eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
Why it matters
Most AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → benchmarks score the end state and ignore the wreckage of failed attempts. For anyone deploying computer-use agents against real records, partial-failure damage is the risk that actually matters.
articleAgentR A Stateful and Recovery-Aware Software Architecture for LLM-based Auditable WorkflowsRiya Samanta, Bidyut Saha, Soumya Kanti Ghosh, Rajkumar Buyya
articleLegacy System Modernization with Coding Agents: A Case StudyIago da Silva Rodrigues Alves, Cristiano Politowski, Jo\~ao Eduardo Montandon
articleDo GUI Agents Know When Not to Act? Enabling Conflict-Aware Termination for Multimodal GUI AgentsZhaoyuan Huang, Tianjie Ju, Pengzhou Cheng, Zheng Wu, Yansi Li, Chuanbiao Song, Jun Lan, Huijia Zhu, Weiqiang Wang, Zhuosheng Zhang