scispace - formally typeset
T

Tae Hwan Jung

Publications -  2

Tae Hwan Jung is an academic researcher. The author has contributed to research in topics: Key (cryptography) & Product (mathematics). The author has co-authored 2 publications.

Papers
More filters
Posted Content

Large Product Key Memory for Pretrained Language Models

TL;DR: This article proposed a new memory usage metric, and careful observation using this metric reveals that most memory slots remain outdated during the training of PKM-augmented models, and propose simple but effective solutions: (1) initialization from the model weights pretrained without memory and (2) augmenting PKM by addition rather than replacing a feed-forward network.
Proceedings ArticleDOI

Large Product Key Memory for Pretrained Language Models

TL;DR: A new memory usage metric is defined, and careful observation reveals that most memory slots remain outdated during the training of PKM-augmented models, enhancing memory utilization and downstream performance.