Applications of Agentic Systems for Scientific Discovery

This directory contains example applications demonstrating multi-agent systems for scientific discovery.

Examples

PolynomialDiscovery

Multiple autonomous agents collaboratively learn a hidden cubic polynomial through noisy observations and peer-to-peer communication. Demonstrates:

• Asynchronous multi-agent coordination
• Communication topology effects (ring vs random)
• Scaling behavior with agent count
• Collective learning from noisy data

RuleDiscovery

LLM-powered agents discover hidden rules governing a simulated world through experimentation and natural language communication. Demonstrates:

• LLM agents for scientific hypothesis generation
• Natural language experiment design
• Peer-to-peer knowledge sharing
• Multiple LLM backend support (Anthropic, OpenAI-compatible, Mock)

Related Resources

Our partners in the DOE Genesis Mission’s Common Agentic Framework (CAF) project operate an Agent of the Week program to highlight interesting agentic applications.

AotW#1 is OPAL - Plant Phenotyping Analysis Agent

AotW#2 is IGNOS — Fusion Foundation Model