Researchers From Nyu And Meta Ai Studies Improving Social Conversational Agents By Learning From Natural Dialogue Between Users And A Deployed Model, Without Extra Annotations

This article examines the use of implicit feedback signals from natural user discussions to improve dialogue models in reinforcement learning with human feedback. Researchers from New York University and Meta AI use publicly available, de-identified data from the BlenderBot online deployment to investigate this problem. Their novel models are found to be superior to the baseline replies through both automated and human judgments.