Benchmarking GNN Models on Molecular Regression Tasks with CKA-Based Representation Analysis
Rajan, Ishaan Gupta · 2026
Read Benchmarking GNN Models on Molecular Regression Tasks with CKA-Based Representation Analysis by Rajan, Ishaan Gupta (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-20573.txt.
Description
Molecules are often written as SMILES strings, which convert to graphs with atoms as nodes and bonds as edges for neural net training. Rajan and Ishaan Gupta note that graph neural nets for molecular property prediction still lack a rigorous benchmark. Their paper proposes MolGraphBench, a benchmark of four common graph neural net models for molecular regression tasks, judged on absolute performance, training speed, transfer learning, and prediction quality.
Results point to graph convolutional nets and graph isomorphism nets as the best architectures for molecular graph regression. The study also finds that molecular fingerprints add nothing complementary in fused graph-fingerprint models. The models match or beat current state-of-the-art graph baselines across three data sets. The paper concludes that the choice of graph layer should be treated as a tunable setting rather than a fixed design choice.
Catalog record
| CREATOR | Rajan, Ishaan Gupta |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 22 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2602-20573