cs.NENov 19, 2026

Hierarchical Spatiotemporal World Modeling with Content-Adaptive Spatial Addressing

Authors: Tiantian Li, Junfeng Lu, Yifei Wang, Zhaoyang Zhang, Lei Wang, Tianyi Qian

Organizations: QiYuanLab · Peking University · Academy of Military Sciences

Abstract

Long-horizon world modeling in partially observable environments requires compact memories that preserve spatial structure despite accumulated localization drift. Inspired by hippocampal episodic and spatial memory and the multi-periodic coding of entorhinal grid cells, we propose HiSTFormer, a hierarchical spatiotemporal world model, which adopts a layer-wise organization of local temporal RoPE and multi-periodic spatial RoPE, together with content-conditioned address offsets, to capture recent temporal context and long-range spatial structure. A fixed-capacity memory keeps KV storage and per-step attention cost bounded as the sequence grows. On MemoryMaze and HM3D, HiSTFormer outperforms strong world-model baselines within the training sequence length and generalizes substantially better to prediction horizons far beyond those seen during training, while using a compact model. The hierarchical spatiotemporal design remains effective across deeper Transformer stacks, with consistent gains as depth increases. These results show that hierarchical spatiotemporal organization and content-adaptive spatial addressing provide effective inductive biases for efficient long-horizon world modeling.

Figures & tables

Appendix figures & tables16 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

CardsList
  1. Composition of Memory Experts for Diffusion World Models

    May 12, 2026Sebastian Stapf, Pablo Acuaviva Huertos, Aram Davtyan +1World Model LearningMemory-Augmented Agents

  2. Spatial Memory Intelligence: Endowing World Models with Understanding-Driven Long-Term Memory

    Oct 1, 2026Ying Yang, Guiyu Zhang, Lianghua Huang +5Memory-Augmented Video GenerationWorld Models

  3. MemWM: Associative Memory-Enhanced Latent World Models for Long-Horizon Planning

    Oct 14, 2026Yuwei Sun, Siyu ZhuWorld Model LearningWorld Models for Robotics