: Automating Agentic Workflow Generation via Self-Adaptive Abstraction Operators
Source: arXiv Authors: Mingming Zhao, Xiaokang Wei, Yuanqi Shao, Kaiwen Zhou, Lin Yang, Siwei Rao, Junhui Zhan, Zhitang Chen Date: 2025-11-23 Primary category: cs.AI All categories: cs.AI, cs.MA
Abstract
A^2Flow removes dependence on a fixed hand-authored operator library by extracting, clustering, and abstracting operators from examples into reusable workflow building blocks. It then uses those learned operators during node-level workflow search with operator memory. The paper is especially relevant to any system that wants workflow evolution to include the vocabulary of workflow steps, not only the arrangement of already-known steps.