跳到正文
The Decoder· Jonathan Kemper·· 3 小时前AI 评分60

Sakana AI 与 UC Berkeley 提出 MASS 自我改进方法,共享盲区成递归自我改进隐患

Sakana AI's self-improving model gets better at solving and judging, but can't catch its own blind spots

AI 导读

Sakana AI 与 UC Berkeley 提出 Multi-Agent Self-Supervision(MASS),让同一个语言模型在无更强教师模型的情况下兼任解题、评判与优化三种角色,实现递归自我改进。

来源:The Decoder · the-decoder.com