Reproduce “APPA: Recoverable Information-Flow Control for Real-World LLM Agents” 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
Arseny Kravchenko, Vadim Liventsev, Innokentii Konstantinov, Ildar Iskhakov, Matvey Kukuy
APPA (Agentic Permissions Policy Algebra) lets agents recover from information-flow restrictions: a monitor checks calls before dispatch and returns before they enter context, and offers remedies such as child branches, sanitization, or call-scoped rulings. Across 1,320 episodes, it recorded zero attacks. In a Luna ablation, utility was 88.0%, versus 56.5% without branching and 35.0% without remedies; these results are evidence.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.