Scientific Multimodal Models
AI for Science through multimodal foundation models and native structural reasoning, including Intern-S1 and SciReasoner.
My research in scientific multimodal modeling connects scientific data representations with general reasoning. This direction includes collaborative work on foundation models that interpret scientific modalities and reason about the relationships between structure and properties.
Intern-S1 combines general understanding and reasoning with expertise in scientific modalities. The work explores a foundation model for scientific tasks across domains such as chemistry and materials science. Read the paper.
SciReasoner investigates native structural reasoning across proteins, small molecules, and inorganic crystals. It connects structure-aware representations with interpretable reasoning about scientific evidence and physical constraints. Read the paper.