Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

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

About

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.

Linting Xue, Aditya Barua, Noah Constant, Rami Al-Rfou, Sharan Narang, Mihir Kale, Adam Roberts, Colin Raffel• 2021

Related benchmarks

TaskDatasetResultRank
Natural Language UnderstandingGLUE
SST-291.6
452
Text ClassificationAG News (test)
Accuracy93.6
210
Natural Language UnderstandingGLUE (val)--
170
Natural Language InferenceXNLI
Accuracy85.7
111
Text ClassificationIMDB (test)
CA91.5
79
Natural Language UnderstandingSuperGLUE (test)
BoolQ Accuracy79.2
63
Paraphrase IdentificationPAWS-X
Accuracy91.7
57
Multilingual Question AnsweringTyDiQA
Accuracy81.9
44
Named Entity RecognitionWikiAnn
F1 Score93.7
32
Comment ClassificationCivil Comments
Accuracy82.8
21
Showing 10 of 38 rows

Other info

Code

Follow for update