Google Research

GAP Shared Task Overview

  • Kellie Webster
  • Marta R. Costa-jussà
  • Christian Hardmeier
  • Will Radford


We overview a shared task we ran using the GAP coreference challenge, including the logistics of running a task with 263 active participants and the modeling trends we observed. We found that fine-tuning BERT with gender balanced data produced a fair model, which serves as a recommendation to the community about one way to approach fairness in NLP modeling.

Learn more about how we do research

We maintain a portfolio of research projects, providing individuals and teams the freedom to emphasize specific types of work