M.J.F. Valdez

Evaluating Retrieval-Augmented Generation on Social Bias Benchmarks across Small Language Models

A fundamental disconnect is suggested between a model's capacity for factual accuracy and its ability to maintain social fairness, highlighting the need for multi-dimensional evaluation frameworks for small-scale systems.