The International Conference on Learning Representations (ICLR) took place from April 23–27, 2026, in Rio de Janeiro, Brazil. It is one of the world’s leading conferences on deep learning and AI research, showcasing advances in machine learning, computer vision, language processing, and robotics.
Trung Nguyen from the Interactive Machine Learning (IML) Department presented the paper “FACET: A Fragment-Aware Conformer Ensemble Transformer“. The research introduces FACET, a scalable structure-aware graph transformer that efficiently combines 2D molecular graphs with large ensembles of 3D conformers for molecular property prediction. The approach enables faster and more flexible integration of structural information while achieving state-of-the-art results in molecular prediction and reaction modeling tasks.
Trung Nguyen presented a poster at the ICLR 2026
