Jan Leike· janleike·· 2025-07-25AI 评分57
Jan Leike 分享 AI 安全对齐研究新进展
In March we published a paper on alignment audits: teams of humans were tasked to find the problems in model we trained to be misaligned....
AI 导读
Jan Leike 在 X 上发布了关于 AI 安全对齐领域的最新研究成果或观点。该动态聚焦于提升模型行为与人类意图一致性的技术路径。
来源：Jan Leike · x.com