Papers
arxiv:2610.21453

CarryOn: A Dual-Loop Harness for Personalized and Reliable Robot Assistance

Published on Oct 15
· Submitted by
Zixuan WANG
on Oct 16
Authors:
,
,
,
,
,
,

Abstract

CarryOn enhances frozen vision-language-action policies with persistent personalized planning and supervised execution with recovery skills to improve safety and adaptability.

While generalist vision-language-action (VLA) models have enabled diverse open-world manipulation, real-world robotic assistance requires two critical capabilities they currently lack: personalization and reliable execution. First, executing an instruction like "make me a sandwich" requires grounding subtask decomposition in persistent user preferences rather than relying on a single generic solution. Second, reliable deployment demands explainable safety assurance that proactively prevents hazards and gracefully recovers from occasional physical failures. To address these gaps, we present CarryOn, a dual-loop harness designed to augment a frozen VLA policy. The planning loop maintains persistent user-level and scene-level context that is continually grounded against new observations and revised through interaction, enabling personalized task selection and execution grounding across conversations and changing scenes. The execution loop actively supervises policy rollouts, shielding actions from predicted hazardous contact. When execution fails, the system invokes reusable control-based recovery skills that restore task progress and are refined through reflection on earlier failures. On a real robot and in simulation, CarryOn dispatches preference-consistent subtasks across users and scenes, reduces the violation rate of an already safety-finetuned VLA from 52.6% to 6.1%, and increases recovery success from 13% to 80% when refined skills are reused.

Community

Paper author Paper submitter
•
edited Oct 16

CarryOn combines personalized planning, runtime protection, and offline recovery refinement around a frozen VLA.

The planning loop maintains persistent, revisable user and scene context to guide personalized task decisions. The execution loop combines proactive geometric shielding with reactive recovery, whose recovery skills improve through offline reflection.

Two highlights from our evaluations:

  • On LIBERO-Safety, CarryOn further reduces the safety-finetuned policy’s violation rate from 52.6% to 6.1%.
  • On perturbed LIBERO-Object tasks, success rate increases from 13% with VLA alone to 80.0% after two successive reflection-and-retry rounds, with each round refining and retrying the remaining failures.

Please also check our real-robot demo to see how remembered preferences, new observations, shielding, and recovery work together within an assistance task 😊

architecture

This is an automated message from the Librarian Bot. I found the following papers similar to this paper.

The following papers were recommended by the Semantic Scholar API

Please give a thumbs up to this comment if you found it helpful!

If you want recommendations for any Paper on Hugging Face checkout this Space

You can directly ask Librarian Bot for paper recommendations by tagging it in a comment: @librarian-bot recommend

Thank you for sharing your work. CarryOn combines persistent preferences, action supervision, and recovery around a frozen VLA. Its evaluation reports preference-consistent subtasks, a reduction in violation rate from 52.6% to 6.1%, and success on perturbed LIBERO-Object tasks rising from 13% to 80% after two successive reflection-and-retry rounds on the remaining failures.

We would like to discuss your work in Section 3.4 (“Skill- and agent-based execution”) of our survey, In-Context Learning for Robots: Methods and Applications, to explain how task-level personalization, action supervision, and reusable recovery affect different stages of execution, with the recovery results also informing our failure-correction discussion.

If the survey offers a useful reference for your discussion of personalized robot assistance, we would appreciate your considering it for citation.

Sign up or log in to comment

Get this paper in your agent:

hf papers read 2610.21453
Don't have the latest CLI?
curl -LsSf https://hf.co/cli/install.sh | bash

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2610.21453 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2610.21453 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2610.21453 in a Space README.md to link it from this page.

Collections including this paper 1