LLM-SecDB benchmark
收藏官方服务：
资源简介：
This is the LLMSec-DB benchmark, containing 300 entries. It is designed to evaluate the effectiveness of different large language models (LLMs) in identifying embedded secrets and personally identifiable information (PII) within comments and text. The benchmark is part of the publication "You Have Been LaTeXpOsEd: A Large-Scale Systematic Analysis of Information Leakage in Preprint Archives Using Large Language Models", which has been accepted at the USENIX WOOT 2026.
提供机构：
Zenodo创建时间：
2026-05-06
