Too Big to Think: Capacity, Memorization, and Generalization in Pre-Trained Transformers
This research addresses the problem of understanding learning trade-offs in large language models for AI researchers, offering insights into model design, though it is incremental in its controlled setting.
The study investigated the relationship between memorization and generalization in pre-trained Transformers using synthetic tasks, finding that small models generalize but fail to memorize, while larger models memorize but fail to generalize, with no model succeeding at both tasks when trained jointly.
The relationship between memorization and generalization in large language models (LLMs) remains an open area of research, with growing evidence that the two are deeply intertwined. In this work, we investigate this relationship by pre-training a series of capacity-limited Transformer models from scratch on two synthetic character-level tasks designed to separately probe generalization (via arithmetic extrapolation) and memorization (via factual recall). We observe a consistent trade-off: small models extrapolate to unseen arithmetic cases but fail to memorize facts, while larger models memorize but fail to extrapolate. An intermediate-capacity model exhibits a similar shift toward memorization. When trained on both tasks jointly, no model (regardless of size) succeeds at extrapolation. These findings suggest that pre-training may intrinsically favor one learning mode over the other. By isolating these dynamics in a controlled setting, our study offers insight into how model capacity shapes learning behavior and offers broader implications for the design and deployment of small language models.