5 papers
Closed-Form Concept Erasure via Double Projections
Chi Zhang, Jingpu Cheng, Zhixian Wang +1
While modern generative models such as diffusion-based architectures have enabled impressive creative capabilities, they also raise important safety and ethical risks. These concer…
Machine Unlearning under Retain-Forget Entanglement
Jingpu Cheng, Ping Liu, Qianxiao Li +1
Forgetting a subset in machine unlearning is rarely an isolated task. Often, retained samples that are closely related to the forget set can be unintentionally affected, particular…
Deep learning and the rate of approximation by flows
Jingpu Cheng, Qianxiao Li, Ting Lin +1
We investigate the dependence of the approximation capacity of deep residual networks on its depth in a continuous dynamical systems setting. This can be formulated as the general…
A unified framework for establishing the universal approximation of transformer-type architectures
Jingpu Cheng, Ting Lin, Zuowei Shen +1
We investigate the universal approximation property (UAP) of transformer-type architectures, providing a unified theoretical framework that extends prior results on residual networ…
Allocation of Parameters in Transformers
Ruoxi Yu, Haotian Jiang, Jingpu Cheng +3
Transformers have achieved remarkable successes across a wide range of applications, yet the theoretical foundation of their model efficiency remains underexplored. In this work, w…