verifiability

2 items

RESEARCHarXiv CS.AI·15d ago

EVE-Agent: Evidence-Verifiable Self-Evolving Agents

EVE-Agent introduces an Evidence-Verifiable Self-Evolving Agent designed to ensure trustworthiness in self-evolution by requiring verifiable evidence for generated instances. This agent modifies the proposer-solver framework, where a proposer creates questions, answers, and evidence spans, which are then evaluated by an evidence verifier.

verifiability trustworthy AI self-evolving systems AI agents

RESEARCHarXiv CS.AI·22d ago

Verifiable Agentic Infrastructure: Proof-Derived Authorization for Sovereign AI Systems

Modern cloud systems rely on identity-centric authorization, but autonomous AI agents invalidate this assumption, creating significant operational risks, especially in sovereign AI systems. This work introduces a Distributed Trust Framework (DTF) to make authorization decisions verifiable and distributed.

security authorization verifiability trust frameworks