GraphAgent: Agentic Graph Language Assistant

TL;DR

GraphAgent tackles the challenge of analyzing graph-structured data by unifying structured and unstructured data through Semantic Knowledge Graphs using a trio of agents: a Graph Generator Agent, a Task Planning Agent, and a Graph Action Agent. The framework grounds user queries into graph-augmented tasks, automatically constructs SKGs from text, and executes tasks with graph-aware language models, enabling both predictive and generative capabilities. Empirical results across graph prediction and graph-enhanced text generation show superior performance over baselines, including smaller LLMs in zero-shot settings, with ablations confirming the value of SKG construction, alignment, and curriculum training. The work contributes an open-source, end-to-end platform for automating graph-aware reasoning with natural language, broadening accessibility to graph analytics.

Abstract

Real-world data is represented in both structured (e.g., graph connections) and unstructured (e.g., textual, visual information) formats, encompassing complex relationships that include explicit links (such as social connections and user behaviors) and implicit interdependencies among semantic entities, often illustrated through knowledge graphs. In this work, we propose GraphAgent, an automated agent pipeline that addresses both explicit graph dependencies and implicit graph-enhanced semantic inter-dependencies, aligning with practical data scenarios for predictive tasks (e.g., node classification) and generative tasks (e.g., text generation). GraphAgent comprises three key components: (i) a Graph Generator Agent that builds knowledge graphs to reflect complex semantic dependencies; (ii) a Task Planning Agent that interprets diverse user queries and formulates corresponding tasks through agentic self-planning; and (iii) a Task Execution Agent that efficiently executes planned tasks while automating tool matching and invocation in response to user queries. These agents collaborate seamlessly, integrating language models with graph language models to uncover intricate relational information and data semantic dependencies. Through extensive experiments on various graph-related predictive and text generative tasks on diverse datasets, we demonstrate the effectiveness of our GraphAgent across various settings. We have made our proposed GraphAgent open-source at: https://github.com/HKUDS/GraphAgent.

Figures (8)

• Figure 1: GraphAgent processes both structured and unstructured data, adapting seamlessly to various downstream tasks across diverse scenarios.
• Figure 2: The overall framework of the proposed GraphAgent.
• Figure 3: Performance comparison with state-of-the-art LLMs on complex graph prediction tasks involving implicit semantic relationships. Results marked with * indicate statistical significance (p<0.01) compared to the second-best performer.
• Figure 4: Performances on ACL-EMNLP related works content generation. Light grey denotes that the score is computed with the same-family model.
• Figure 5: GovReport summarization performance. Evaluation scores are presented with same-family model comparisons highlighted in light grey.