GNNs as Predictors of Agentic Workflow Performances

Yuanshuo Zhang, Yuchen Hou, Bohan Tang, Shuo Chen, Muhan Zhang, Xiaowen Dong, Siheng Chen · 2025

Read GNNs as Predictors of Agentic Workflow Performances by Yuanshuo Zhang, Yuchen Hou, Bohan Tang, Shuo Chen, Muhan Zhang, Xiaowen Dong, Siheng Chen (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2503-11301.txt.

Description

Yuanshuo Zhang and co-authors argue in a 2025 position paper that graph networks can predict how well AI agent workflows will do. Such workflows chain large language model calls to solve hard tasks, but testing each design by running it is slow and costly. The authors recast a workflow as a graph, with steps as nodes and links as edges.

Graph neural networks, which learn from graph shape, can then score a workflow without running the costly models. To ground the claim, the team built a benchmark platform for this task and ran wide tests. The networks proved simple yet strong predictors. The authors say this opens a path to tuning agent workflows automatically, and they released their code, models, and data.

Catalog record

CREATORYuanshuo Zhang, Yuchen Hou, Bohan Tang, Shuo Chen, Muhan Zhang, Xiaowen Dong, Siheng Chen
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE50 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2503-11301

Browse the full library · About the Tower of Babel