Materium: An Autoregressive Approach for Material Generation
This work addresses material generation for researchers and engineers, offering a faster alternative to diffusion methods, though it appears incremental as it adapts existing autoregressive techniques to a specific domain.
The authors tackled the problem of generating crystal structures by introducing Materium, an autoregressive transformer that converts 3D material representations into token sequences, enabling fast and scalable generation with training in a few hours on a single GPU and outperforming diffusion-based approaches in speed.
We present Materium: an autoregressive transformer for generating crystal structures that converts 3D material representations into token sequences. These sequences include elements with oxidation states, fractional coordinates and lattice parameters. Unlike diffusion approaches, which refine atomic positions iteratively through many denoising steps, Materium places atoms at precise fractional coordinates, enabling fast, scalable generation. With this design, the model can be trained in a few hours on a single GPU and generate samples much faster on GPUs and CPUs than diffusion-based approaches. The model was trained and evaluated using multiple properties as conditions, including fundamental properties, such as density and space group, as well as more practical targets, such as band gap and magnetic density. In both single and combined conditions, the model performs consistently well, producing candidates that align with the requested inputs.