New Framework TRUST Aims to Decentralize AI Verification for High-Stakes Domains

A new framework named TRUST (v.0.1) has been proposed to address verification challenges in Large Reasoning Models (LRMs) and Multi-Agent Systems (MAS). The framework aims to overcome four key limitations of centralized approaches: robustness, scalability, bias, and single points of failure. By decentralizing verification, TRUST seeks to provide more reliable and trustworthy AI services, especially in high-stakes applications. The preprint is available on arXiv (2604.27132).