Conversational Agents:
From RAG to LTM

Outline

• Opening

• Towards Trustworthy Conversational Agents

  • A General Framework for Trustworthy Conversational Agents {Yue}

  • How Do Conversational Agents Reason {Yue}

  • Reduce Hallucinations for Conversational Agents {Yue}

  • Teaching Conversational Agents to Use Externals Tools {Yue}

  • Retrieval-Augmented Generation (RAG) for Conversational Agents {Yue}

• Inclusive Multimodal Conversational Search Examples

  • Inclusive Information Access {Haiming}

  • Inclusive Symbol-based Conversational Search  {Haiming}

  • DoodleShoper  {Haiming}

• Long-Term Memory (LTM) for Conversational Agents

  • Why Do We Need LTM {Dell}

  • Approaches to LTM  {Changzhi & Jixiang}

    • Text LTM

    • Graph LTM

    • Parametric LTM

  • Evaluation of LTM {Xiangyu}

  • Looking Forward {Dell}

• Closing

Authors

• Dell Zhang, Institute of Artificial Intelligence (TeleAI), China Telecom, China

• Yue Feng, University of Birmingham, UK,

• Haiming Liu, University of Southampton, UK

• Changzhi Sun, Institute of Artificial Intelligence (TeleAI), China Telecom, China

• Jixiang Luo, Institute of Artificial Intelligence (TeleAI), China Telecom, China

• Xiangyu Chen, Institute of Artificial Intelligence (TeleAI), China Telecom, China

• Xuelong Li, Institute of Artificial Intelligence (TeleAI), China Telecom, China

Paper

• https://dl.acm.org/doi/10.1145/3767695.3769671 

@inproceedings{10.1145/3767695.3769671,

        author = {Zhang, Dell and Feng, Yue and Liu, Haiming and Sun, Changzhi and Luo, Jixiang and Chen, Xiangyu and Li, Xuelong},

        title = {Conversational Agents: From RAG to LTM},

        year = {2025},

        isbn = {9798400722189},

        doi = {10.1145/3767695.3769671},

        booktitle = {Proceedings of the 2025 Annual International ACM SIGIR Conference on Research and Development in Information Retrieval in the Asia Pacific Region (SIGIR-AP)},

        pages = {447–452},

        location = {China}

}

References

• https://github.com/TeleAI-UAGI/Awesome-Agent-Memory

• https://github.com/TeleAI-UAGI/telemem