We present BabyBabelLM, a multilingual collection of datasets modeling the language a person observes from birth until they acquire a native language. We curate developmentally plausible pretraining data aiming to cover the equivalent of 100M English words of content in each of 45 languages. We compile evaluation suites and train baseline models in each language. BabyBabelLM aims to facilitate multilingual pretraining and cognitive modeling.
@article{arxiv.2510.10159,
title = {BabyBabelLM: A Multilingual Benchmark of Developmentally Plausible Training Data},
author = {Jaap Jumelet and Abdellah Fourtassi and Akari Haga and Bastian Bunzeck and Bhargav Shandilya and Diana Galvan-Sosa and Faiz Ghifari Haznitrama and Francesca Padovani and Francois Meyer and Hai Hu and Julen Etxaniz and Laurent Prévot and Linyang He and María Grandury and Mila Marcheva and Negar Foroutan and Nikitas Theodoropoulos and Pouya Sadeghi and Siyuan Song and Suchir Salhan and Susana Zhou and Yurii Paniv and Ziyin Zhang and Arianna Bisazza and Alex Warstadt and Leshem Choshen},
journal= {arXiv preprint arXiv:2510.10159},
year = {2025}
}