Reimagines 3D molecules as continuous vector fields rather than discrete graphs, decoupling structure learning from atom types.
arXiv · March 16, 2026 · 2603.12734
Why it matters
This avoids the 'heterogeneous modality entanglement' of typical 3D graph models, providing a more stable continuous objective for diffusion-based drug discovery.
From the abstract
Generative modeling of three-dimensional (3D) molecules is a fundamental yet challenging problem in drug discovery and materials science. Existing approaches typically represent molecules as 3D graphs and co-generate discrete atom types with continuous atomic coordinates, leading to intrinsic learning difficulties such as heterogeneous modality entanglement and geometry-chemistry coherence constraints. We propose VecMol, a paradigm-shifting framework that reimagines molecular representation by m