Topic

#privacy-auditing

1 article exploring privacy-auditing. Expert insights and analysis from our editorial team.

Showing 1–1 of 1 articles
Related topics:
differential-privacy llm-benchmarks ai-security machine-learning

Articles

Newest first
Security

DPrivBench: LLMs Score 99.5% on Textbook DP but Collapse on Advanced Reasoning

DPrivBench tests 11 LLMs on 713 differential-privacy instances. GPT-5-High hits 0.995 on textbook checks, but the best model reaches only F1 0.829 on advanced DP — and fails.
Browse All Topics