ATLAS：面向多语言预训练、微调及破解多语言诅咒的自适应迁移缩放法则

摘要

当前关于缩放定律的研究过度集中于英语领域，而最前沿的AI模型实际服务着数十亿国际用户。本研究开展了迄今规模最大的多语言缩放定律分析，累计完成774项多语言训练实验，覆盖模型参数量级从1000万至80亿，训练语言超400种，评估语言达48种。我们提出适用于单语与多语预训练的自适应迁移缩放定律（ATLAS），其样本外泛化能力较现有缩放定律普遍提升超过0.3的R²值。通过实验分析，我们揭示了多语言学习动态机制、语言间迁移特性以及多语言性诅咒现象：首先推导出跨语言迁移矩阵，实证测量38×38=1444组语言对间的互惠分值；其次建立语言无关的缩放定律，揭示在扩展语言种类时如何优化模型规模与数据配置以保持性能；最后确定了从头预训练与基于多语言检查点微调的计算临界点。这些发现有望为跨语言缩放定律的民主化奠定科学基础，助力实践者突破英语优先的AI开发现状，实现模型的高效扩展。

English

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.

ATLAS：面向多语言预训练、微调及破解多语言诅咒的自适应迁移缩放法则

ATLAS: Adaptive Transfer Scaling Laws for Multilingual Pretraining, Finetuning, and Decoding the Curse of Multilinguality

摘要

Support