Trust, but Verify

Computer 2026 Open access

Read full text
Authors: Michael J. Yuan, Carlos Lospoy, Sydney Lai, James Snewin, J. B. Long
DOI: 10.1109/mc.2025.3600630
Cited by: 0
References: 4

Abstract

We demonstrate that in a cluster of mostly honest nodes, we can detect nodes that run unauthorized or incorrect large language models (LLMs) through the social consensus of their peers. We discuss an intersubjective validation system to encourage honest behavior from LLM nodes.

Fields of Study & Subjects

Access Control and Trust Blockchain Technology Applications and Security Adversarial Robustness in Machine Learning

Publication details

Journal / source
Computer
Volume
59
Issue
2
Pages
28–38
Publication date
1 February 2026
Language
EN