English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Zendy Plus

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Zendy Tools

Zendy Open

What is the relationship between model architecture and the ability toperform in-context learning? In this empirical study, we take the first stepstoward answering this question. We evaluate thirteen model architecturescapable of causal language modeling across a suite of synthetic in-contextlearning tasks. These selected architectures represent a broad range ofparadigms, including recurrent and convolution-based neural networks,transformers, state space model inspired, and other emerging attentionalternatives. We discover that all the considered architectures can performin-context learning under a wider range of conditions than previouslydocumented. Additionally, we observe stark differences in statisticalefficiency and consistency by varying the number of in-context examples andtask difficulty. We also measure each architecture's predisposition towardsin-context learning when presented with the option to memorize rather thanleverage in-context examples. Finally, and somewhat surprisingly, we find thatseveral attention alternatives are sometimes competitive with or betterin-context learners than transformers. However, no single architecturedemonstrates consistency across all tasks, with performance either plateauingor declining when confronted with a significantly larger number of in-contextexamples than those encountered during gradient-based training.

Is attention required for ICL? Exploring the Relationship Between Model Architecture and In-Context Learning Ability