Beyond Pipelines A Survey Of The Paradigm Shift Toward Modelnative Agentic Ai Jitao Sang Jinlin Xiao Jiarun Han Jilin Chen Xiaoyi Chen Shuyu Wei Yongjie Sun Yuhang Wang by Jitao Sang & Jinlin Xiao & Jiarun Han & Jilin Chen & Xiaoyi Chen & Shuyu Wei & Yongjie Sun & Yuhang Wang instant download after payment.

1 IntroductionIn recent years, the field of Artificial Intelligence (AI) has been dominated by the rapid progress ofgenerative AI, which excels at producing human-like text, images, and other modalities [13, 164, 210].Yet, the outputs of generative AI remain largely reactive: it produces content as prompted but doesnot pursue goals, sustain long-horizon reasoning, or interact with environments. To move beyondpassive generation toward autonomous action, researchers have increasingly focused on agentic AI,which emphasizes self-directed behavior, complex reasoning abilities, and environment interaction.The rise of agentic AI is widely regarded as the next stage in the evolution of AI systems.Across both academia and industry, three core capabilities are consistently highlighted as centralto agentic AI:• Planning: to decompose high-level goals into coherent, multi-step strategies.• Tool use: to invoke and coordinate external resources such as APIs, databases, or othermodels.• Memory: to retain, retrieve, and manage information over extended horizons.The development of agentic AI is deeply intertwined with the evolution of how its core capabilitiesare implemented, a process that has been undergoing a profound paradigm shift.