Structure Abstraction and Generalization in a Hippocampal-Entorhinal Inspired World Model

arXiv 2026 49.6 method

TLDR

A brain-inspired hierarchical model extracts abstract structures from dynamics using HPC-MEC coupling for structural generalization.

Reasoning

The paper presents a novel computational framework inspired by hippocampal-entorhinal circuits for structural abstraction and generalization in world models. Its strength lies in the biologically plausible architecture and demonstrated capacity for structural reuse, but it is limited to primitive transformation benchmarks without real-world validation or interactive/RL contexts.

Read-first score

Read-first score 49.6, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 28.

Recency 6%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Citation impact 18%

80.9

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.8088977

Methodology quality 18%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=baseline,benchmark

Topical relevance 29%

Uses existing LLM keyword relevance scores normalized to 0-100. world model,world simulator,generative world model,interactive world model,video world model,world dynamics prediction,model-based reinforcement learning world model

Reproducibility 18%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Citation velocity 12%

Citation velocity estimates citations per publication-year to reduce old-paper bias. velocity=0.00

Field roles

FoundationFrontierBridgeMethodology anchor

Rank sensitivity

Stability: volatile; rank range: 296.

Keyword Scores

world model

world dynamics prediction

generative world model

video world model

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Simultaneous inference of latent transitions and construction of a predictive visual world model in a brain-inspired hierarchical architecture
Dissociation of relational structures (MEC) from integrated episodic scenes (HPC) via an HPC-MEC coupling model
Use of velocity-driven path integration to enable structural abstraction and reuse across diverse contexts

Methodology

The paper proposes a hierarchical model combining an inverse model for structural extraction with an HPC-MEC coupling mechanism. The model is trained using self-supervised learning on primitive transformation dynamics, leveraging velocity-driven path integration to infer latent transitions and generate predictions.

Key Results

The model demonstrates capacity for structural abstraction on primitive transformation dynamics and achieves structural generalization through robust prediction and structural reuse across different contexts.

Limitations

Only evaluated on primitive transformation dynamics, leaving generalization to complex real-world scenarios unaddressed
No explicit comparison to alternative models or baselines is mentioned in the abstract
Biological plausibility of the specific computational mechanisms (e.g., inverse model, coupling) is not validated against neural data

Resources

Paper