7 papers
DiffRatio: Training One-Step Diffusion Models Without Teacher Supervision
Wenlin Chen, Mingtian Zhang, Jiajun He +4
Score-based distillation methods (e.g., variational score distillation) train one-step diffusion models by first pre-training a teacher score model and then distilling it into a on…
Towards Healing the Blindness of Score Matching
Mingtian Zhang, Oscar Key, Peter Hayes +3
Score-based divergences have been widely used in machine learning and statistics applications. Despite their empirical success, a blindness problem has been observed when using the…
VarDiU: A Variational Diffusive Upper Bound for One-Step Diffusion Distillation
Leyang Wang, Mingtian Zhang, Zijing Ou +1
Recently, diffusion distillation methods have compressed thousand-step teacher diffusion models into one-step student generators while preserving sample quality. Most existing appr…
Training Neural Samplers with Reverse Diffusive KL Divergence
Jiajun He, Wenlin Chen, Mingtian Zhang +2
Training generative models to sample from unnormalized density functions is an important and challenging task in machine learning. Traditional training methods often rely on the re…
Variational f-divergence Minimization
Mingtian Zhang, Thomas Bird, Raza Habib +2
Probabilistic models are often trained by maximum likelihood, which corresponds to minimizing a specific f-divergence between the model and data distribution. In light of recent su…
Active Preference Learning for Large Language Models
William Muldrew, Peter Hayes, Mingtian Zhang +1
As large language models (LLMs) become more capable, fine-tuning techniques for aligning with human intent are increasingly important. A key consideration for aligning these models…