Reproduce “BobFlow: Towards Human-Agent Collaborative Data Science Using Dataflows” 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

Jiadong Bai, Yicong Huang, Chen Li

BobFlow represents an agent’s analysis as a shared directed acyclic graph of executable operators, which analysts can inspect and revise. The agent builds and runs the graph, while retained outputs expose intermediate results. On PrepBench, Gemini 3.8 Flash scored 89.5 versus 70.9 for Script ReAct, but this end-to-end comparison does not isolate the graph’s effect.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.