Uses AlphaEvolve to discover interpretable models of human and LLM strategic behavior from data. Analysis on iterated rock-paper-scissors shows frontier LLMs capable of deeper strategy than humans. Provides foundation for understanding behavioral differences in interactions.
Key Points
- 1.Program discovery for BGT models
- 2.LLMs exceed human strategic depth
- 3.Open-ended structural factor analysis
Impact Analysis
Enables better prediction of LLM behavior in social scenarios, aiding safe deployment.