PolicyBench

A mini-benchmark comparing Large Language Model estimates to PolicyEngine calculations of US tax/benefit programs.

Requirements

• Python 3.9+
• policyengine-us for ground truth
• edsl for LLM queries
• (Optional) pytest, etc. for tests

Installation

1. Clone the repo:

   git clone https://github.com/YOUR_USERNAME/policybench.git
   cd policybench