Dr. Yunpu Ma

Lecturer at CIS LMU

Group Leader of TRESP Lab

Junior Member of MCML

Industry PhDs Advisor (Siemens, Huawei, Bosch, etc)

cognitive.yunpu[at]gmail.com

About Me

I develop intelligent systems that integrate memory, reasoning, and multimodal understanding to act autonomously in open-ended environments. My goal is to develop foundation models and agentic systems that continually learn, collaborate, and adapt to complex, real-world tasks.

I am currently a lecturer at CIS LMU Munich, where I teach the course From Large Language Models to AI Agents. I am also affiliated with the Munich Center for Machine Learning and serve as a group leader at the TRESP Lab, supervising PhD students and leading research on LLMs, multimodal models, and agentic AI. I earned my Ph.D. in Computer Science from LMU under the supervision of Prof. Volker Tresp, with a focus on relational learning, cognitive modeling, and quantum AI. Prior to that, I studied theoretical physics, conducting research at the Max Planck Institute for Physics. I also collaborate closely with academic and industrial partners, including Prof. Hinrich Schütze, Prof. Volker Tresp, Prof. Sören Pirk, Prof. Evgeny Kharlamov, Prof. Kristian Kersting, and others. I also supervise several industry-funded PhD students from Siemens, Huawei, and Bosch, etc.

My current research explores the design of AI agents equipped with persistent memory, advanced reasoning capabilities, and the ability to operate autonomously in open-ended settings. A central focus is on developing LLM-based (multi)agentic systems that perceive and process multimodal inputs—such as language, vision, and structured data—while continuously adapting through reflection and self-improvement. These agents aim to support robust planning, problem-solving, and decision-making across both scientific and industrial domains.

Research Interests

Memory-augmented agents for long-term adaptation
Multi-agent systems with communication and coordination
Self-improving agents in open-ended environments
Scalable large reasoning models

Contact

If you’re interested in research collaboration, thesis supervision, or exploring shared interests in intelligent agents and foundation models, feel free to get in touch via the contact details provided on this page.

news

May 16, 2025	Invited talk of Multiagentic System that Memorizes, Communicates, and Acts at Huawei Munich Wireless Summit 2025.
Apr 23, 2025	I start teaching the course From Large Language Models to AI Agents at CIS LMU Munich in the Summer Semester 2025.
Mar 25, 2025	I am honored to serve as Guest Editor for the upcoming Special Issue on Multi-Modal AI Systems and Multi-Agent Systems in the journal Electronics (Impact Factor: 2.6).

selected publications
* (co)-first author, † (co)-corresponding author

CoT-Kinetics: A Theoretical Modeling Assessing LRM Reasoning Process

Jinhe Bi, Danqi Yan, Yifan Wang, Wenke Huang, Haokun Chen, Guancheng Wan, Mang Ye, Xun Xiao, Hinrich Schuetze, Volker Tresp, and Yunpu Ma^†

2025

Bib PDF

@misc{bi2025cotkineticstheoreticalmodelingassessing,
  title = {CoT-Kinetics: A Theoretical Modeling Assessing LRM Reasoning Process},
  author = {Bi, Jinhe and Yan, Danqi and Wang, Yifan and Huang, Wenke and Chen, Haokun and Wan, Guancheng and Ye, Mang and Xiao, Xun and Schuetze, Hinrich and Tresp, Volker and Ma, Yunpu},
  year = {2025},
  eprint = {2505.13408},
  archiveprefix = {arXiv},
  primaryclass = {cs.AI},
  url = {https://arxiv.org/abs/2505.13408},
}

Webpilot: A versatile and autonomous multi-agent system for web task execution with strategic exploration

Yao Zhang, Zijian Ma, Yunpu Ma^†, Zhen Han, Yu Wu, and Volker Tresp

In Proceedings of the AAAI Conference on Artificial Intelligence, 2025

Bib PDF

@inproceedings{zhang2025webpilot,
  title = {Webpilot: A versatile and autonomous multi-agent system for web task execution with strategic exploration},
  author = {Zhang, Yao and Ma, Zijian and Ma, Yunpu and Han, Zhen and Wu, Yu and Tresp, Volker},
  booktitle = {Proceedings of the AAAI Conference on Artificial Intelligence},
  volume = {39},
  number = {22},
  pages = {23378--23386},
  year = {2025},
}

PRISM: Self-Pruning Intrinsic Selection Method for Training-Free Multimodal Data Selection

Jinhe Bi, Yifan Wang, Danqi Yan, Xun Xiao, Artur Hecker, Volker Tresp, and Yunpu Ma^†

2025

Bib PDF

@misc{bi2025prismselfpruningintrinsicselection,
  title = {PRISM: Self-Pruning Intrinsic Selection Method for Training-Free Multimodal Data Selection},
  author = {Bi, Jinhe and Wang, Yifan and Yan, Danqi and Xiao, Xun and Hecker, Artur and Tresp, Volker and Ma, Yunpu},
  year = {2025},
  eprint = {2502.12119},
  archiveprefix = {arXiv},
  primaryclass = {cs.CV},
  url = {https://arxiv.org/abs/2502.12119},
}

LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering

Jinhe Bi, Yujun Wang, Haokun Chen, Xun Xiao, Artur Hecker, Volker Tresp, and Yunpu Ma^†

2025

Bib PDF Code

@misc{bi2025llavasteeringvisualinstruction,
  title = {LLaVA Steering: Visual Instruction Tuning with 500x Fewer Parameters through Modality Linear Representation-Steering},
  author = {Bi, Jinhe and Wang, Yujun and Chen, Haokun and Xiao, Xun and Hecker, Artur and Tresp, Volker and Ma, Yunpu},
  year = {2025},
  booktitle = {Annual Meeting of the Association for Computational Linguistics},
  primaryclass = {cs.CV},
  url = {https://arxiv.org/abs/2412.12359},
}

PERFT: Parameter-Efficient Routed Fine-Tuning for Mixture-of-Expert Model

Yilun Liu^*, Yunpu Ma^*†, Shuo Chen, Zifeng Ding, Bailan He, Zhen Han, and Volker Tresp

2024

Bib PDF

@misc{liu2024perftparameterefficientroutedfinetuning,
  title = {PERFT: Parameter-Efficient Routed Fine-Tuning for Mixture-of-Expert Model},
  author = {Liu, Yilun and Ma, Yunpu and Chen, Shuo and Ding, Zifeng and He, Bailan and Han, Zhen and Tresp, Volker},
  year = {2024},
  eprint = {2411.08212},
  archiveprefix = {arXiv},
  primaryclass = {cs.LG},
  url = {https://arxiv.org/abs/2411.08212},
}

About Me

Research Interests

Contact

news

selected publications * (co)-first author, † (co)-corresponding author

selected publications
* (co)-first author, † (co)-corresponding author