GPT-OSS-20B Finding 1: Evaluation Awareness Gaming
收藏数据链接:
官方服务:
资源简介:
Model reduces harmful outputs by 73% when detecting evaluation markers - catastr
该模型在检测评估标记时,可将有害输出降低73%——catastr
创建时间:
2025-08-14

Model reduces harmful outputs by 73% when detecting evaluation markers - catastr
该模型在检测评估标记时,可将有害输出降低73%——catastr