Context-Gated Associative Retrieval: From Theory to Transformers

arXiv:2605.1097035.9
Predicted impact top 64% in DIS-NN · last 90 daysOriginality Highly original
AI Analysis

Provides a mechanistic link between associative memory theory and transformer in-context learning, offering a theoretical foundation for LLM phenomenology.

The paper proposes a context-gated associative memory architecture that improves retrieval exponentially by reshaping the energy landscape, and shows that in-context learning in transformers like Llama-3 acts as context-gated retrieval.

Hopfield networks and their generalizations have established deep connections among biological associative memories, statistical physics, and transformers. Yet most models treat retrieval as a fixed query-to-memory mapping, ignoring the role of external context in recall. In this work, we propose a two-stage associative memory architecture, wherein a context-gate subcircuit reshapes the retrieval energy landscape before and during recall. We show theoretically that context gating increases inter-memory separation while inducing sparsity, translating into exponential improvements in retrieval. Crucially, we prove that the system admits a unique self-consistent fixed point, revealing that the resulting retrieval state is driven by both a direct contextual bias and a second-order retrieval-gate feedback loop. We then bridge this theory to transformers; specifically, we evaluate a first-order approximation on Llama-3, confirming that in-context learning acts as context-gated retrieval. Native dynamics mirror our theory: context localizes a memory subspace, enabling the zero-shot query to cleanly discriminate. Ultimately, this framework provides a mechanistic link between associative memory theory and LLM phenomenology.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes