Back to homepage

Verifiable AI

This page highlights Mengyuan Li's research at the intersection of verifiable AI, trustworthy machine learning systems, and privacy-preserving oversight. The focus is on verifying LLM inference, understanding failure modes in proof-based AI systems, and building mechanisms that improve the integrity and auditability of deployed models.

Related searches this page is designed to serve: verifiable AI, AI verification, zero-knowledge verification of LLM inference, trustworthy AI systems, model oversight.

Representative Papers

Hollow-LLM Attack: Computationally Trivial Weights in Zero-Knowledge Verification of LLM Inference
IEEE S&P 2026

Studies a concrete security problem in proof-based verification of LLM inference, showing why verifiable AI systems must be designed with stronger threat models.

ASPLOS 2026

Explores privacy-preserving oversight of model execution and also fits naturally into verifiable AI because it studies how to monitor and reason about model behavior in a trustworthy way.

Why This Matters

As AI systems become part of high-stakes workflows, users increasingly need evidence about what model ran, how it ran, and whether results can be trusted without revealing sensitive data. Verifiable AI sits at that boundary between systems, security, and machine learning.

Related Pages

See AI Agent Security for work on runtime monitoring and trusted infrastructure for LLM systems and agents. WAVE is a cross-cutting paper that also connects naturally to that direction. See TEE and Confidential Computing for the secure execution mechanisms that often support deployable verification pipelines.