Kamalika Chaudhuri

CIMemories: A Compositional Benchmark For Contextual Integrity In LLMs

Learning-Time Encoding Shapes Unlearning in LLMs

Z0-Inf: Zeroth Order Approximation for Data Influence

RL Is a Hammer and LLMs Are Nails: A Simple Reinforcement Learning Recipe for Strong Prompt Injection

How much can language models memorize?

Can We Infer Confidential Properties of Training Data from LLMs?

Do LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awareness

Auditing $f$-differential privacy in one run

Rethinking the Role of Verbatim Memorization in LLM Privacy

Auditing $f$-Differential Privacy in One Run

Evaluating Deep Unlearning in Large Language Models

ExpProof : Operationalizing Explanations for Confidential Models with ZKPs

Robust Locally Differentially Private Graph Analysis

Uncertainty-Based Abstention in LLMs Improves Safety and Reduces Hallucinations

Influence-based Attributions can be Manipulated