Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A Versatile Graph Learning Approach through LLM-based Agent

Lanning Wei, Huan Zhao, Xiaohan Zheng, Zhiqiang He, Quanming Yao

TL;DR

Graph learning must adapt to diverse graphs and tasks. This paper introduces GL-Agent, a framework of LLM-based agents organized as a manager, data, configuration, searching, tuning, and response agents that collaboratively tailor AutoML-driven graph learning pipelines to task- and data-specific settings. Across 11 datasets covering node, graph, and link tasks, GL-Agent achieves comparable or superior performance to AutoML baselines while delivering correct agent outputs and demonstrating robustness to complex instructions. The approach offers low resource cost and potential to leverage open-source LLMs, making versatile graph learning accessible to non-experts and broad real-world applications.

Abstract

Designing versatile graph learning approaches is important, considering the diverse graphs and tasks existing in real-world applications. Existing methods have attempted to achieve this target through automated machine learning techniques, pre-training and fine-tuning strategies, and large language models. However, these methods are not versatile enough for graph learning, as they work on either limited types of graphs or a single task. In this paper, we propose to explore versatile graph learning approaches with LLM-based agents, and the key insight is customizing the graph learning procedures for diverse graphs and tasks. To achieve this, we develop several LLM-based agents, equipped with diverse profiles, tools, functions and human experience. They collaborate to configure each procedure with task and data-specific settings step by step towards versatile solutions, and the proposed method is dubbed GL-Agent. By evaluating on diverse tasks and graphs, the correct results of the agent and its comparable performance showcase the versatility of the proposed method, especially in complex scenarios.The low resource cost and the potential to use open-source LLMs highlight the efficiency of GL-Agent.

A Versatile Graph Learning Approach through LLM-based Agent

TL;DR

Graph learning must adapt to diverse graphs and tasks. This paper introduces GL-Agent, a framework of LLM-based agents organized as a manager, data, configuration, searching, tuning, and response agents that collaboratively tailor AutoML-driven graph learning pipelines to task- and data-specific settings. Across 11 datasets covering node, graph, and link tasks, GL-Agent achieves comparable or superior performance to AutoML baselines while delivering correct agent outputs and demonstrating robustness to complex instructions. The approach offers low resource cost and potential to leverage open-source LLMs, making versatile graph learning accessible to non-experts and broad real-world applications.

Abstract

Designing versatile graph learning approaches is important, considering the diverse graphs and tasks existing in real-world applications. Existing methods have attempted to achieve this target through automated machine learning techniques, pre-training and fine-tuning strategies, and large language models. However, these methods are not versatile enough for graph learning, as they work on either limited types of graphs or a single task. In this paper, we propose to explore versatile graph learning approaches with LLM-based agents, and the key insight is customizing the graph learning procedures for diverse graphs and tasks. To achieve this, we develop several LLM-based agents, equipped with diverse profiles, tools, functions and human experience. They collaborate to configure each procedure with task and data-specific settings step by step towards versatile solutions, and the proposed method is dubbed GL-Agent. By evaluating on diverse tasks and graphs, the correct results of the agent and its comparable performance showcase the versatility of the proposed method, especially in complex scenarios.The low resource cost and the potential to use open-source LLMs highlight the efficiency of GL-Agent.
Paper Structure (42 sections, 7 figures, 26 tables)

This paper contains 42 sections, 7 figures, 26 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Versatile Graph Learning Methods
  4. LLM-empowered Machine Learning
  5. Method
  6. Preliminary on Technical Support
  7. Instruction and Manager Agent
  8. Graph Learning Agents
  9. Data Agent
  10. Configuration Agent
  11. Searching Agent
  12. Tuning Agent
  13. Response Agent
  14. Discussion
  15. Experiments
  16. ...and 27 more sections

Figures (7)

  • Figure 1: The framework designed to customize the graph learning procedures with LLM-based agents.
  • Figure 2: The introduction of procedures and agents used when learning on graphs.
  • Figure 3: The designed architecture backbone from wei2022designing.
  • Figure 4: The framework used in LRGNN wei2023search, in which the yellow color nodes represent the pre- and port-processing operations and the green nodes represent the module of selection, fusion and aggregation.
  • Figure 5: The framework used in Prof-CF wang2022profiling.
  • ...and 2 more figures