Generative Modeling of Scientific Data

We are focused on the following topics:

• Biological Sequence Modeling
• Deep Generative Models for Geometric Graph Generation
• Scaled Structure-based Generative Models

Large Language Model

We propose the Decoupled Clip and Dynamic sAmpling Policy Optimization (DAPO) algorithm, and fully open-source a state-of-the-art large-scale RL system that achieves 50 points on AIME 2024 using Qwen2.5-32B base model. [code] [paper]

Foundation Models of AI4S

We aim to build the universal foundation model of science, some milestones are listed:

• Steering Protein Family Design through Profile Bayesian Flow
• ESM All-Atom: Multi-scale Protein Language Model for Unified Molecular Modeling