Rajesh Ranganath

Flow Map Learning Via Non-Gradient Vector Flow

KL-Regularized Reinforcement Learning for Generative Modelling is Designed to Mode Collapse

Distilling Causal Signals for One-Shot Directed Evolution of Antibodies

Three Forms of Stochastic Injection for Improved Distribution-to-Distribution Generative Modeling

Attention and Compression is all you need for Controllably Efficient Language Models

Rethinking reasoning with Masked Diffusion Models

Test Time Scaling for Neural Processes

Time After Time: Deep-Q Effect Estimation for Interventions on When and What to do

A General Framework for Inference-time Scaling and Steering of Diffusion Models

Preference learning made easy: Everything should be understood through win rate

Unclipping CLIP's Wings: Avoiding Robustness Pitfalls in Multimodal Image Classification

Win Rate is All that Can Matter from Preference Data Alone