Unified Vision-Language-Action Model

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Wang, Yuqi, Li, Xinghang, Wang, Wenxuan, Zhang, Junbo, Li, Yingyan, Chen, Yuntao, Wang, Xinlong, Zhang, Zhaoxiang
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866909659188690944
author Wang, Yuqi
Li, Xinghang
Wang, Wenxuan
Zhang, Junbo
Li, Yingyan
Chen, Yuntao
Wang, Xinlong
Zhang, Zhaoxiang
author_facet Wang, Yuqi
Li, Xinghang
Wang, Wenxuan
Zhang, Junbo
Li, Yingyan
Chen, Yuntao
Wang, Xinlong
Zhang, Zhaoxiang
contents Vision-language-action models (VLAs) have garnered significant attention for their potential in advancing robotic manipulation. However, previous approaches predominantly rely on the general comprehension capabilities of vision-language models (VLMs) to generate action signals, often overlooking the rich temporal and causal structure embedded in visual observations. In this paper, we present UniVLA, a unified and native multimodal VLA model that autoregressively models vision, language, and action signals as discrete token sequences. This formulation enables flexible multimodal tasks learning, particularly from large-scale video data. By incorporating world modeling during post-training, UniVLA captures causal dynamics from videos, facilitating effective transfer to downstream policy learning--especially for long-horizon tasks. Our approach sets new state-of-the-art results across several widely used simulation benchmarks, including CALVIN, LIBERO, and Simplenv-Bridge, significantly surpassing previous methods. For example, UniVLA achieves 95.5% average success rate on LIBERO benchmark, surpassing pi0-FAST's 85.5%. We further demonstrate its broad applicability on real-world ALOHA manipulation and autonomous driving.
format Preprint
id arxiv_https___arxiv_org_abs_2506_19850
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Unified Vision-Language-Action Model
Wang, Yuqi
Li, Xinghang
Wang, Wenxuan
Zhang, Junbo
Li, Yingyan
Chen, Yuntao
Wang, Xinlong
Zhang, Zhaoxiang
Computer Vision and Pattern Recognition
Robotics
Vision-language-action models (VLAs) have garnered significant attention for their potential in advancing robotic manipulation. However, previous approaches predominantly rely on the general comprehension capabilities of vision-language models (VLMs) to generate action signals, often overlooking the rich temporal and causal structure embedded in visual observations. In this paper, we present UniVLA, a unified and native multimodal VLA model that autoregressively models vision, language, and action signals as discrete token sequences. This formulation enables flexible multimodal tasks learning, particularly from large-scale video data. By incorporating world modeling during post-training, UniVLA captures causal dynamics from videos, facilitating effective transfer to downstream policy learning--especially for long-horizon tasks. Our approach sets new state-of-the-art results across several widely used simulation benchmarks, including CALVIN, LIBERO, and Simplenv-Bridge, significantly surpassing previous methods. For example, UniVLA achieves 95.5% average success rate on LIBERO benchmark, surpassing pi0-FAST's 85.5%. We further demonstrate its broad applicability on real-world ALOHA manipulation and autonomous driving.
title Unified Vision-Language-Action Model
topic Computer Vision and Pattern Recognition
Robotics
url https://arxiv.org/abs/2506.19850