Bias Benchmark for Question Answering (BBQ)

The Bias Benchmark for Question Answering (BBQ) is a dataset used to evaluate the presence of demographic, gender, and geographic biases in large language models (LLMs) and the effectiveness of debiasing methods.

Papers using Bias Benchmark for Question Answering (BBQ) (3)

CatRAG: Functor-Guided Structural Debiasing with Retrieval Augmentation for Fair LLMs2026Breaking the Benchmark: Revealing LLM Bias via Minimal Contextual Augmentation2025Evaluating the Effect of Retrieval Augmentation on Social Biases2025