You can also find my articles on my Google Scholar profile. Names marked with an asterisk denote equal contribution.
Score Forgetting Distillation: A Swift, Data-Free Method for Machine Unlearning in Diffusion Models
A swift, data-free machine unlearning algorithm that rapidly removes target classes or concepts from diffusion models while preserving the original generative capability of others — with up to 1000× sampling speedup, ...
A Dense Reward View on Aligning Text-to-Image Diffusion with Preference
A dense-reward perspective on aligning text-to-image diffusion with human preference, turning the entire denoising trajectory — not just the final sample — into the optimization signal.
Beta Diffusion
A diffusion model defined directly on the simplex via Beta distributions, well-suited for bounded, normalized, or compositional data such as images and probability vectors.
Learning to Jump: Thinning and Thickening Latent Counts for Generative Modeling
We propose a binomial/Poisson-based hierarchical variational autoencoding framework that are well suited for modeling non-standard non-negative distributions that exhibit sparsity, skewedness, heavy-tailedness and/or ...
ASK: Adversarial Soft k-Nearest Neighbor Attack and Defense
In this work, we first introduce an information-theoretic surrogate loss for DkNN-based classification, based upon which we then propose an attack algorithm and a defense algorithm achieve SOTA adversarial results on ...
RAILS: A Robust Adversarial Immune-Inspired Learning System
We develop a novel adversarial defense framework inspired by the adaptive immune system: the Robust Adversarial Immune-inspired Learning System (RAILS)
Immuno-mimetic Deep Neural Networks (Immuno-Net)
In this paper we introduce a different type of biomimetic model, one that borrows concepts from the immune system, for designing robust deep neural networks.