New TTGBench benchmark reveals LLM vs TGNN divide in temporal graph learning

Researchers have introduced TTGBench, a new benchmark designed to evaluate both structural and semantic evolution in temporal graph learning. This benchmark addresses limitations in existing methods that primarily focus on structural changes, offering a more comprehensive assessment of models' capabilities. TTGBench includes six real-world datasets with text-rich attributes and "Dual Volatility," supporting multi-class and multi-label temporal node classification to better capture semantic drift. Initial evaluations indicate a distinct performance gap between Temporal Graph Neural Networks (TGNNs), which excel at structural prediction, and Large Language Model (LLM)-based approaches, which perform better on semantic tracking. AI

IMPACT Highlights a capability divide between TGNNs and LLMs in temporal graph analysis, guiding future model development.

RANK_REASON The cluster describes a new benchmark and research paper for evaluating temporal graph learning models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• large language model
• ScienceCast
• Temporal Graph Neural Networks
• TTGBench

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →