FlowSteer: Towards Agents Designing Agentic Workflows via Reinforced Progressive Canvas Editing

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Zhang, Mingda, Liu, Wenjin, Shen, Tiesunlong, Lin, Qika, Mao, Rui, Cambria, Erik, Tang, Xiaoying, Luo, Haoran
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910217903538176
author Zhang, Mingda
Liu, Wenjin
Shen, Tiesunlong
Lin, Qika
Mao, Rui
Cambria, Erik
Tang, Xiaoying
Luo, Haoran
author_facet Zhang, Mingda
Liu, Wenjin
Shen, Tiesunlong
Lin, Qika
Mao, Rui
Cambria, Erik
Tang, Xiaoying
Luo, Haoran
contents In recent years, agentic workflows have been widely applied to solve complex human tasks. However, existing workflow construction still faces key challenges, including human-dependent workflow construction, the lack of graph-level execution feedback, and the inability to repair errors in-loop during long-horizon construction. To address these challenges, we propose FlowSteer, a new paradigm of Agent Designing Agentic Workflows - a single agent itself end-to-end designs the workflow that a downstream executor runs. To support this paradigm, we introduce the Workflow Canvas, a novel executable graph-state environment that returns syntax-checked execution feedback for every atomic edit. Built on the canvas, we further propose Reinforced Progressive Canvas Editing, in which a lightweight policy agent issues one atomic edit per turn conditioned on real canvas feedback, and is trained end-to-end via reinforcement learning. Moreover, FlowSteer provides a plug-and-play framework that supports diverse operator libraries and interchangeable LLM backends. Experimental results on twelve datasets show that FlowSteer significantly outperforms baselines across various tasks. Our code is available at https://anonymous.4open.science/r/FlowSteer-9B2E.
format Preprint
id arxiv_https___arxiv_org_abs_2602_01664
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle FlowSteer: Towards Agents Designing Agentic Workflows via Reinforced Progressive Canvas Editing
Zhang, Mingda
Liu, Wenjin
Shen, Tiesunlong
Lin, Qika
Mao, Rui
Cambria, Erik
Tang, Xiaoying
Luo, Haoran
Artificial Intelligence
Machine Learning
I.2.6; I.2.11
In recent years, agentic workflows have been widely applied to solve complex human tasks. However, existing workflow construction still faces key challenges, including human-dependent workflow construction, the lack of graph-level execution feedback, and the inability to repair errors in-loop during long-horizon construction. To address these challenges, we propose FlowSteer, a new paradigm of Agent Designing Agentic Workflows - a single agent itself end-to-end designs the workflow that a downstream executor runs. To support this paradigm, we introduce the Workflow Canvas, a novel executable graph-state environment that returns syntax-checked execution feedback for every atomic edit. Built on the canvas, we further propose Reinforced Progressive Canvas Editing, in which a lightweight policy agent issues one atomic edit per turn conditioned on real canvas feedback, and is trained end-to-end via reinforcement learning. Moreover, FlowSteer provides a plug-and-play framework that supports diverse operator libraries and interchangeable LLM backends. Experimental results on twelve datasets show that FlowSteer significantly outperforms baselines across various tasks. Our code is available at https://anonymous.4open.science/r/FlowSteer-9B2E.
title FlowSteer: Towards Agents Designing Agentic Workflows via Reinforced Progressive Canvas Editing
topic Artificial Intelligence
Machine Learning
I.2.6; I.2.11
url https://arxiv.org/abs/2602.01664