← back to paper
arxiv: 2604.15851 · 2 revisions
DPrivBench: Benchmarking LLMs' Reasoning for Differential Privacy