cao and 7 more
DrugTargetWorld creates synthetic cardiovascular biobanks with hidden causal answers to evaluate agents that construct disease phenotypes, identify protein drivers, choose intervention directions and optionally run experiments. Across 540 episodes, Opus 5 averaged 39.98/100 and GPT-5.6 Sol 35.38/100. No agent reliably rejected misleading non-causal proteins; these scores come from designed cardiovascular worlds, not real patients.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.