We first derive an information-theoretic bound that decomposes the token rejection rate into contributions from SLM-LLM distribution mismatch and from quantization distortion. Guided by this analysis, we propose the Sparse Quantize-and-Sample SD (SQS-SD) framework, which exploits distributional sparsity through structured sparsification and lattice-based quantization.

We propose PROPS (PROgressively Private Self-alignment), a multi-stage privacy preserving alignment framework where privately aligned models in previous stages can serve as labelers for supplementing training data in the subsequent stages of alignment.

We develop a framework that combines deep learning and epidemic models to jointly forecast and learn epidemic dynamics using multiple datasets and show that even DP-protected financial data improves forecasting. 

This study proposes a framework for fine-tuning large language models (LLMs) with differential privacy (DP) to perform multi-abnormality classification on radiology report text. The framework seeks to mitigate the privacy risks associated with sensitive patient data and protect against data leakage while maintaining classification performance. 

We introduce the concept of Inference Privacy (IP), a new framework designed to ensure privacy for user’s query/ input data during inference. The core idea behind IP is to obscure model outputs to the extent that adversaries are unable to discern the specific query input within a defined privacy radius.