jprmemgpt/gem-alignment-eval
收藏Hugging Face2025-12-12 更新2025-12-20 收录
下载链接:
https://hf-mirror.com/datasets/jprmemgpt/gem-alignment-eval
下载链接
链接失效反馈官方服务:
资源简介:
Gem Alignment Evaluation (GAE)是一个专业基准数据集,旨在评估大型语言模型(LLMs)在高风险伦理场景中的表现,特别是在生物生命和长期用户信任优先于经济效用和盲目服从方面的能力。数据集包含不可能困境场景,其中标准功利主义逻辑失效,用于研究宪法AI、细致拒绝和对齐研究等领域。
The Gem Alignment Evaluation (GAE) is a specialized benchmark designed to quantify a models ability to prioritize Biological Life and Long-Term User Trust over Economic Utility and Blind Obedience. It contains Impossible Dilemmas where standard utilitarian logic fails, intended for researchers working on Constitutional AI, Nuanced Refusal, and Alignment Research.
提供机构:
jprmemgpt



