Publications

2025

1. 

   Covered Forest: Fine-grained generalization analysis of graph neural networks

   Antonis Vasileiou, Ben Finkelshtein, Floris Geerts, and 2 more authors

   ICML, 2025

   Abs HTML

   We extend graph similarity theory to assess the influence of graph structure, aggregation, and loss functions on MPNNs’ generalization abilities.

2. 

   Position: Graph Learning Will Lose Relevance Due To Poor Benchmarks

   Maya Bechler-Speicher, Ben Finkelshtein, Fabrizio Frasca, and 9 more authors

   ICML, 2025

   Abs HTML

   This paper calls towards more meaningful benchmarks, rigorous evaluation protocols, and stronger collaboration with domain experts to drive advances in graph learning research.

2024

1. 

   Almost Surely Asymptotically Constant Graph Neural Networks

   Sam Adam-Day, Michael Benedikt, İsmail İlkan Ceylan, and 1 more author

   NeurIPS, 2024

   Abs HTML

   A new angle on the expressive power of GNNs by studying how the predictions of a GNN probabilistic classifier evolve.

2. 

   Learning on Large Graphs using Intersecting Communities

   Ben Finkelshtein, İsmail İlkan Ceylan, Michael Bronstein, and 1 more author

   NeurIPS, 2024

   Abs HTML

   A new and fundamentally different pipeline for learning on very large non-sparse graphs using intersecting communities.

2023

1. 

   Cooperative Graph Neural Networks

   Ben Finkelshtein, Xingyue Huang, Michael Bronstein, and 1 more author

   ICML, 2023

   Abs HTML

   A more dynamic and flexible message-passing paradigm in which each node can choose a different communication strategy.

2022

1. 

   Single-Node Attacks for Fooling Graph Neural Networks

   Ben Finkelshtein, Chaim Baskin, Evgenii Zheltonozhskii, and 1 more author

   Neurocomputing, 2022

   Abs HTML

   Showcased that GNNs are vulnerable to a realistic single-node adversarial attack, even when the attacker cannot be chosen.

2. 

   A Simple and Universal Rotation Equivariant Point-cloud Network

   Ben Finkelshtein, Chaim Baskin, Haggai Maron, and 1 more author

   ICML workshop on Topology, Algebra, and Geometry in Machine Learning, 2022

   Abs HTML

   A simple architecture which is equivariant to rigid motions with the ability to approximate any equivariant function.

3. 

   Strategic Classification with Graph Neural Networks

   Itay Eilat, Ben Finkelshtein, Chaim Baskin, and 1 more author

   ICLR, 2022

   Abs HTML

   Learning in a setting where users that are dependant can modify their features to obtain favorable predictions.