Distributed statistical techniques for multi-agent systems

Develop distributed versions of statistical verification and control techniques (e.g., conformal prediction and statistical abstractions) that can be applied across agents in large-scale multi-agent systems while preserving safety guarantees and scalability.

Background

The paper argues statistical techniques improve scalability for single-agent settings. Extending these to distributed multi-agent contexts raises challenges in coordination, independence assumptions, and guarantee preservation.

The authors explicitly state that it is unclear how statistical techniques can be distributed, framing a concrete methodological gap.