ByT5: Towards a token-free future with pre-trained byte-to-byte models

TL;DR

This paper introduces ByT5, a byte-level Transformer model that operates directly on raw text, offering robustness to noise and language independence, while maintaining competitive performance with token-based models.

Contribution

The paper demonstrates that standard Transformer architectures can be effectively adapted for byte-level processing, and releases new pre-trained byte-level models based on T5.

Findings

Byte-level models are more robust to noise.

Byte-level models perform well on spelling and pronunciation sensitive tasks.

Byte-level models are competitive with token-level models in efficiency.

Abstract

Most widely-used pre-trained language models operate on sequences of tokens corresponding to word or subword units. By comparison, token-free models that operate directly on raw text (bytes or characters) have many benefits: they can process text in any language out of the box, they are more robust to noise, and they minimize technical debt by removing complex and error-prone text preprocessing pipelines. Since byte or character sequences are longer than token sequences, past work on token-free models has often introduced new model architectures designed to amortize the cost of operating directly on raw text. In this paper, we show that a standard Transformer architecture can be used with minimal modifications to process byte sequences. We characterize the trade-offs in terms of parameter count, training FLOPs, and inference speed, and show that byte-level models are competitive with their token-level counterparts. We also demonstrate that byte-level models are significantly more robust to noise and perform better on tasks that are sensitive to spelling and pronunciation. As part of our contribution, we release a new set of pre-trained byte-level Transformer models based on the T5 architecture, as well as all code and data used in our experiments.