Causal Estimation of Tokenisation Bias


Loading...

Date

2025-07

Publication Type

Conference Paper

ETH Bibliography

yes

Citations

Altmetric

Data

Abstract

Modern language models are typically trained over subword sequences, but ultimately define probabilities over character-strings. Ideally, the choice of the tokeniser—which maps character-strings to subwords—should not affect the probability assigned to the underlying character-string; in practice, it does. We define this mismatch as **tokenisation bias**. In this work, we quantify one particular type of tokenisation bias: the effect of including or not a subword (e.g., ⟨ hello ⟩) in a tokeniser’s vocabulary on the probability a trained model assigns to the corresponding characters (i.e., “hello”). Estimating this effect is challenging because each model is trained with only one tokeniser. We address this by framing tokenisation bias as a causal effect and estimating it using the regression discontinuity design. Specifically, we exploit the fact that tokenisation algorithms rank subwords and add the first K to a tokeniser’s vocabulary, where K is an arbitrary cutoff point. As such, we can estimate a causal effect by comparing similar subwords around this cutoff. Experimentally, we find that tokenisation consistently affects models’ outputs across scales, vocabularies, and tokenisers. Notably, a subword’s presence in a small model’s vocabulary may increase its characters’ probability by up to 17 times, highlighting tokenisation as a key design choice in language modelling.

Publication status

published

Book title

Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)

Journal / series

Volume

Pages / Article No.

28325 - 28340

Publisher

Association for Computational Linguistics

Event

63rd Annual Meeting of the Association for Computational Linguistics (ACL 2025)

Edition / version

Methods

Software

Geographic location

Date collected

Date created

Subject

Organisational unit

09462 - Hofmann, Thomas / Hofmann, Thomas check_circle

Notes

Funding

Related publications and datasets