Shuchen Xue

Fast-dLLM: Training-free Acceleration of Diffusion LLM by Enabling KV Cache and Parallel Decoding

Fast-dLLM v2: Efficient Block-Diffusion LLM

Variational Autoencoding Discrete Diffusion with Enhanced Dimensional Correlations Modeling

Any-Order GPT as Masked Diffusion Model: Decoupling Formulation and Architecture

Advantage Weighted Matching: Aligning RL with Pretraining in Diffusion Models

Graffe: Graph Representation Learning Enabled via Diffusion Probabilistic Models

Improved Diffusion-based Generative Model with Better Adversarial Robustness

Noise Consistency Training: A Native Approach for One-step Generator in Learning Additional Controls

Learning Single Index Models with Diffusion Priors