HeTGB: A Comprehensive Benchmark for Heterophilic Text-Attributed Graphs
Abstract
AbstractGraph neural networks (GNNs) have demon- strated success in modeling relational data pri- marily under the assumption of homophily. However, many real-world graphs exhibit het- erophily, where linked nodes belong to dif- ferent categories or possess diverse attributes, such as webpages, Wikipedia articles, social networks, and e-commerce platforms. Addi- tionally, nodes in many domains are associated with textual descriptions, forming heterophilic text-attributed graphs (TAGs). Despite their sig- nificance, heterophilic TAGs remain underex- plored due to the lack of dedicated benchmarks that jointly capture heterophilic structures and rich textual attributes. To address this gap, we introduce the Heterophilic Text-attributed Graph Benchmark (HeTGB), a novel bench- mark comprising five real-world heterophilic graph datasets from diverse domains, with nodes enriched by extensive textual descrip- tions. HeTGB enables systematic evaluation of GNNs, pre-trained language models (PLMs) and co-training methods on the node classifi- cation task. Through extensive benchmarking experiments, we showcase the utility of text attributes in heterophilic graphs, analyze the challenges posed by heterophilic TAGs and the limitations of existing models, and provide in- sights into the interplay between graph struc- tures and textual attributes.