Dentate gyrus of large language models: A scenario cognition diagnosis and enhancement framework based on pattern separation

Boxiang, Ma, Ru, Li, Shaoru, Guo, Yinghao, Li and Gutiérrez Basulto, Víctor ORCID: https://orcid.org/0000-0002-6117-5459 2026. Dentate gyrus of large language models: A scenario cognition diagnosis and enhancement framework based on pattern separation. Journal of Computer Research and Development 63 (9) , pp. 2187-2201. 10.7544/issn1000-1239.202660437

Preview

PDF - Accepted Post-Print Version Available under License Creative Commons Attribution. Download (5MB) | Preview

Abstract

Large language models (LLMs) exhibit systematic generalization limitations in scenario cognition tasks, revealing potential functional deficiencies in their knowledge memorization processes. Inspired by the pattern separation mechanism of the dentate gyrus in the human hippocampus, we hypothesize that failures in scenario cognition are associated with limitations in the internal pattern separation capability of LLMs. Based on this hypothesis, we first design a cross-entity representation diagnosis (CERD) method, which quantitatively measures entity representation confusion across different layers without additional training. Our analysis reveals that standard autoregressive training fails to induce sufficiently effective pattern separation capabilities, leaving substantial cross-entity correlation confusion within LLM representations. Building upon this observation, we further propose latent-space pattern separation (LPS), which explicitly incorporates pattern separation constraints into the training process by introducing a triplet-based contrastive regularization at target layers with severe entity confusion, thereby reducing correlations among easily confused knowledge representations. Extensive experiments across LLMs ranging from 0.8B to 9B parameters demonstrate that LPS consistently improves scenario cognition capabilities. Compared with standard autoregressive training, our method achieves up to 101.5% relative improvement in exact match (EM), providing a new perspective for understanding and enhancing scenario cognition and knowledge memorization mechanisms in LLMs.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Schools > Computational & Mathematical Sciences Schools > Computer Science & Informatics
Additional Information:	RRS policy applied
Language other than English:	Chinese
Date of First Compliant Deposit:	29 October 2026
Last Modified:	29 Oct 2026 12:00
URI:	https://orca.cardiff.ac.uk/id/eprint/190510

Actions (repository staff only)

Edit Item