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

N/A

Published as a conference paper at ICLR 2026 ATLAS : A DAPTIVE T RANSFER S CALING L AWS FOR M ULTILINGUAL P RETRAINING , F INETUNING , AND D ECODING THE C URSE OF M ULTILINGUALITY Shayne Longpre∗1 Sneha Kudugunta2 Niklas Muennighoff3 I-Hung Hsu4 Isaac Caswell5 Alex Pentland1 Sercan Ö. Arık4 Chen-Yu Lee4 Sayna Ebrahimi5 1…

related reading

• Scaling Laws, Carefully | Lil'Loglilianweng.github.io
• [2605.12715] Scaling Laws for Mixture Pretraining Under Data Constraintsarxiv.org
• [2005.14165] Language Models are Few-Shot Learnersarxiv.org
• frontier model training methodologies | Alex Wa's Blogdjdumpling.github.io
• [2001.08361] Scaling Laws for Neural Language Modelsarxiv.org
• [2405.10938] Observational Scaling Laws and the Predictability of Language Model Performancearxiv.org
• Scaling is subtler than it seemsberen.io
• MLAS-LoRA: Language-Aware Parameters Detection and LoRA-Based Knowledge Transfer for Multilingual Machine Translation - ACL Anthologyaclanthology.org
• Recent Advances in Language Model Fine-tuningruder.io
• [2509.14786] Pre-training under infinite computearxiv.org
• Pathways Language Model (PaLM): Scaling to 540 Billion Parameters for Breakthrouai.googleblog.com
• Large Language Diffusion Modelsarxiv.org