Portrait of Jamyoung Xu
Academic Homepage

Jamyoung Xu 徐文江

我目前是中国科学院自动化研究所（CASIA）博士研究生（2024-），本科毕业于南京大学自动化系（2020-2024）。我的研究兴趣包括具身智能、智能体、复杂推理与机器人操作规划。更具体地说，我关注机器人在动态真实环境中的落地问题：如何让机器人有效地思考并行动，如何在交互过程中保障人和环境的安全，以及如何赋予机器人持续学习和适应新任务的能力。欢迎有相同兴趣的研究者联系我交流合作；我也在积极寻找工业界实习机会。

I am a Ph.D. student at the Institute of Automation, Chinese Academy of Sciences (CASIA, 2024-), and received my B.Eng. from the Department of Automation at Nanjing University (2020-2024). My research interests span embodied AI, agents, complex reasoning, and robotic manipulation planning. I am particularly interested in deploying robots in dynamic real-world environments: enabling robots to reason and act effectively, ensuring the safety of humans and surroundings during interaction, and developing continual learning abilities for adapting to new tasks. I welcome collaborations with researchers who share similar interests, and I am actively seeking internship opportunities in industry.

Embodied AI Agent Large Language Models Robotics Planning

Selected Work

Publications

Google Scholar: 151 citations
2026/8/19

Embodied Tree of Thoughts teaser
Featured RA-L 2026 First Author 5 citations

Embodied tree of thoughts: Deliberate manipulation planning with embodied world model

W Xu, M Zhang, C Wang, R Fang, L Li, J Xu, J Gu, Z Zeng, R Chen

IEEE Robotics and Automation Letters, 2026

This paper brings tree-structured deliberation into embodied manipulation planning. It couples reasoning over candidate action branches with an embodied world model, enabling the agent to evaluate physical consequences before committing to long-horizon manipulation steps.

Project Figure Scholar
Logic-of-Thought teaser
Featured NAACL 2025 CCF B First Author Excluding Corresponding Advisor 61 citations

Logic-of-thought: Injecting logic into contexts for full reasoning in large language models

T Liu*, W Xu*, W Huang, Y Zeng, J Wang, X Wang, H Yang, J Li

* Equal contribution

Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics

This work studies how logic-aware context injection can strengthen full-chain reasoning in large language models. By making implicit logical relations more explicit before generation, the method aims to improve consistency, completeness, and reliability on complex reasoning tasks.

Figure Scholar
GroupDebate Multi-agent debate with group discussion for more efficient coordination.
AAMAS 2026 Preprint 82 citations

Groupdebate: Enhancing the efficiency of multi-agent debate using group discussion

T Liu, X Wang, W Huang, W Xu, Y Zeng, L Jiang, H Yang, J Li

Autonomous Agents and Multiagent Systems, 2026

GroupDebate improves multi-agent debate by organizing agents into discussion groups, reducing redundant exchanges while preserving the benefits of diverse viewpoints and collaborative refinement.

Scholar
Critic in the Loop A tri-system VLA framework for robust long-horizon manipulation.
arXiv 2026 Preprint 3 citations

Critic in the loop: A tri-system vla framework for robust long-horizon manipulation

P Yi, Y Ma, W Xu, Y Hao, S Gan, W Li, S Zhong

arXiv preprint arXiv:2603.05185, 2026

This work investigates robust long-horizon manipulation through a tri-system vision-language-action framework, introducing a critic component that monitors and improves action execution across extended tasks.

Scholar