AstroAgents
Published:
A Multi-Agent AI for Hypothesis Generation from Mass Spectrometry Data
Agentic AI · LangChain · AI for Science · Hypothesis Generation — ICLR 2025 Workshop on Agentic AI for Science
Project website Paper (arXiv) Nature News
Inspired by LifeTracer, AstroAgents is a multi-agent AI system for hypothesis generation. A team of specialized LLM agents — a planner, scientist, data analyst, and critic — collaborate to propose and iteratively refine hypotheses about the origin of organic compounds in meteorite samples.
With upcoming sample-return missions across the solar system, AstroAgents introduces a paradigm for AI-driven scientific discovery, and we study the comparative performance of Claude 3.5 Sonnet and Gemini 2.0 Flash as the underlying reasoning engines.
