Revision, Not Restart: Revisable Visual Plans for Closed-Loop World-Action Models
cs.RO, cs.CV
Submitted: 2026-09-28
Updated: 2026-09-28
Comments: 27 pages, 4 figures. Project Page: https://PLACEHOLDER.github.io/RTP/
Project page: https://placeholder.github.io/RTP
License: http://creativecommons.org/licenses/by/4.0/
The gist: World-action models use predicted visual futures to condition robot actions, yet execution feedback can invalidate parts of a prediction while leaving its task structure useful.
Terminology
Abstract
World-action models use predicted visual futures to condition robot actions, yet execution feedback can invalidate parts of a prediction while leaving its task structure useful. We propose Revisable Temporal Planning (RTP), which maintains the visual future as a persistent action condition and revises it after feedback. Its central mechanism is a learned revision bridge: it resumes an intermediate state saved during visual generation and adapts its continuation to current observations. Visual and action supervision connect this revision to subsequent control. Time-aware history supplies observed evidence, and an adaptive policy selects retention, bridge revision, or fresh replanning from new noise before decoding the next action. On RoboMME and RMBench, RTP achieves task-averaged success rates of 48.6% and 84.8%, respectively. Matched comparisons support learned continuation; estimated checkpoint-source and action-prefix effects are positive but less precisely resolved. These results connect feedback-driven visual-plan revision to closed-loop task performance. Project Page: https://PLACEHOLDER.github.io/RTP/
Sources
- Feedback World Model Enables Precise Guidance of Diffusion Policy
- Motus: A Unified Latent Action World Model
- AHA-WAM:Asynchronous Horizon-Adaptive World-Action Modeling with Observation-Guided Context Routing
- Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusion
- LaWAM: Latent World Action Models for Efficient Dynamics-Aware Robot Policies
- RoboTwin 2.0: A Scalable Data Generator and Benchmark with Strong Domain Randomization for Robust Bimanual Robotic Manipulation
- RMBench: Memory-Dependent Robotic Manipulation Benchmark with Insights into Policy Design
- FlowWAM: Optical Flow as a Unified Action Representation for World Action Models
- Diffusion Policy: Visuomotor Policy Learning via Action Diffusion
- RoboMME: Benchmarking and Understanding Memory for Robotic Generalist Policies
- SAM2Act: Integrating Visual Foundation Model with A Memory Architecture for Robotic Manipulation
- Unified 4D World Action Modeling from Video Priors with Asynchronous Denoising
- Chameleon: Control-Indexed Prospective Memory for Visuomotor Manipulation
- Streaming Diffusion Policy: Fast Policy Synthesis with Variable Noise Diffusion Models
- FutureRTC: Real-Time Robot Execution with Anticipatory-Conditioned Action Chunking
- OpenVLA: An Open-Source Vision-Language-Action Model
- Diffusion ReRoll: Revisable Denoising for Robotic Sequential Prediction
- Causal World Modeling for Robot Control
- OVO-S-Bench: A Hierarchical Benchmark for Streaming Spatial Intelligence in Multimodal LLMs
- Light-WAM: Efficient World Action Models with State-Fusion Action Decoding
Related papers
- FMT x: An Efficient and Asymptotically Optimal Extension of the Fast Marching Tree for Dynamic Replanning
- MPCFormer: A physics-informed data-driven approach for explainable socially-aware autonomous driving
- RoboLab: A High-Fidelity Simulation Benchmark for Analysis of Task Generalist Policies
- ARC: A Reasoning Recipe for Robot Foundation Models
- HRDexDB: A 4D Dexterous Grasping Dataset Across Human and Multiple Robot Embodiments
- APT: Action Expert Pretraining Improves Instruction Generalization of Vision-Language-Action Policies