Fast Inference-Time Policy Refinement via Action-Affine World Models
cs.RO
Submitted: 2026-11-11
Updated: 2026-11-19
Terminology
Sources
- Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
- Residual Policy Learning
- Residual Reinforcement Learning for Robot Control
- From Imitation to Refinement -- Residual RL for Precise Assembly
- Policy Decorator: Model-Agnostic Online Refinement for Large Policy Model
- Robust Online Residual Refinement via Koopman-Guided Dynamics Modeling
- Steering Your Generalists: Improving Robotic Foundation Models via Value Guidance
- Latent Policy Barrier: Learning Robust Visuomotor Policies by Staying In-Distribution
- Mastering Diverse Domains through World Models
- TD-MPC2: Scalable, Robust World Models for Continuous Control
- Deep Visual Foresight for Planning Robot Motion
- DINO-WM: World Models on Pre-trained Visual Features enable Zero-shot Planning
- LaDi-WM: A Latent Diffusion-based World Model for Predictive Manipulation
- Inference-Time Enhancement of Generative Robot Policies via Predictive World Modeling
- Embed to Control: A Locally Linear Latent Dynamics Model for Control from Raw Images
- A Data-Driven Approximation of the Koopman Operator: Extending Dynamic Mode Decomposition
- Deep learning for universal linear embeddings of nonlinear dynamics
- Model-Based Control Using Koopman Operators
- Imitation Learning with Limited Actions via Diffusion Planners and Deep Koopman Controllers
- Linear predictors for nonlinear dynamical systems: Koopman operator meets model predictive control
Related papers
- Rethinking Actions: Continuity-Aware Priors for World-Model Planning
- MOSAIC: Modular Obstacle and Spatial-Aware Framework for Manipulation In Cluttered Scenes
- SAGAS: Semantic-Aware Graph-Assisted Stitching for Offline Temporal Logic Planning
- DexSeed: Expanding Human Hand-Object Interactions into Robot Demonstrations for Dexterous Policy Learning
- LaST0.5: A Generalist World Action Model with Interaction-Centric Latent Reasoning
- CIALLO: Learning Shared Intent Representations for Cross-Embodiment Manipulation