Deliberate Evolution: Agentic Reasoning for Sample-Efficient Symbolic Regression with LLMs

In three linesDeliberate Evolution (DE) is an agentic framework decoupling symbolic generation from search control for symbolic regression. DE guides LLM proposals via adaptive operators, analytical tools, and reflective memory. On LLM-SRBench, DE outperforms baselines using only 40% of standard sample budget.

Summary generated by Claude — human-verified