mlsanityData to 7 Oct 2026

EDELINE: Enhancing Memory in Diffusion-based World Models via Linear-Time Sequence Modeling

arXiv 2502.00466 ↗First posted 1 Feb 2025cs.LGcited by 3

On the map: World models → Latent dynamics models and model-based RL

Abstract (arXiv)

World models represent a promising approach for training reinforcement learning agents with significantly improved sample efficiency. While most world model methods primarily rely on sequences of discrete latent variables to model environment dynamics, this compression often neglects critical visual details essential for reinforcement learning. Recent diffusion-based world models condition generation on a fixed context length of frames to predict the next observation, using separate recurrent neural networks to model rewards and termination signals. Although this architecture effectively enhances visual fidelity, the fixed context length approach inherently limits memory capacity. In this paper, we introduce EDELINE, a unified world model architecture that integrates state space models with diffusion models. Our approach outperforms existing baselines across visually challenging Atari 100k tasks, memory-demanding Crafter benchmark, and 3D first-person ViZDoom environments, demonstrating superior performance in all these diverse challenges.

Key facts. Cited by 3 papers in the map (data through 7 Oct 2026). Placed in Latent dynamics models and model-based RL, one of 11 directions in World models. Builds on 58 of its 72 references that are in the map. First posted on arXiv 1 Feb 2025.

Builds on

58 of 72 references.

Closest ancestors · 15

Same direction, newest first

  1. 2024Drama: Mamba-Enabled Model-Based Reinforcement Learning Is Sample and Parameter Efficient11 Oct
  2. 2024Efficient World Models with Context-Aware Tokenization27 Jun
  3. 2024Diffusion for World Modeling: Visual Details Matter in Atari20 May
  4. 2024Mastering Memory Tasks with World Models7 Mar
  5. 2024Improving Token-Based World Models with Parallel Observation Prediction8 Feb
  6. 2023STORM: Efficient Stochastic Transformer based World Models for Reinforcement Learning14 Oct
  7. 2023HarmonyDream: Task Harmonization Inside World Models30 Sep
  8. 2023Transformer-based World Models Are Happy With 100k Interactions13 Mar
  9. 2023Mastering Diverse Domains through World Models10 Jan
  10. 2022Transformers are Sample-Efficient World Models1 Sep
  11. 2022DayDreamer: World Models for Physical Robot Learning28 Jun
  12. 2021Mastering Atari Games with Limited Data30 Oct
3 more
  1. 2020Mastering Atari with Discrete World Models5 Oct
  2. 2019Dream to Control: Learning Behaviors by Latent Imagination3 Dec
  3. 2018Recurrent World Models Facilitate Policy Evolution4 Sep

Other references · 43

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2021High-Resolution Image Synthesis with Latent Diffusion Modelscited by 18,694
  4. 2020Denoising Diffusion Implicit Modelscited by 9,655
  5. 2020Score-Based Generative Modeling through Stochastic Differential Equationscited by 8,964
  6. 2021Diffusion Models Beat GANs on Image Synthesiscited by 8,185
  7. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  8. 2023Mamba: Linear-Time Sequence Modeling with Selective State Spacescited by 4,616
  9. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  10. 2022Elucidating the Design Space of Diffusion-Based Generative Modelscited by 2,994
  11. 2021Efficiently Modeling Long Sequences with Structured State Spacescited by 2,424
  12. 2014Empirical Evaluation of Gated Recurrent Neural Networks on Sequence Modelingcited by 2,341
31 more
  1. 2021Decision Transformer: Reinforcement Learning via Sequence Modelingcited by 1,528
  2. 2019Mastering Atari, Go, Chess and Shogi by Planning with a Learned Modelcited by 1,327
  3. 2024Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Dualitycited by 1,283
  4. 2020Improved Techniques for Training Score-Based Generative Modelscited by 1,036
  5. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,023
  6. 2021Deep Reinforcement Learning at the Edge of the Statistical Precipicecited by 781
  7. 2022Simplified State Space Layers for Sequence Modelingcited by 776
  8. 2015Dueling Network Architectures for Deep Reinforcement Learningcited by 769
  9. 2024Genie: Generative Interactive Environmentscited by 706
  10. 2020HiPPO: Recurrent Memory with Optimal Polynomial Projectionscited by 600
  11. 2022Diffusion Policies as an Expressive Policy Class for Offline Reinforcement Learningcited by 543
  12. 2022Is Conditional Generative Modeling all you need for Decision-Making?cited by 504
  13. 2019Model-Based Reinforcement Learning for Ataricited by 495
  14. 2022On the Parameterization and Initialization of Diagonal State Space Modelscited by 418
  15. 2022Diagonal State Spaces are as Effective as Structured State Spacescited by 350
  16. 2023Minigrid & Miniworld: Modular & Customizable Reinforcement Learning Environments for Goal-Oriented Taskscited by 320
  17. 2024Diffusion Models Are Real-Time Game Enginescited by 282
  18. 2023Imitating Human Behaviour with Diffusion Modelscited by 260
  19. 2023Reduce, Reuse, Recycle: Compositional Generation with Energy-Based Diffusion Models and MCMCcited by 216
  20. 2021Benchmarking the Spectrum of Agent Capabilitiescited by 193
  21. 2016ViZDoom: A Doom-based AI Research Platform for Visual Reinforcement Learningcited by 178
  22. 2023AdaptDiffuser: Diffusion Models as Adaptive Self-evolving Plannerscited by 147
  23. 2022Liquid Structural State-Space Modelscited by 134
  24. 2023Structured State Space Models for In-Context Reinforcement Learningcited by 109
  25. 2023Synthetic Experience Replaycited by 100
  26. 2020Learning to Simulate Dynamic Environments with GameGANcited by 86
  27. 2023MetaDiffuser: Diffusion Model as Conditional Planner for Offline Meta-RLcited by 65
  28. 2022Denoising Likelihood Score Matching for Conditional Score-based Data Generationcited by 45
  29. 2024Decision Mamba: Reinforcement Learning via Hybrid Selective Sequence Modelingcited by 18
  30. 2023Extracting Reward Functions from Diffusion Modelscited by 17
  31. 2018Improving the Closed-Loop Tracking Performance Using the First-Order Hold Sensing Technique with Experimentscited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2025: 1; 2026: 2.

3 papers.

  1. 2026World Models: A Comprehensive Survey of Architectures, Methodologies, Reasoning Paradigms, and Applicationscited by 3
  2. 2025On Memory: A comparison of memory mechanisms in world modelscited by 2
  3. 2026HaM-World: Soft-Hamiltonian World Models with Selective Memory for Planningcited by 1

Similar, not cited

Close in wording, no citation either way

  1. 2026Composition of Memory Experts for Diffusion World Models
  2. 2026Mixture-of-World Models: Scaling Multi-Task Reinforcement Learning with Modular Latent Dynamics
  3. 2020Smaller World Models for Reinforcement Learning
  4. 2023Facing Off World Model Backbones: RNNs, Transformers, and S4
  5. 2019Continual Learning Using World Models for Pseudo-Rehearsal
  6. 2024Learning to Play Atari in a World of Tokens
  7. 2025Zero-shot World Models via Search in Memory
  8. 2026Scaling World-Model Reinforcement Learning Through Diffusion Policy Optimization