NYU Machine Learning for Language
datasetactive

BBQ: A Hand-Built Bias Benchmark for Question Answering

View original resource

BBQ is a hand-built benchmark that measures social bias in question-answering models across nine demographic dimensions, testing how model outputs shift with and without disambiguating context.

Tags

benchmarkbiasfairnessquestion answering

At a glance

Published

2022

Jurisdiction

Global

Category

Datasets and benchmarks

Access

Public access

Build your AI governance program

VerifyWise helps you implement AI governance frameworks, track compliance, and manage risk across your AI systems.

BBQ: A Hand-Built Bias Benchmark for Question Answering | VerifyWise AI Governance Library