SciReasoner: Deep Native Structural Reasoning Across Proteins, Molecules, and Crystals

The paper introduces SciReasoner, a multimodal scientific foundation model for native structural reasoning across proteins, small molecules, and inorganic crystals. It discretizes coordinates, topologies, and periodic connectivities into a unified structure-aware vocabulary, allowing structural tokens to serve as addressable evidence during reasoning. The abstract reports improvements in homology-controlled Gene Ontology prediction, single-step retrosynthesis, and materials band-gap regime separation. Across 86 benchmarks, SciReasoner reportedly reaches state-of-the-art results on 67 tasks. A double-blind expert evaluation found its reasoning traces preferred or at least comparable to those of a frontier LLM in 98% of cases.