Zatom-2: One Atomistic Generative Model Unifies Molecules, Materials and Proteins
rishabh16_ · x · 2026-10-11
Zatom-2 is a single atomistic foundation model spanning molecules, periodic materials, and (via finetuning) proteins, aiming to replace domain-specific modeling stacks with one shared backbone.
- Pretraining: 5M DFT-labeled structures, combining OMol25 (4M molecular systems with diverse chemistry and non-equilibrium geometries) and OMat24 (1M periodic inorganic systems), all providing energies and forces for physics-grounded auxiliary supervision.
- Architecture: a multiscale U-shaped Transformer (atom encoder → token trunk → atom decoder) that keeps atom-level features for local geometry and token-level features for global exchange, with pair-biased attention and adaptive normalization.
- Open source: training code, paper configs, data prep tools, inference CLI and weights are available in the Nucleus GitHub repo, with an arXiv paper and PyPI install.
More from Research
- Looped LM paper: 1.6B model matches full-cache baseline with 3x smaller KV cache — rupspace · 2026-10-11
- Pure RL discovers superhuman robot strategies in sim, transfers zero-shot to real hardware — KyleMorgenstein · 2026-10-11
- Softmax picks probabilities, cross-entropy picks the target: a 3-class walkthrough — techNmak · 2026-10-11
- PartLLM brings LLM-powered 3D mesh part segmentation to SIGGRAPH Asia with code released — Promptmethus · 2026-10-11
- Duo Bregman pseudo-divergence yields closed-form KL divergence between truncated Gaussians — FrnkNlsn · 2026-10-11
- 228 tasks, 26 models: METR's time horizon curve is nearly flat from 2-30 min — lulzxdxdxd · 2026-10-11