analiza aiqbench
The paper addresses a practical problem in learned robot control: action-chunking policies often regenerate entire action sequences when only a short stretch is wrong. CMAP instead uses an uncertainty score to select one contiguous span and regenerate just that segment, conditioned on executed history and retained predictions. Results are reported on 13 tasks (nine simulated, four real-world), but the gains are described as overall performance over baselines, so readers should check per-task numbers and the real-world setup before drawing conclusions.
Predict in Chunks, Revise in Spans: History-conditioned Action Regeneration for Robotic Manipulation
arXiv:2610.14864v1 Announce Type: new Abstract: Action chunking improves temporal coherence in visuomotor imitation learning by jointly predicting sequences of future actions. However, many existing policies regenerate entire action chunks even when errors are localized, potentially replacing otherwise valid predictions. We introduce Contextual Masked Action Policy (CMAP), a visuomotor policy that refines continuous action chunks through history-conditioned local regeneration. Specifically, CMAP introduces a history-conditioned action tape that combines executed commands with masked future positions to produce coarse action predictions and per-step uncertainty scores. An uncertainty-guided span regeneration mechanism then selects a high-uncertainty contiguous span and reconstructs it using the executed history and retained future predictions as context. By jointly reconstructing neighboring actions, CMAP models local temporal dependencies while preserving all predictions outside the selected span. Extensive experiments on nine simulated and four real-world manipulation tasks demonstrate that CMAP achieves superior overall performance over several competitive baselines. Ablation studies validate the complementary contributions of its key components, while held-out trajectory analysis shows that local regeneration reduces prediction errors within selected spans.
Abstract:Action chunking improves temporal coherence in visuomotor imitation learning by jointly predicting sequences of future actions. However, many existing policies regenerate entire action chunks even when errors are localized, potentially replacing otherwise valid predictions. We introduce Contextual Masked Action Policy (CMAP), a visuomotor policy that refines continuous action chunks through history-conditioned local regeneration. Specifically, CMAP introduces a history-conditioned action tape that combines executed commands with masked future positions to produce coarse action predictions and per-step uncertainty scores. An uncertainty-guided span regeneration mechanism then selects a high-uncertainty contiguous span and reconstructs it using the executed history and retained future predictions as context. By jointly reconstructing neighboring actions, CMAP models local temporal dependencies while preserving all predictions outside the selected span. Extensive experiments on nine simulated and four real-world manipulation tasks demonstrate that CMAP achieves superior overall performance over several competitive baselines. Ablation studies validate the complementary contributions of its key components, while held-out trajectory analysis shows that local regeneration reduces prediction errors within selected spans.
Robotics (cs.RO)
arXiv:2610.14864 [cs.RO]
(or arXiv:2610.14864v1 [cs.RO] for this version)
https://doi.org/10.48550/arXiv.2610.14864 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 257 cuvinte
Subiecte conexe
Building trust in AI agents: A production-ready evaluation framework with Amazon Bedrock AgentCore
AWS · 16 oct. 2026
Debugging Logprob Mismatches in LLM Reinforcement Learning
AMD · 23 sept. 2026
TacPAC: Tactile Prediction and Real-Time Action Correction in World-Action Models for Contact-Rich Manipulation
arXiv · 7 sept. 2026
ForceDelta-VLA: Distilling Force-Conditioned ActionCorrections for Contact-Rich Manipulation
arXiv · 17 sept. 2026