ByT5: Towards a Token-Free Future with Pre-trained Byte-to-Byte Models
Abstract
ByT5 studies byte-level sequence modeling at scale and demonstrates strong multilingual performance without tokenization.
Type
Publication
Transactions of the Association for Computational Linguistics
Rami Al-Rfou
Member of Technical Staff - TLM
My research interests include language modeling, embodied AI, motion forecasting, and multilingual modeling.