Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials
NeurIPS 2026
Amir Masoud Nourollah
PhD student, Department of Computer Science, Cardiff University
I work on scientific machine learning for chemistry and molecular systems. My research asks when machine learning models can be trusted outside the data they were trained on: how they generalise under distribution shift, and how we can interpret and explain what they have learnt. Currently I study compositional generalisation in machine learning interatomic potentials for molecular dynamics.
I am a member of the Cardiff NLP and Knowledge Representation and Reasoning groups, supervised by Prof. Steven Schockaert and Prof. Stefano Leoni. Before my PhD I studied chemistry and then computer science, and worked for several years as a software engineer and machine learning researcher. More about me →
News
- 2026 Benchmarking Compositional Generalisation for Machine Learning Interatomic Potentials published at NeurIPS 2026.
- Jul 2025 Large Language and Reasoning Models are Shallow Disjunctive Reasoners published at ACL 2025.
- 2025 Presented my first-year review at Cardiff University.
- 2024 Started my PhD at Cardiff University.