UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent

Jianke Zhang, Yanjiang Guo, Yucheng Hu, Xiaoyu Chen, Xiang Zhu, Jianyu Chen
Proceedings of the 42nd International Conference on Machine Learning, PMLR 267:74911-74922, 2025.

Abstract

Recent advancements in Vision-Language-Action (VLA) models have leveraged pre-trained Vision-Language Models (VLMs) to improve the generalization capabilities. VLMs, typically pre-trained on vision-language understanding tasks, provide rich semantic knowledge and reasoning abilities. However, prior research has shown that VLMs often focus on high-level semantic content and neglect low-level features, limiting their ability to capture detailed spatial information and understand physical dynamics. These aspects, which are crucial for embodied control tasks, remain underexplored in existing pre-training paradigms. In this paper, we investigate the training paradigm for VLAs, and introduce UP-VLA, a Unified VLA model training with both multi-modal Understanding and future Prediction objectives, enhancing both high-level semantic comprehension and low-level spatial understanding. Experimental results show that UP-VLA achieves a 33% improvement on the Calvin ABC-D benchmark compared to the previous state-of-the-art method. Additionally, UP-VLA demonstrates improved success rates in real-world manipulation tasks, particularly those requiring precise spatial information.

Cite this Paper

BibTeX
@InProceedings{pmlr-v267-zhang25w, title = {{UP}-{VLA}: A Unified Understanding and Prediction Model for Embodied Agent}, author = {Zhang, Jianke and Guo, Yanjiang and Hu, Yucheng and Chen, Xiaoyu and Zhu, Xiang and Chen, Jianyu}, booktitle = {Proceedings of the 42nd International Conference on Machine Learning}, pages = {74911--74922}, year = {2025}, editor = {Singh, Aarti and Fazel, Maryam and Hsu, Daniel and Lacoste-Julien, Simon and Berkenkamp, Felix and Maharaj, Tegan and Wagstaff, Kiri and Zhu, Jerry}, volume = {267}, series = {Proceedings of Machine Learning Research}, month = {13--19 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v267/main/assets/zhang25w/zhang25w.pdf}, url = {https://proceedings.mlr.press/v267/zhang25w.html}, abstract = {Recent advancements in Vision-Language-Action (VLA) models have leveraged pre-trained Vision-Language Models (VLMs) to improve the generalization capabilities. VLMs, typically pre-trained on vision-language understanding tasks, provide rich semantic knowledge and reasoning abilities. However, prior research has shown that VLMs often focus on high-level semantic content and neglect low-level features, limiting their ability to capture detailed spatial information and understand physical dynamics. These aspects, which are crucial for embodied control tasks, remain underexplored in existing pre-training paradigms. In this paper, we investigate the training paradigm for VLAs, and introduce UP-VLA, a Unified VLA model training with both multi-modal Understanding and future Prediction objectives, enhancing both high-level semantic comprehension and low-level spatial understanding. Experimental results show that UP-VLA achieves a 33% improvement on the Calvin ABC-D benchmark compared to the previous state-of-the-art method. Additionally, UP-VLA demonstrates improved success rates in real-world manipulation tasks, particularly those requiring precise spatial information.} }
Endnote
%0 Conference Paper %T UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent %A Jianke Zhang %A Yanjiang Guo %A Yucheng Hu %A Xiaoyu Chen %A Xiang Zhu %A Jianyu Chen %B Proceedings of the 42nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2025 %E Aarti Singh %E Maryam Fazel %E Daniel Hsu %E Simon Lacoste-Julien %E Felix Berkenkamp %E Tegan Maharaj %E Kiri Wagstaff %E Jerry Zhu %F pmlr-v267-zhang25w %I PMLR %P 74911--74922 %U https://proceedings.mlr.press/v267/zhang25w.html %V 267 %X Recent advancements in Vision-Language-Action (VLA) models have leveraged pre-trained Vision-Language Models (VLMs) to improve the generalization capabilities. VLMs, typically pre-trained on vision-language understanding tasks, provide rich semantic knowledge and reasoning abilities. However, prior research has shown that VLMs often focus on high-level semantic content and neglect low-level features, limiting their ability to capture detailed spatial information and understand physical dynamics. These aspects, which are crucial for embodied control tasks, remain underexplored in existing pre-training paradigms. In this paper, we investigate the training paradigm for VLAs, and introduce UP-VLA, a Unified VLA model training with both multi-modal Understanding and future Prediction objectives, enhancing both high-level semantic comprehension and low-level spatial understanding. Experimental results show that UP-VLA achieves a 33% improvement on the Calvin ABC-D benchmark compared to the previous state-of-the-art method. Additionally, UP-VLA demonstrates improved success rates in real-world manipulation tasks, particularly those requiring precise spatial information.
APA
Zhang, J., Guo, Y., Hu, Y., Chen, X., Zhu, X. & Chen, J.. (2025). UP-VLA: A Unified Understanding and Prediction Model for Embodied Agent. Proceedings of the 42nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 267:74911-74922 Available from https://proceedings.mlr.press/v267/zhang25w.html.

Related Material