GDGB: A Benchmark for Generative Dynamic Text-Attributed Graph Learning

2025-07-03

AI Summary

Existing DyTAG datasets suffer from low textual quality and lack standardized definitions and evaluation protocols for dynamic text-attributed graph (DTAG) generation tasks. Method: We introduce GDGB, a benchmark comprising eight high-quality dynamic text-attributed graph datasets; formally define two generation tasks—Temporal-Driven Graph Generation (TDGG) and Identity-Driven Graph Generation (IDGG); and propose GAG-General, an LLM-powered multi-agent framework that jointly models structural, temporal, and textual features. We further design comprehensive, reproducible multi-dimensional evaluation metrics. Contribution/Results: Experiments demonstrate that joint modeling of structure and text significantly improves generated graph quality. GDGB establishes the first standardized benchmark for DTAG generation, addressing a critical gap in the field and advancing both practical applicability and methodological rigor in dynamic graph generation research.

Technology Category

• Natural Language Processing: Generation
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Data Mining & Knowledge Management: Linked Open Data, Knowledge Graphs & KB Completion

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics

Abstract

Dynamic Text-Attributed Graphs (DyTAGs), which intricately integrate structural, temporal, and textual attributes, are crucial for modeling complex real-world systems. However, most of the existing DyTAG datasets exhibit poor textual quality, which severely limits their utility for DyTAG generation tasks requiring semantically rich inputs. Additionally, prior work mainly focuses on discriminative tasks on DyTAGs, resulting in a lack of standardized task formulations and evaluation protocols tailored for DyTAG generation. To address these critical issues, we propose Generative DyTAG Benchmark (GDGB), which comprises eight meticulously curated DyTAG datasets with high-quality textual features for both nodes and edges, overcoming limitations of prior datasets. Building on GDGB, we define two novel DyTAG generation tasks: Transductive Dynamic Graph Generation (TDGG) and Inductive Dynamic Graph Generation (IDGG). TDGG transductively generates a target DyTAG based on the given source and destination node sets, while the more challenging IDGG introduces new node generation to inductively model the dynamic expansion of real-world graph data. To enable holistic evaluation, we design multifaceted metrics that assess the structural, temporal, and textual quality of the generated DyTAGs. We further propose GAG-General, an LLM-based multi-agent generative framework tailored for reproducible and robust benchmarking of DyTAG generation. Experimental results demonstrate that GDGB enables rigorous evaluation of TDGG and IDGG, with key insights revealing the critical interplay of structural and textual features in DyTAG generation. These findings establish GDGB as a foundational resource for advancing generative DyTAG research and unlocking further practical applications in DyTAG generation. GDGB datasets, source codes, and leaderboards are available at href{https://gdgb-algo.github.io/}{here}.

Problem

• Poor textual quality in existing DyTAG datasets limits generation tasks
• Lack standardized task formulations evaluation protocols DyTAG generation
• Need multifaceted metrics assess structural temporal textual quality

Innovation

• Proposes GDGB benchmark with high-quality DyTAG datasets
• Introduces TDGG and IDGG for dynamic graph generation
• Develops LLM-based GAG-General framework for evaluation