# How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models

Source: [How Good is Your Tokenizer? On the Monolingual Performance of Multilingual Language Models](<https://www.sota2.com/research/paper/68fa6a0f1e209131cea25eac>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-25T23:15:37.395Z

Authors: Phillip Rust, Jonas Pfeiffer, Ivan Vuli\\'c, Sebastian Ruder, Iryna Gurevych

Published: 2020-12-31T14:11:00.000Z

## Abstract

In this work, we provide a systematic and comprehensive empirical comparison of pretrained multilingual language models versus their monolingual counterparts with regard to their monolingual task performance. We study a set of nine typologically diverse languages with readily available pretrained monolingual models on a set of five diverse monolingual downstream tasks. We first aim to establish, via fair and controlled comparisons, if a gap between the multilingual and the corresponding monolingual representation of that language exists, and subsequently investigate the reason for any performance difference. To disentangle conflating factors, we train new monolingual models on the same data, with monolingually and multilingually trained tokenizers. We find that while the pretraining data size is an important factor, a designated monolingual tokenizer plays an equally important role in the downstream performance. Our results show that languages that are adequately represented in the multilingual model's vocabulary exhibit negligible performance decreases over their monolingual counterparts. We further find that replacing the original multilingual tokenizer with the specialized monolingual tokenizer improves the downstream performance of the multilingual model for almost every task and language.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Pearson correlation analysis on Tasks sensitive to language variation](<https://www.sota2.com/research/sota/pearson-correlation-analysis-on-tasks-sensitive-to-language-variation.md>): Pearson correlation analysis; Tasks sensitive to language variation; Pearson Correlation: 0.37.

- [Pearson correlation analysis on Tasks robust to language variation](<https://www.sota2.com/research/sota/pearson-correlation-analysis-on-tasks-robust-to-language-variation.md>): Pearson correlation analysis; Tasks robust to language variation; Pearson Correlation (r): -0.45.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2012.15613>)

[PDF](<https://arxiv.org/pdf/2012.15613.pdf>)