A Benchmark Dataset for Graph Regression with Homogeneous and Multi-Relational Variants

TL;DR

This paper introduces RelSC, a new diverse graph-regression benchmark dataset derived from program graphs, to evaluate models on homogeneous and multi-relational structures with continuous targets.

Contribution

RelSC provides two variants capturing different relational complexities, enabling analysis of how structural representation affects graph regression model performance.

Findings

Models perform differently on homogeneous versus multi-relational graphs.

RelSC reveals the impact of structural representation on regression accuracy.

The dataset offers a challenging benchmark for future graph regression research.

Abstract

Graph-level regression underpins many real-world applications, yet public benchmarks remain heavily skewed toward molecular graphs and citation networks. This limited diversity hinders progress on models that must generalize across both homogeneous and heterogeneous graph structures. We introduce RelSC, a new graph-regression dataset built from program graphs that combine syntactic and semantic information extracted from source code. Each graph is labelled with the execution-time cost of the corresponding program, providing a continuous target variable that differs markedly from those found in existing benchmarks. RelSC is released in two complementary variants. RelSC-H supplies rich node features under a single (homogeneous) edge type, while RelSC-M preserves the original multi-relational structure, connecting nodes through multiple edge types that encode distinct semantic relationships. Together, these variants let researchers probe how representation choice influences model behaviour. We evaluate a diverse set of graph neural network architectures on both variants of RelSC. The results reveal consistent performance differences between the homogeneous and multi-relational settings, emphasising the importance of structural representation. These findings demonstrate RelSC's value as a challenging and versatile benchmark for advancing graph regression methods.