parameter-sweep

What problem does it solve?

This Skill automates the process of testing different governance configurations in SWARM simulations, allowing users to identify optimal parameters for desired outcomes like welfare or toxicity reduction.

Core Features & Use Cases

• Parameter Grid Sweeps: Systematically explore a range of values for specific governance parameters (e.g., transaction tax rate).
• Multi-Seed Simulations: Run simulations with multiple random seeds for each parameter configuration to ensure robust results.
• Summary Statistics Generation: Consolidate results into easily digestible CSV and JSON files, including mean/std deviations and best-performing configurations.
• Use Case: A researcher wants to find the optimal transaction tax rate in SWARM that maximizes agent welfare while keeping toxicity below a certain threshold. This Skill can run hundreds of simulations to pinpoint that rate.

Quick Start

Run a parameter sweep for the 'governance.transaction_tax_rate' parameter across values [0.0, 0.05, 0.10] with 3 seeds, saving results to '/tmp/sweep_output'.