Reproduce “Aleks: AI powered Multi Agent System for Autonomous Scientific Discovery via Data-Driven Approaches in Plant 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
Daoyuan Jin, Nick Gunner, Niko Carvajal Janke, Shivranjani Baruah, Kaitlin M. Gold, Yu Jiang
Cornell University researchers developed "Aleks," an AI-powered multi-agent system designed for autonomous scientific discovery, which successfully formulates problems, explores modeling strategies, and refines solutions in data-driven plant science. Applied to Grapevine Red Blotch Disease prediction, it autonomously identified biologically meaningful features and converged on robust predictive models, reducing the research cycle from ideation to feedback to just a few hours per experiment.
Key insights
- The Domain Scientist agent is crucial for guiding coherent, biologically relevant feature engineering, preventing the system from solely pursuing statistically correlated but biologically unmeaningful features.
- Maintaining a full, shared experimental history in the agent memory is essential for robust reasoning across iterations, ensuring consistency, and preventing errors like data leakage.
- Aleks autonomously explores diverse problem formulations (e.g., classification vs. regression) and successfully identifies and prioritizes biologically meaningful original and derived features.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.