Research topics
- Graph representations and message passing
- Equivariance and constrained optimization
- Robustness and interpretability of structured models
Recent methods include learning approximate equivariance through constrained optimization and adapting graph connectivity to capture long-range dependencies.
Selected publications
- Learning the Neighborhood: Contrast-Free Multimodal Self-Supervised Molecular Graph Pretraining 2026 · ICML
- Protein Fold Classification at Scale: Benchmarking and Pretraining 2026 · ICML · Oral presentation
- FACET: A Fragment-Aware Conformer Ensemble Transformer 2026 · ICLR
- Learning (Approximately) Equivariant Networks via Constrained Optimization 2025 · NeurIPS
- Adaptive Message Passing: A General Framework to Mitigate Oversmoothing, Oversquashing, and Underreaching 2025 · ICML
- Higher-Rank Irreducible Cartesian Tensors for Equivariant Message Passing 2024 · NeurIPS
- Probabilistic Graph Rewiring via Virtual Nodes 2024 · NeurIPS
- Ordered Subgraph Aggregation Networks 2022 · NeurIPS
- Learning Discrete Structures for Graph Neural Networks 2019 · ICML
- Learning Convolutional Neural Networks for Graphs 2016 · ICML
Information reviewed 26 September 2026 · University research topic