elvis@omarsar0
47AI 编辑部评分,满分 100
2026-08-01 00:55· 15分钟前
跳到正文
AI 摘要

OpenMLE 开源了面向机器学习工程递归自我改进的全栈系统,并后训练出 35B 元进化智能体 Frontis-MA1。在 MLE-Bench Lite 上,Medal Average 从 39.39% 提升至 60.61%,配合异步搜索和经验先验达 71.21%,超过 GPT-5.5 with Codex。

Great release for recursive self-improving in ML Engineering.

DAIR.AIVery interesting paper on recursive self-improvement. The whole stack is released. Machine learning engineering gives recursive self-improvement a concrete, exe...
elvis · @omarsar0 · X·2026-08-01 00:55·15分钟前
在 X 看原推· x.com
AI 摘要

OpenMLE 开源了面向机器学习工程递归自我改进的全栈系统,并后训练出 35B 元进化智能体 Frontis-MA1。在 MLE-Bench Lite 上,Medal Average 从 39.39% 提升至 60.61%,配合异步搜索和经验先验达 71.21%,超过 GPT-5.5 with Codex。

Great release for recursive self-improving in ML Engineering.

DAIR.AIVery interesting paper on recursive self-improvement. The whole stack is released. Machine learning engineering gives recursive self-improvement a concrete, exe...
在 X 查看原推x.com