Symmetry-Breaking De Novo Crystal Generation via Markovian Jump Diffusion
SbCD generates full crystal structure specifications by moving across space groups instead of sampling them from empirical distributions.
Nguyen and Kalousis frame the method as a diffusion model that reverses from lowest-symmetry priors, borrowing the idea of spontaneous symmetry breaking from physics. Its Markovian jump-diffusion process is meant to model transitions between space groups during generation. In de novo tests on MP20 and MPTS-52, the paper says SbCD substantially outperforms a symmetry-preserving counterpart. The paper is under review. ArXiv · AI/CL/LG's note
Nguyen and Kalousis frame the method as a diffusion model that reverses from lowest-symmetry priors, borrowing the idea of spontaneous symmetry breaking from physics. Its Markovian jump-diffusion process is meant to model transitions between space groups during generation. In de novo tests on MP20 and MPTS-52, the paper says SbCD substantially outperforms a symmetry-preserving counterpart. The paper is under review. ArXiv · AI/CL/LG's note
score 5