Ya-Ping Hsieh

When Scores Learn Geometry: Rate Separations under the Manifold Hypothesis

Verifier-Constrained Flow Expansion for Discovery Beyond the Data

Reward-Guided Flow Merging via Implicit Density Operators

Flow Density Control: Generative Optimization Beyond Entropy-Regularized Fine-Tuning

Provable Maximum Entropy Manifold Exploration via Diffusion Models