Hi There, I'm

Rebecca Westhäußer

PhD Student

I am a PhD student at Mercedes-Benz, in cooperation with Ulm University. My research explores personal AI agents, with a focus on defining and measuring personalization and improving agents’ ability to adapt to individual users over time.

See my publications

Portrait of Rebecca Westhäußer

Publications

  1. Abstract

    Personalizing Long-Term Interactions with LLM-Based Agents

    Rebecca Westhäußer

    Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence (IJCAI 2026)

    DOI 10.24963/ijcai.2026/953

    LLM-based agents need effective ways to remember and reuse user-specific information across sessions. This doctoral consortium abstract outlines research on combining memory augmentation and agentic decision-making to enable more personalized long-term interactions, including the evaluation of CAIM and user-specific memory modules.

    Read paper
    Cite this paper

    Citation

    Westhäußer, R. (2026). Personalizing Long-Term Interactions with LLM-Based Agents. Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence (IJCAI 2026), 8349–8350. https://doi.org/10.24963/ijcai.2026/953

    BibTeX

    @inproceedings{westhaeusser2026personalizing,
  author = {Rebecca Westhäußer},
  title = {Personalizing Long-Term Interactions with {LLM}-Based Agents},
  booktitle = {Proceedings of the Thirty-Fifth International Joint Conference on Artificial Intelligence},
  year = {2026},
  pages = {8349--8350},
  doi = {10.24963/ijcai.2026/953},
  url = {https://www.ijcai.org/proceedings/2026/953}
}

  2. Conference paper

    CAIM: Development and Evaluation of a Cognitive AI Memory Framework for Long-Term Interaction with Intelligent Agents

    Rebecca Westhäußer , Wolfgang Minker , Sebastian Zepf

    Proceedings of the 31st International Conference on Intelligent User Interfaces (IUI 2026)

    DOI 10.1145/3742413.3789222

    Long-term conversations require agents to retain relevant context while adapting to individual users. This paper develops and evaluates CAIM, a cognitive AI memory framework with dedicated memory control, retrieval, and maintenance modules. The evaluation shows how this approach can improve context-aware, personalized interaction over time.

    Read paper GitHub project
    Cite this paper

    Citation

    Westhäußer, R., Minker, W., & Zepf, S. (2026). CAIM: Development and Evaluation of a Cognitive AI Memory Framework for Long-Term Interaction with Intelligent Agents. Proceedings of the 31st International Conference on Intelligent User Interfaces (IUI 2026). https://doi.org/10.1145/3742413.3789222

    BibTeX

    @inproceedings{westhaeusser2026caim,
  author = {Rebecca Westhäußer and Wolfgang Minker and Sebastian Zepf},
  title = {CAIM: Development and Evaluation of a Cognitive AI Memory Framework for Long-Term Interaction with Intelligent Agents},
  booktitle = {Proceedings of the 31st International Conference on Intelligent User Interfaces},
  year = {2026},
  doi = {10.1145/3742413.3789222},
  url = {https://doi.org/10.1145/3742413.3789222}
}

  3. Preprint

    Enabling Personalized Long-term Interactions in LLM-based Agents through Persistent Memory and User Profiles

    Rebecca Westhäußer , Wolfgang Minker , Sebastian Zepf

    arXiv preprint arXiv:2510.07925

    DOI 10.48550/arXiv.2510.07925

    Personalized AI agents need to combine conversational context with user-specific information over time. This paper presents a framework that integrates persistent memory, dynamic coordination, self-validation, and evolving user profiles. Its evaluation across three public datasets and a five-day pilot study offers early evidence for more adaptive, personalized long-term interactions.

