Research / Selected publications
Exploring generative intelligence, one paper at a time.
A selection of work on diffusion, generation, and alignment.
20 publications
arXiv preprint, 2026
A controlled comparison of standalone drafting and continuous diffusion under matched training budgets, examining recurrent memory and iterative full-sequence refinement without target-model verification.
arXiv preprint, 2026
LangDiT improves continuous diffusion language generation through normalization, noise conditioning, a calibrated likelihood shortcut, and confidence-gated self-conditioning.
ICML 2026
Reinforcement learning for few-step generative models.
ICLR 2026
Distill a discrete diffusion language model into a few-step generator through integral KL matching.
ICLR 2025
Easy Consistency Tuning adapts pre-trained diffusion models for efficient consistency generation.
ICML , 2025
preprint, 2025
NeurIPS, 2025
IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI) , 2025
NeurIPS, 2025
Transactions on Machine Learning Research (TMLR) , 2024
Bringing human preferences to one-step image generation.
preprint , 2024
CVPR , 2024
NeurIPS, 2024
Data-free one-step diffusion distillation through implicit score-divergence gradients.
ICML , 2024
Adaptive variational scores enable efficient diffusion transport and simulation-free score training.
NeurIPS , 2023
NeurIPS , 2023
Transferring the knowledge of diffusion models into one-step generators.
arXiv preprint , arXiv:2304.04262, 2023
A structured overview of diffusion distillation, implicit generators, and training-free accelerated sampling.
NeurIPS , 2023
NeurIPS , 2023
Try another keyword or choose a different year.
* indicates equal contribution or corresponding author. Explore the research ›