Getting started

Introduction

What the score means and what it does not.

Groundwork answers one question: for each claim in this answer, is there a passage in the retrieved context that supports it? It does not measure whether the claim is true in the world.

bash

pip install groundwork-eval

groundwork run \
  --dataset ./eval/legal-qa.jsonl \
  --threshold 0.92 \
  --report ./out/report.json