paperSeptember 2026Unreviewed

PolicyMem: Geometric Policy Memory for LLM Governance

Yuanchen Bei, Zhengzhang Chen, Yanjun Zhao, Haoyu Wang, Hanghang Tong, Haifeng Chen

Abstract

As large language models (LLMs) are increasingly deployed in real-world high-stakes applications, effective governance has become essential. Existing safeguards largely follow two paradigms: learning-based guards provide strong semantic discrimination but couple policy behavior to trained models and taxonomies, while programmable frameworks offer flexible control but require substantial manual prompt and workflow engineering. Neither externalizes policies as reusable operational states, making i

Cite

@misc{bei2026policymem,
  title = {{PolicyMem: Geometric Policy Memory for LLM Governance}},
  author = {Yuanchen Bei and Zhengzhang Chen and Yanjun Zhao and Haoyu Wang and Hanghang Tong and Haifeng Chen},
  year = {2026},
  month = sep,
  eprint = {2609.13734},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.13734}
}