Saved in:
Bibliographic Details
Main Authors: Sun, Nan, Li, Yongchang, Wang, Chenxu, Li, Huiying, Liu, Huaping
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2509.14889
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916956187131904
author Sun, Nan
Li, Yongchang
Wang, Chenxu
Li, Huiying
Liu, Huaping
author_facet Sun, Nan
Li, Yongchang
Wang, Chenxu
Li, Huiying
Liu, Huaping
contents In this work, we present CollabVLA, a self-reflective vision-language-action framework that transforms a standard visuomotor policy into a collaborative assistant. CollabVLA tackles key limitations of prior VLAs, including domain overfitting, non-interpretable reasoning, and the high latency of auxiliary generative models, by integrating VLM-based reflective reasoning with diffusion-based action generation under a mixture-of-experts design. Through a two-stage training recipe of action grounding and reflection tuning, it supports explicit self-reflection and proactively solicits human guidance when confronted with uncertainty or repeated failure. It cuts normalized Time by ~2x and Dream counts by ~4x vs. generative agents, achieving higher success rates, improved interpretability, and balanced low latency compared with existing methods. This work takes a pioneering step toward shifting VLAs from opaque controllers to genuinely assistive agents capable of reasoning, acting, and collaborating with humans.
format Preprint
id arxiv_https___arxiv_org_abs_2509_14889
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle CollabVLA: Self-Reflective Vision-Language-Action Model Dreaming Together with Human
Sun, Nan
Li, Yongchang
Wang, Chenxu
Li, Huiying
Liu, Huaping
Robotics
In this work, we present CollabVLA, a self-reflective vision-language-action framework that transforms a standard visuomotor policy into a collaborative assistant. CollabVLA tackles key limitations of prior VLAs, including domain overfitting, non-interpretable reasoning, and the high latency of auxiliary generative models, by integrating VLM-based reflective reasoning with diffusion-based action generation under a mixture-of-experts design. Through a two-stage training recipe of action grounding and reflection tuning, it supports explicit self-reflection and proactively solicits human guidance when confronted with uncertainty or repeated failure. It cuts normalized Time by ~2x and Dream counts by ~4x vs. generative agents, achieving higher success rates, improved interpretability, and balanced low latency compared with existing methods. This work takes a pioneering step toward shifting VLAs from opaque controllers to genuinely assistive agents capable of reasoning, acting, and collaborating with humans.
title CollabVLA: Self-Reflective Vision-Language-Action Model Dreaming Together with Human
topic Robotics
url https://arxiv.org/abs/2509.14889