Qinghua Zhou

This article addresses stability and accuracy issues in deep learning models.

New techniques aim to fix errors in language models without complete retraining.

New compression techniques speed up training for large language models while maintaining accuracy.