mlsanityData to 7 Oct 2026

Inference-Time Enhancement of Generative Robot Policies via Predictive World Modeling

arXiv 2502.00622 ↗First posted 2 Feb 2025cs.RO cs.CV cs.LGcited by 22

On the map: World models → Embodied and robot world models

also in 1 more
Abstract (arXiv)

We present Generative Predictive Control (GPC), an inference-time method for improving pretrained behavior-cloning policies without retraining. GPC augments a frozen diffusion policy at deployment with an action-conditioned world model trained on expert demonstrations and random exploration rollouts. The world model predicts the consequences of action proposals generated by the diffusion policy and enables lightweight online planning that ranks and refines these proposals through model-based look-ahead. By combining a generative prior with predictive foresight, GPC enables test-time adaptation while keeping the original policy fixed. Across diverse robotic manipulation tasks, including state- and vision-based settings in both simulation and real-world experiments, GPC consistently outperforms standard behavior cloning and compares favorably with other inference-time adaptation baselines.

Key facts. Cited by 22 papers in the map (data through 7 Oct 2026). Placed in Embodied and robot world models, one of 11 directions in World models. Builds on 35 of its 47 references that are in the map. First posted on arXiv 2 Feb 2025.

Builds on

35 of 47 references.

Closest ancestors · 2

Same direction, newest first

  1. 2023GenSim: Generating Robotic Simulation Tasks via Large Language Models2 Oct
  2. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusion7 Mar

Other references · 33

  1. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  2. 2024OpenVLA: An Open-Source Vision-Language-Action Modelcited by 2,777
  3. 2022Video Diffusion Modelscited by 2,280
  4. 2018World Modelscited by 1,444
  5. 2023Mastering Diverse Domains through World Modelscited by 1,193
  6. 2023Learning Universal Policies via Text-Guided Video Generationcited by 506
  7. 2024Evaluating Real-World Robot Manipulation Policies in Simulationcited by 484
  8. 2023TD-MPC2: Scalable, Robust World Models for Continuous Controlcited by 482
  9. 2023Learning Interactive Real-World Simulatorscited by 440
  10. 2024DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planningcited by 357
  11. 2022Transformers are Sample-Efficient World Modelscited by 317
  12. 2024Diffusion for World Modeling: Visual Details Matter in Ataricited by 285
21 more
  1. 2016Deep Visual Foresight for Planning Robot Motioncited by 276
  2. 2024Navigation World Modelscited by 262
  3. 2023Foundation Models in Robotics: Applications, Challenges, and the Futurecited by 212
  4. 2023RoboGen: Towards Unleashing Infinite Data for Automated Robot Learning via Generative Simulationcited by 206
  5. 2015Action-Conditional Video Prediction using Deep Networks in Atari Gamescited by 171
  6. 2023Video Language Planningcited by 160
  7. 2023Learning to Act from Actionless Videos through Dense Correspondencescited by 159
  8. 2024Behavior Generation with Latent Actionscited by 159
  9. 2023Motion Planning Diffusion: Learning and Planning of Robot Motions with Diffusion Modelscited by 155
  10. 2024Gen2Act: Human Video Generation in Novel Scenarios enables Generalizable Robot Manipulationcited by 142
  11. 2020A Review on Deep Learning Techniques for Video Predictioncited by 114
  12. 2023GenAug: Retargeting behaviors to unseen situations via Generative Augmentationcited by 103
  13. 2024Steering Your Generalists: Improving Robotic Foundation Models via Value Guidancecited by 100
  14. 2024Dreamitate: Real-World Visuomotor Policy Learning via Video Generationcited by 84
  15. 2024From Imitation to Refinement -- Residual RL for Precise Assemblycited by 80
  16. 2024Diffusion-ES: Gradient-free Planning with Diffusion for Autonomous Driving and Zero-Shot Instruction Followingcited by 41
  17. 2024Potential Based Diffusion Motion Planningcited by 40
  18. 2024Deep Generative Models in Robotics: A Survey on Learning from Multimodal Demonstrationscited by 35
  19. 2024Control-oriented Clustering of Visual Latent Representationcited by 8
  20. 2017SE3-Pose-Nets: Structured Deep Dynamics Models for Visuomotor Planning and Controlcited by 6
  21. 2024Learning Multiple Initial Solutions to Optimization Problemscited by 4

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 22.

22 papers.

Sort:

By area:

  1. 2026World Action Models: The Next Frontier in Embodied AIcited by 46
  2. 2026World Model for Robot Learning: A Comprehensive Surveycited by 42
  3. 2026Self-Refining Video Samplingcited by 14
  4. 2026World Action Verifier: Self-Improving World Models via Forward-Inverse Asymmetrycited by 13
  5. 2026Masked Visual Actions for Unified World Modelingcited by 7
  6. 2026PACT-WAM: Predicting Actions and Visual Foresight with Compact Temporal Encoding for Robot Manipulationcited by 4
  7. 2026World Models for Robotic Manipulation: A Surveycited by 3
  8. 2026From World Models to World Action Models: A Concise Tutorial for Roboticscited by 3
  9. 2026FACT: Failure-Aware Causal Training for World-Action Modelscited by 3
  10. 2026Robot Self-Improvement via Human-Video Dynamics Modelscited by 2
  11. 2026Structured 4D Latent Predictive Model for Robot Planningcited by 2
  12. 2026DriftWorld: Fast World Modeling through Driftingcited by 2
10 more
  1. 2026Cortex 2.0: Grounding World Models in Real-World Industrial Deploymentcited by 1
  2. 2026DreamAvoid: Critical-Phase Test-Time Dreaming to Avoid Failures in VLA Policiescited by 1
  3. 2026PhysCoRe: Physics-Corrected Residual World Models for Material-Aware Deformable Dynamicscited by 1
  4. 2026SyncWorld: Visual Calibration Enables World Models as Zero-Shot Simulatorscited by 1
  5. 2026World-Action Models for Robot Learning and Control: A Surveycited by 1
  6. 2026ICPRL: Acquiring Physical Intuition from Interactive Controlcited by 0
  7. 2026Data-Asymmetric Latent Imagination and Reranking for 3D Robotic Imitation Learningcited by 0
  8. 2026ReGuide: From Test-Time Guidance to Self-Improving Diffusion Policiescited by 0
  9. 2026CereVLA: Cerebellum-Inspired Consequence-Aware Residual Governance for Efficient Vision-Language-Action Executioncited by 0
  10. 2026LeWAM: A JEPA World Action Model with Diffusion-Steering-Based MPCcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Where Should Action Generation Begin? A Learnable Source Prior for Generative Robot Policies
  2. 2025Generative Predictive Control: Flow Matching Policies for Dynamic and Difficult-to-Demonstrate Tasks
  3. 2025Improving Generative Behavior Cloning via Self-Guidance and Adaptive Chunking
  4. 2025Much Ado About Noising: Dispelling the Myths of Generative Robotic Control
  5. 2026Generative Control as Optimization: Time Unconditional Flow Matching for Adaptive and Robust Robotic Control
  6. 2025World4RL: Diffusion World Models for Policy Refinement with Reinforcement Learning for Robotic Manipulation
  7. 2026VLS: Steering Pretrained Robot Policies via Vision-Language Models
  8. 2026Feedback World Model Enables Precise Guidance of Diffusion Policy