TextWorld

First TextWorld Problems: A Reinforcement and Language Learning Challenge

The goal of the First TextWorld Problems (FTWP) was to build an AI agent that could play efficiently and win simplified text-based games. This competition highlighted the limitation of existing reinforcement learning models when combined with natural language processing. Head over to CodaLab for further details on the past competition and to our Microsoft Research blog post to learn more about the competition winner’s strategy.

Check back soon for when the training and evaluation games are available for download.