Why Larger Models Learn More: Effects of Capacity, Interference, and Rare-Task Retention

2026-05-28

AI Summary

This work investigates why large language models can master rare and complex tasks that smaller models fail to learn. Through synthetic task mixtures and real pretraining experiments with the OLMo model series (ranging from 4M to 4B parameters), combined with gradient interference analysis and representational probing, the study proposes a “data-centric” perspective: sufficiently large models alleviate neuron resource competition caused by imbalanced data distributions, substantially reducing inter-task gradient interference and thereby effectively preserving features of rare tasks. Experimental results demonstrate that only large models successfully acquire low-frequency complex tasks, exhibiting richer task-specific information in their internal representations. These findings underscore the critical role of resource allocation and interference dynamics in model scaling.

Abstract

Larger models learn tasks smaller models do not. What drives this phenomenon? We develop a simple phenomenological argument that power-law scaling already suggests that a larger model will be able to learn a part of the data distribution that a smaller model fails to learn, even with infinite training data. To validate this claim and identify its causes, we study the effects of model scaling on a synthetic setup consisting of a mixture of tasks that show monotonic scaling curves. The results point to a data-induced competition over resources (neurons). Specifically, smaller models allocate their neurons to high frequency or low complexity tasks, and so they learn solutions that perform poorly on rare and complex tasks. Moreover, this happens even when solutions capable of expressing the desired task exist. We then assess how a larger model circumvents this data-centric bottleneck, finding that it traces to a reduced interference mechanism: larger models can allocate enough resources to common tasks that the gradient updates for those tasks become weak, which means that they do not overwrite rare-task features as they slowly accumulate. Finally, to further validate these claims, we pretrain OLMo models (4M to 4B parameters) on novel tasks of varying frequency and complexity. The results mirror those from our synthetic data experiments: only the larger OLMo models learn the infrequent and complex tasks, and these larger models embed more task features in their representations and show less gradient interference between tasks. Overall, we offer a data-centric account of why larger models learn tasks that smaller models fail to. This helps explain why larger models are better in practice, and it can inform practical questions concerning model sizing and training data mixtures.

Problem

• model scaling
• rare-task retention
• gradient interference
• data distribution
• task complexity

Innovation

• model scaling
• gradient interference
• task retention
• data-induced competition
• rare-task learning