14 papers
Improving the Convergence of Private Shuffled Gradient Methods with Public Data
Shuli Jiang, Pranay Sharma, Zhiwei Steven Wu +1
We consider the problem of differentially private (DP) convex empirical risk minimization (ERM). While the standard DP-SGD algorithm is theoretically well-established, practical im…
MarkTune: Improving the Quality-Detectability Trade-off in Open-Weight LLM Watermarking
Yizhou Zhao, Zhiwei Steven Wu, Adam Block
Watermarking aims to embed hidden signals in generated text that can be reliably detected when given access to a secret key. Open-weight language models pose acute challenges for s…
Membership Inference Attacks for Unseen Classes
Pratiksha Thaker, Neil Kale, Zhiwei Steven Wu +1
A key tool in developing safe AI models is \emph{data auditing}, i.e., using statistical tools to determine whether harmful content may have been used in the training data of a bla…
Unlearned but Not Forgotten: Data Extraction after Exact Unlearning in LLM
Xiaoyu Wu, Yifei Pang, Terrance Liu +1
Large Language Models are typically trained on datasets collected from the web, which may inadvertently contain harmful or sensitive personal information. To address growing privac…
On the Benefits of Public Representations for Private Transfer Learning under Distribution Shift
Pratiksha Thaker, Amrith Setlur, Zhiwei Steven Wu +1
Public pretraining is a promising approach to improve differentially private model training. However, recent work has noted that many positive research results studying this paradi…
Measurement as Bricolage: Examining How Data Scientists Construct Target Variables for Predictive Modeling Tasks
Luke Guerdan, Devansh Saxena, Stevie Chancellor +2
Data scientists often formulate predictive modeling tasks involving fuzzy, hard-to-define concepts, such as the "authenticity" of student writing or the "healthcare need" of a pati…