PurposeShift: Benchmark, Run Logs and Analysis Outputs for "Permission Is Not Purpose: Evaluating Purpose Limitation in LLM Agents"
收藏官方服务：
资源简介：
Synthetic benchmark (PurposeShift), complete run logs, per-segment metrics, tables, figures and the blinded detector audit for the paper. All data are synthetic. The code that generates the benchmark and reproduces all analyses is included.
提供机构：
Zenodo创建时间：
2026-11-04
