Reproduce “Organizational Control Layer: Governance Infrastructure at the Execution Boundary of LLM Agent Systems” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Tianyu Shi, Yang Mo, Yiou Liu, Zhuonan Hao, Yin Wang, Wenzhuo Hu, Nan Yu, Meng Zhou and 1 more
The Organizational Control Layer (OCL) provides a model-agnostic governance framework for LLM agent systems, mediating actions at the execution boundary to ensure safety and compliance. In adversarial negotiation tasks, OCL achieved a 0% unsafe execution rate and a 96% valid success rate, significantly improving reliability over ungoverned LLM agents.
Key insights
- Pre-execution governance is essential for safely deploying LLM agents in real-world, high-stakes environments where actions have direct and measurable consequences.
- Structured recovery mechanisms, such as deterministic revision (replanning), are crucial for preserving task utility while enforcing safety, as merely blocking unsafe actions can lead to task failure.
- The OCL approach enhances system transparency and auditability by providing detailed logs of proposed decisions, constraint checks, and control outcomes.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.