Is attention required for ICL? Exploring the Relationship Between Model Architecture and In-Context Learning Ability
This work addresses the problem of understanding architectural impacts on in-context learning for AI researchers, providing empirical insights but is incremental as it builds on existing knowledge without introducing new methods.
The study explored how different model architectures affect in-context learning ability, finding that all tested architectures could perform in-context learning under broader conditions than previously known, with some attention alternatives being competitive or better than transformers in certain tasks, though no architecture was consistent across all tasks.
What is the relationship between model architecture and the ability to perform in-context learning? In this empirical study, we take the first steps toward answering this question. We evaluate thirteen model architectures capable of causal language modeling across a suite of synthetic in-context learning tasks. These selected architectures represent a broad range of paradigms, including recurrent and convolution-based neural networks, transformers, state space model inspired, and other emerging attention alternatives. We discover that all the considered architectures can perform in-context learning under a wider range of conditions than previously documented. Additionally, we observe stark differences in statistical efficiency and consistency by varying the number of in-context examples and task difficulty. We also measure each architecture's predisposition towards in-context learning when presented with the option to memorize rather than leverage in-context examples. Finally, and somewhat surprisingly, we find that several attention alternatives are sometimes competitive with or better in-context learners than transformers. However, no single architecture demonstrates consistency across all tasks, with performance either plateauing or declining when confronted with a significantly larger number of in-context examples than those encountered during gradient-based training.