OpenRTAG Benchmark for Text-Attributed Graph Robustness
July 22, 2026
OpenRTAG provides a standardized evaluation framework for text-attributed graph learning across nine degradation scenarios, including sparsity, noise, and imbalance. It utilizes a 3x3 taxonomy to test model sensitivity across nine datasets and three downstream tasks.
HOW THIS AFFECTS YOU
●
researcherYou can use this taxonomy to systematically evaluate how text and structural noise affect your graph learning models.