Reproduce “AutoMind: Adaptive Knowledgeable Agent for Automated Data Science” 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

Yujie Luo, Yixin Ou, Jingsheng Zheng, Lanning Wei, Zhuoyun Yu, Shuofei Qiao, Jintian Zhang, Da Zheng and 4 more

AutoMind helps data-science agents build and test solutions for unfamiliar tasks by retrieving curated Kaggle solution insights and research, exploring candidate plans in a solution tree, and routing code generation between one-pass and stepwise modes based on plan complexity. On a filtered 15-task MLE-Bench subset, DeepSeek-V3 AutoMind reaches 0.41 overall avg@3 versus AIDE’s 0.36, under 24-hour task budgets.

Summary by alphaXiv.

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