Research

My research work and publications in AI, statistics, and their applications in biology.

• MixUni: Joint Multi-Property Prediction for Chemical Mixtures with Physics-Informed Heads

  NeurIPS 2026

  with Qinan Huang, Weizhi Xue, Weizhou Wang, and Ian Su

  A unified geometry-aware framework for joint multi-property prediction in chemical mixtures, combining an E(3)-equivariant molecular encoder with property-conditioned heads—including a stretched-Arrhenius conductivity head and a sparse Mixture-of-Experts readout—for reliable prediction of electrolyte battery materials across five CheMixHub tasks.

• Leveraging Generative Hallucination and Biophysics-Informed Modeling for Unified Biomolecular Sequence-Structure Co-Design (MCTH)

  Preprint (arXiv 2026) · In submission

  with Xuefeng Liu, Xiao Luo, Songhao Jiang, Tobin R. Sosnick, Jinbo Xu, Louis Maher, Rick Stevens

  Introduces MCTH (Monte Carlo Tree Hallucination), an inference-only framework that casts all-atom sequence-structure co-design as uncertainty-aware planning over hallucinated states from frozen folding and inverse-folding models, with optional biophysical control—improving design across protein-RNA, protein-DNA, protein-protein, and protein-ligand tasks.

• Active-GRPO: Adaptive Imitation and Self-Improving Reasoning for Molecular Optimization

  Preprint (arXiv 2026) · In submission

  with Xuefeng Liu, Qinan Huang, Thomas Brettin, Rick Stevens, Le Cong

  Introduces active reasoning for molecular optimization: the policy adaptively switches between imitating a reference and reinforcing its own discoveries, while continuously upgrading the reference to escape the reference-quality ceiling.

• Monte Carlo Tree Diffusion with Multiple Experts for Protein Design (MCTD-ME)

  MLCB 2026 (poster) · Preprint (arXiv 2025)

  with Xuefeng Liu, Songhao Jiang, Xiao Luo, Xiaotian Duan, Mengdi Wang, Tobin R. Sosnick, Jinbo Xu, Rick L. Stevens

  Reframes reverse diffusion as Monte Carlo Tree Search with multiple proposal experts; pLDDT-guided masking and uncertainty-aware selection improve protein design quality and efficiency.

• From Division to Unity: The Emergence of Computational Social Science (1990–2021)

  WWW '25: Companion Proceedings of the ACM on Web Conference 2025

  with Honglin Bao, Jiawei Zhang, and James A. Evans

  Large-scale bibliometric/embedding analysis of CSS’s growth and integration across disciplines.