Population Fidelity: Evaluating Population Representativeness in LLMs
This work introduces Population Fidelity, an evaluation framework that distinguishes key conditions required for a set of LLM-generated responses to represent a population and organizes these features and provides reusable code, data, and trained models for evaluating population fidelity across substantive domains and...
Neemias B. da Silva, Martin Lukk, A. Sutani et al.
· 0 citations