ExBERT: Improving cross-domain performance of a DistilBERT Q&A bot via a Mixture-of-Experts (MoE) architectural modification

A detailed study on improving cross-domain performance of a DistilBERT Q&A bot via a Mixture-of-Experts (MoE) modification.

This project was completed as part of a 2-month long, open-ended study for CS 7643 - Deep Learning at Georgia Tech.

GitHub has difficulty rendering some PDFs--Please download and open 'Bring in the ExBert.pdf' to view the pdf of the article.