# HERO: Hierarchical Extrapolation and Refresh for Efficient World Models

arXiv: https://arxiv.org/abs/2508.17588 · HTML: https://mlsanity.com/paper/2508.17588/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- Placed in Video-generation world models and simulators, one of 11 directions in World models.
- Builds on 33 of its 37 references that are in the map.
- First posted on arXiv 25 Aug 2025.

## Placed in

- World models / [Video-generation world models and simulators](https://mlsanity.com/wm/video-generation-world-models-and-simulators/)
- Diffusion models / [Video generation](https://mlsanity.com/diffusion/video-generation/)

## Builds on (most cited first, 30 of 33)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [LLaMA: Open and Efficient Foundation Language Models](https://mlsanity.com/paper/2302.13971/) (2023)
- [Llama 2: Open Foundation and Fine-Tuned Chat Models](https://mlsanity.com/paper/2307.09288/) (2023)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness](https://mlsanity.com/paper/2205.14135/) (2022)
- [Qwen Technical Report](https://mlsanity.com/paper/2309.16609/) (2023)
- [Wan: Open and Advanced Large-Scale Video Generative Models](https://mlsanity.com/paper/2503.20314/) (2025)
- [CogVideoX: Text-to-Video Diffusion Models with An Expert Transformer](https://mlsanity.com/paper/2408.06072/) (2024)
- [DPM-Solver: A Fast ODE Solver for Diffusion Probabilistic Model Sampling in Around 10 Steps](https://mlsanity.com/paper/2206.00927/) (2022)
- [VBench: Comprehensive Benchmark Suite for Video Generative Models](https://mlsanity.com/paper/2311.17982/) (2023)
- [HunyuanVideo: A Systematic Framework For Large Video Generative Models](https://mlsanity.com/paper/2412.03603/) (2024)
- [Dream to Control: Learning Behaviors by Latent Imagination](https://mlsanity.com/paper/1912.01603/) (2019)
- [DUSt3R: Geometric 3D Vision Made Easy](https://mlsanity.com/paper/2312.14132/) (2023)
- [Mastering Diverse Domains through World Models](https://mlsanity.com/paper/2301.04104/) (2023)
- [One-step Diffusion with Distribution Matching Distillation](https://mlsanity.com/paper/2311.18828/) (2023)
- [Cosmos World Foundation Model Platform for Physical AI](https://mlsanity.com/paper/2501.03575/) (2025)
- [Continuous 3D Perception Model with Persistent State](https://mlsanity.com/paper/2501.12387/) (2025)
- [Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models](https://mlsanity.com/paper/2402.17177/) (2024)
- [DayDreamer: World Models for Physical Robot Learning](https://mlsanity.com/paper/2206.14176/) (2022)
- [Navigation World Models](https://mlsanity.com/paper/2412.03572/) (2024)
- [Post-training Quantization on Diffusion Models](https://mlsanity.com/paper/2211.15736/) (2022)
- [Structural Pruning for Diffusion Models](https://mlsanity.com/paper/2305.10924/) (2023)
- [GAIA-2: A Controllable Multi-View Generative World Model for Autonomous Driving](https://mlsanity.com/paper/2503.20523/) (2025)
- [Accelerating Diffusion Transformers with Token-wise Feature Caching](https://mlsanity.com/paper/2410.05317/) (2024)
- [Transformer-based World Models Are Happy With 100k Interactions](https://mlsanity.com/paper/2303.07109/) (2023)
- [From Reusing to Forecasting: Accelerating Diffusion Models with TaylorSeers](https://mlsanity.com/paper/2503.06923/) (2025)
- [Token Fusion: Bridging the Gap between Token Pruning and Token Merging](https://mlsanity.com/paper/2312.01026/) (2023)
- [Adaptive Caching for Faster Video Generation with Diffusion Transformers](https://mlsanity.com/paper/2411.02397/) (2024)
- [PeRFlow: Piecewise Rectified Flow as Universal Plug-and-Play Accelerator](https://mlsanity.com/paper/2405.07510/) (2024)
- [Aether: Geometric-Aware Unified World Modeling](https://mlsanity.com/paper/2503.18945/) (2025)

## Built on by (most cited first, 9 of 9)

- [A Survey on Cache Methods in Diffusion Models: Toward Efficient Multi-Modal Generation](https://mlsanity.com/paper/2510.19755/) (2025)
- [OmniShow: Unifying Multimodal Conditions for Human-Object Interaction Video Generation](https://mlsanity.com/paper/2604.11804/) (2026)
- [SceneDecorator: Towards Scene-Oriented Story Generation with Scene Planning and Scene Consistency](https://mlsanity.com/paper/2510.22994/) (2025)
- [Efficient Video Diffusion Models: Advancements and Challenges](https://mlsanity.com/paper/2604.15911/) (2026)
- [InnoAds-Composer: Efficient Condition Composition for E-Commerce Poster Generation](https://mlsanity.com/paper/2603.05898/) (2026)
- [WorldCache: Accelerating World Models for Free via Heterogeneous Token Caching](https://mlsanity.com/paper/2603.06331/) (2026)
- [Video Generation Models as World Models: Efficient Paradigms, Architectures and Algorithms](https://mlsanity.com/paper/2603.28489/) (2026)
- [WorldDynCache: Risk-Controlled Latent Dynamics Approximation for Diffusion World Model](https://mlsanity.com/paper/2608.01845/) (2026)
- [SpectralCache: Accelerating Diffusion-Based World Models via Spectral Feature Caching](https://mlsanity.com/paper/2610.02660/) (2026)

## Abstract (arXiv)

Generation-driven world models create immersive virtual environments but suffer slow inference due to the iterative nature of diffusion models. While recent advances have improved diffusion model efficiency, directly applying these techniques to world models introduces limitations such as quality degradation. In this paper, we present HERO, a training-free hierarchical acceleration framework tailored for efficient world models. Owing to the multi-modal nature of world models, we identify a feature coupling phenomenon, wherein shallow layers exhibit high temporal variability, while deeper layers yield more stable feature representations. Motivated by this, HERO adopts hierarchical strategies to accelerate inference: (i) In shallow layers, a patch-wise refresh mechanism efficiently selects tokens for recomputation. With patch-wise sampling and frequency-aware tracking, it avoids extra metric computation and remain compatible with FlashAttention. (ii) In deeper layers, a linear extrapolation scheme directly estimates intermediate features. This completely bypasses the computations in attention modules and feed-forward networks. Our experiments show that HERO achieves a 1.73× speedup with minimal quality degradation, significantly outperforming existing diffusion acceleration methods.