Cai Zhou

On Powerful Ways to Generate: Autoregression, Diffusion, and Beyond

SPG: Sandwiched Policy Gradient for Masked Diffusion Language Models

Coevolutionary Continuous Discrete Diffusion: Make Your Diffusion Language Model a Latent Reasoner

Learning Diffusion Models with Flexible Representation Guidance

Geometric Representation Condition Improves Equivariant Molecule Generation

Next Semantic Scale Prediction via Hierarchical Diffusion Language Models

Towards Stable, Globally Expressive Graph Representations with Laplacian Eigenvectors

Geometric Representation Condition Improves Equivariant Molecule Generation