[TrustNLP@NAACL 2025] BiasEdit: Debiasing Stereotyped Language Models via Model Editing
Some scripts for evaluating stereotypes in LMs
This study investigates social biases in text-to-image models by examining biases in AI-generated images across demographic, occupational, and cultural prompts. Testing diverse inputs and collecting feedback, we aim to identify and analyze biases to propose strategies for fairer, more inclusive AI-generated visuals.
Pipeline LLM para adaptar o benchmark MGS (52k frases) ao contexto jurídico-cultural brasileiro: tradução PT-BR, novas categorias (regiao, orientacao_sexual) e classificação legal (Lei 7.716/1989, STF ADO 26/2019)
This web application is inspired by "Guess Who" game. The goal of the game is to study stereotypes.
How film and TV cast, write about and label women and social groups: IMDb, TMDb and 3M plot summaries, 1920–2026
To associate your repository with the stereotypes topic, visit your repo's landing page and select "manage topics."