Compression Forcing: Hierarchical History Tokenization Enables Flexible World Models
cs.CV
Submitted: 2026-11-02
Updated: 2026-11-02
Comments: 37 pages, 16 figures
Code: https://github.com/eloialonso/diamond2https:
License: http://creativecommons.org/licenses/by/4.0/
The gist: Effective world modeling requires compressing salient history information that facilitates future prediction.
Terminology
Abstract
Effective world modeling requires compressing salient history information that facilitates future prediction. However, most existing approaches inherit tokenization and conditioning mechanisms from video generation, where objectives are centered around perceptual realism rather than predictive utility. This discrepancy limits the world model's effectiveness, especially in environments dominated by visual nuisances or task-irrelevant background details. To this end, we introduce Compression Forcing (CF), a general framework that aligns history compression with the predictive objective of world modeling. Compression Forcing features Dynamics-Aware Tokenization (DAT), a novel tokenization method that compresses history observations into a sequence ordered by decreasing dynamics saliency. Built on a diffusion backbone, this structure enables Hierarchical History Guidance (HG-H), allowing selective amplification of dynamics-salient information during inference. Extensive experiments across various environments demonstrate improved prediction fidelity, physical realism, and downstream decision-making performance, establishing Compression Forcing as a highly flexible and effective approach for world modeling.
Sources
- Stochastic Interpolants: A Unifying Framework for Flows and Diffusions
- Building Normalizing Flows with Stochastic Interpolants
- Diffusion for World Modeling: Visual Details Matter in Atari
- Navigation World Models
- A Short Note about Kinetics-600
- Open X-Embodiment: Robotic Learning Datasets and RT-X Models
- Self-Forcing++: Towards Minute-Scale High-Quality Video Generation
- Ctrl-World: A Controllable Generative World Model for Robot Manipulation
- World Models
- Mastering Diverse Domains through World Models
- Training Agents Inside of Scalable World Models
- Classifier-Free Diffusion Guidance
- Vid2World: Crafting Video Diffusion Models to Interactive World Models
- Nano World Models: A Minimalist Implementation of Future Video Prediction
- Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion
- CoTracker3: Simpler and Better Point Tracking by Pseudo-Labelling Real Videos
- Cosmos Policy: Fine-Tuning Video Models for Visuomotor Control and Planning
- Understanding Diffusion Objectives as the ELBO with Simple Data Augmentation
- Segment Anything
- Latte: Latent Diffusion Transformer for Video Generation
Related papers
- TEA: Text Encoder Alignment for Robust Concept Erasure in Text-to-Image Models
- Do Egocentric Video-Language Models Capture Both Hand- and Object-Centric Cues?
- Aura: Consistent Multi-Subject Video Generation via VLM-Grounded Semantic Alignment
- PVRF: All-in-one Adverse Weather Removal via Prior-modulated and Velocity-constrained Rectified Flow
- DLWM: Diverse Latent World Models for Efficient Multimodal Reasoning
- Out of Sight, Still in Mind: Token Compression for Omni-LLMs