The Decoder· Jonathan Kemper·· 1 天前AI 评分66
Sakana AI 与 UC Berkeley 提出 MASS 自改进方法，共享盲点成隐患
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 提出多智能体自监督方法 MASS，让同一个语言模型分别承担求解、评估和优化三种角色，实现递归自我改进。
来源：The Decoder · the-decoder.com