DecMem: Towards Minute-Long Consistent World Generation with Decoupled Memory

arXiv 2026 53.2 method

TLDR

Proposes DecMem, a decoupled memory architecture for minute-long consistent video generation with fine-grained spatio-temporal consistency.

Reasoning

The paper introduces a novel decoupled memory approach to address long-horizon consistency in video generation, which is a strength. However, the abstract lacks specific quantitative results or dataset details, and the focus is narrow (video generation) without addressing interactive or RL settings.

Read-first score

Read-first score 53.2, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 37.

Recency 6%

100

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

Citation impact 18%

90.1

Uses OpenAlex-shaped citation metadata as a bibliometric attention signal, separate from paper quality. citation_normalized_percentile=0.90115039

Methodology quality 18%

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

Topical relevance 29%

52.9

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

FoundationFrontierBridge

Rank sensitivity

Stability: volatile; rank range: 378.

Keyword Scores

world model

video world model

generative world model

world dynamics prediction

world simulator

interactive world model

model-based reinforcement learning world model

Deep Analysis

Innovations

Decoupled memory architecture with Sparse Global Memory for efficient fine-grained access to global history and Anchored Local Memory for stable and high-quality extrapolation
Identification and systematic analysis of attention dispersion as a fundamental limitation of naïve learnable memory architectures in long-horizon extrapolation
Fine-grained, learnable, and scalable memory for consistent world generation, moving beyond explicit 3D memory and coarse frame-level implicit modeling

Methodology

The paper proposes DecMem, a decoupled memory architecture that employs Sparse Global Memory for efficient fine-grained access to global history and Anchored Local Memory for stable and high-quality extrapolation. It addresses two fundamental limitations of naïve learnable memory: computational inefficiency and attention dispersion. The model is evaluated on long video generation tasks, comparing to state-of-the-art methods.

Key Results

DecMem significantly outperforms current state-of-the-art methods, enabling minute-level controllable long video generation with high fidelity and consistency.

Resources

Paper