THERE IS A long-running joke in artificial intelligence (AI): once we understand how something works, we stop calling it intelligence and start calling it boring software automation. While a capability remains surprising and mysterious, it belongs to AI. Once it becomes understood, reliable, and repeatable, it is treat...
An LLM judge assesses agent steps from text packets. A renderer converts recorded runs into these packets, while software checkers test records and packets against requirements. We ask what evidence shows a packet is what the evaluation requires and propose four input-assurance obligations: required content survives re...
Bo-Ming Xia, Qing-Hua Lu, Run-Ze Li et al.· 0 citations
Agentic AI moves generative AI applications toward systems that coordinate reasoning, enterprise context, model inference, and external actions. Integrating these capabilities into enterprise software introduces diverse architectural concerns, from orchestration, knowledge and memory, and model access to tools and acti...
Ming-Qin Yu, F. Rabhi, Li-Ming Zhu et al.· 0 citations
This work presents a safety-by-design assurance architecture combining model-level supervision, such as Scientist AI, with system-level controls over scaffolds and harnesses, independent verification, monitoring, and evidence infrastructure, supported by governance for accountability and evidence interoperability.
A taxonomy focused on the architectures of foundation-model-based agents is introduced, addressing critical aspects such as functional capabilities and non-functional qualities and unifying and detailing these classifications to improve the design of foundation-model-based agents.