Salvato in:
Dettagli Bibliografici
Autori principali: Gigant, Théo, Peng, Bowen, Quesnelle, Jeffrey
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:https://arxiv.org/abs/2604.27263
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911685256675328
author Gigant, Théo
Peng, Bowen
Quesnelle, Jeffrey
author_facet Gigant, Théo
Peng, Bowen
Quesnelle, Jeffrey
contents Subword tokenization is an essential part of modern large language models (LLMs), yet its specific contributions to training efficiency and model performance remain poorly understood. In this work, we decouple the effects of subword tokenization by isolating them within a controlled byte-level pretraining pipeline. We formulate and test hypotheses across various dimensions, including sample throughput, vocabulary scaling, and the linguistic prior of subword boundaries. By simulating these effects in a byte-level setting, we refine our understanding of why subword models outperform raw byte models and offer insights to improve the pretraining of future byte-level and subword models. Specifically, our experiments highlight the critical role of increased training throughput and the integration of subword boundaries as either explicit priors or inductive biases.
format Preprint
id arxiv_https___arxiv_org_abs_2604_27263
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation
Gigant, Théo
Peng, Bowen
Quesnelle, Jeffrey
Computation and Language
Subword tokenization is an essential part of modern large language models (LLMs), yet its specific contributions to training efficiency and model performance remain poorly understood. In this work, we decouple the effects of subword tokenization by isolating them within a controlled byte-level pretraining pipeline. We formulate and test hypotheses across various dimensions, including sample throughput, vocabulary scaling, and the linguistic prior of subword boundaries. By simulating these effects in a byte-level setting, we refine our understanding of why subword models outperform raw byte models and offer insights to improve the pretraining of future byte-level and subword models. Specifically, our experiments highlight the critical role of increased training throughput and the integration of subword boundaries as either explicit priors or inductive biases.
title Decoupling the Benefits of Subword Tokenization for Language Model Training via Byte-level Simulation
topic Computation and Language
url https://arxiv.org/abs/2604.27263