Event Details

Scaling laws research has focused overwhelmingly on English — yet the most prominent AI models explicitly serve billions of international users. In this work, we undertake the largest multilingual scaling laws study to date, totaling 774 multilingual training experiments, spanning 10M-8B model parameters, 400+ training languages and 48 evaluation languages. We introduce the Adaptive Transfer Scaling Law (ATLAS) for both monolingual and multilingual pretraining, which outperforms existing scaling laws’ out-of-sample generalization often by more than 0.3 R^2. Our analyses of the experiments shed light on multilingual learning dynamics, transfer properties between languages, and the curse of multilinguality. First, we derive a cross-lingual transfer matrix, empirically measuring mutual benefit scores between 38 x 38=1444 language pairs. Second, we derive a language-agnostic scaling law that reveals how to optimally scale model size and data when adding languages without sacrificing performance. Third, we identify the computational crossover points for when to pretrain from scratch versus finetune from multilingual checkpoints. We hope these findings provide the scientific foundation for democratizing scaling laws across languages, and enable practitioners to efficiently scale models — beyond English-first AI.

Zoom:
Password: 2025

Host: DJ Ashok
POC: Maura Covaci

Speaker Bio

Shayne is a PhD candidate at MIT, whose research investigates the intersection of AI and its global impact by developing data-centric methods for system building and auditing AI’s influence on the web, economy, and society. He leads the Data Provenance Initiative, and has contributed to training models such as Bloom, Aya, and Flan-T5/PaLM. His work has earned Best Paper Awards at ACL 2024 and NAACL 2024, and has been featured in prominent outlets including the New York Times, Washington Post, The Atlantic, and MIT Tech Review.