    Read paper GitHub project
    Cite this paper

    Citation

    Westhäußer, R., Minker, W., & Zepf, S. (2025). Enabling Personalized Long-term Interactions in LLM-based Agents through Persistent Memory and User Profiles. arXiv preprint arXiv:2510.07925. https://doi.org/10.48550/arXiv.2510.07925

    BibTeX

    @article{westhaeusser2025personalized,
  author = {Rebecca Westhäußer and Wolfgang Minker and Sebastian Zepf},
  title = {Enabling Personalized Long-term Interactions in {LLM}-based Agents through Persistent Memory and User Profiles},
  journal = {arXiv preprint arXiv:2510.07925},
  year = {2025},
  doi = {10.48550/arXiv.2510.07925},
  url = {https://arxiv.org/abs/2510.07925}
}

  4. Workshop paper

    BEHAVE AI: BEst Practices and Guidelines for Human-Centric Design and EvAluation of ProactiVE AI Agents

    Matthias Kraus , Sebastian Zepf , Rebecca Westhäußer , Isabel Feustel , Nima Zargham , Ilhan Aslan , Justin Edwards , Sven Mayer , Dimosthenis Kontogiorgos , Nicolas Wagner , Elisabeth André

    Companion Proceedings of the 30th International Conference on Intelligent User Interfaces (IUI 2025)

    DOI 10.1145/3708557.3716155

    Proactive AI agents must be helpful without becoming intrusive or poorly timed. This paper brings together best practices and guidelines for the human-centric design and evaluation of proactive agents. It frames the challenges that researchers and practitioners need to address when building systems that act on a user’s behalf.

    Read paper
    Cite this paper

    Citation

    Kraus, M., Zepf, S., Westhäußer, R., Feustel, I., Zargham, N., Aslan, I., Edwards, J., Mayer, S., Kontogiorgos, D., Wagner, N., & André, E. (2025). BEHAVE AI: BEst Practices and Guidelines for Human-Centric Design and EvAluation of ProactiVE AI Agents. Companion Proceedings of the 30th International Conference on Intelligent User Interfaces (IUI 2025). https://doi.org/10.1145/3708557.3716155

    BibTeX

    @inproceedings{kraus2025behave,
  author = {Matthias Kraus and Sebastian Zepf and Rebecca Westhäußer and Isabel Feustel and Nima Zargham and Ilhan Aslan and Justin Edwards and Sven Mayer and Dimosthenis Kontogiorgos and Nicolas Wagner and Elisabeth André},
  title = {{BEHAVE AI}: {B}Est Practices and Guidelines for Human-Centric Design and {E}vAluation of Proacti{VE} {AI} Agents},
  booktitle = {Companion Proceedings of the 30th International Conference on Intelligent User Interfaces},
  year = {2025},
  doi = {10.1145/3708557.3716155},
  url = {https://doi.org/10.1145/3708557.3716155}
}

  5. Abstract

    CAIM: A Cognitive AI Memory Framework for Long-term Interaction with LLMs

    Rebecca Westhäußer , Sebastian Zepf , Wolfgang Minker

    Companion Proceedings of the 30th International Conference on Intelligent User Interfaces (IUI 2025)

    DOI 10.1145/3708557.3716342

    Large language models need a structured way to carry context across extended interactions. This paper introduces CAIM, a cognitive AI memory framework that combines a memory controller, retrieval, and post-thinking. It establishes the architecture and its potential to support more relevant long-term relationships between people and language-model agents.

    Read paper GitHub project
    Cite this paper

    Citation

    Westhäußer, R., Zepf, S., & Minker, W. (2025). CAIM: A Cognitive AI Memory Framework for Long-term Interaction with LLMs. Companion Proceedings of the 30th International Conference on Intelligent User Interfaces (IUI 2025). https://doi.org/10.1145/3708557.3716342

    BibTeX

    @inproceedings{westhaeusser2025caim,
  author = {Rebecca Westhäußer and Sebastian Zepf and Wolfgang Minker},
  title = {CAIM: A Cognitive AI Memory Framework for Long-term Interaction with {LLMs}},
  booktitle = {Companion Proceedings of the 30th International Conference on Intelligent User Interfaces},
  year = {2025},
  doi = {10.1145/3708557.3716342},
  url = {https://doi.org/10.1145/3708557.3716342}
}