The reviewed record of science sign in
Pith

arxiv: 2506.18639 · v1 · pith:B6QMVEAV · submitted 2025-06-23 · cs.CL

ByteSpan: Information-Driven Subword Tokenisation

Reviewed by Pithpith:B6QMVEAVopen to challenge →

classification cs.CL
keywords bytespansubwordbytesexperimentsinformation-drivenpredictablerepresentationstokenisation
0
0 comments X
read the original abstract

Recent dynamic tokenisation methods operate directly on bytes and pool their latent representations into patches. This bears similarities to computational models of word segmentation that determine lexical boundaries using spikes in an autoregressive model's prediction error. Inspired by this connection, we explore whether grouping predictable bytes - rather than pooling their representations - can yield a useful fixed subword vocabulary. We propose a new information-driven subword tokeniser, ByteSpan, that uses an external byte-level LM during training to identify contiguous predictable byte sequences and group them into subwords. Experiments show that ByteSpan yields efficient vocabularies with higher morphological alignment scores than BPE for English. Multilingual experiments show similar compression and R\'enyi efficiency for 25 languages.

This paper has not been read by Pith yet.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Faster Superword Tokenization

    cs.CL 2026-04 accept novelty 7.0

    Frequency aggregation of supermerge candidates and a two-phase formulation make BoundlessBPE and SuperBPE training over 600x faster on 1GB data while preserving identical results, with open-source Python and Rust code.