难得看到一个把“AI 改进 AI”落到完整工程闭环的项目。 OpenRSI 用 Gym 封装任务、隔离执行和自动评测，RL 基于反馈训练 Draft、Improve、Debug、Crossover 四类原子算子，Evo 再将它们编排成长程异步搜索。分层设计既能拆开验证模型与搜索的增益，也方便替换任务、模型和算力后端。组件环境相互独立，许可限非商业用途。

清华+Frontis开源的RSI项目OpenRSI：把"AI 改进 AI"做成了可执行的工程问题，框架和思想值得学习。近期递归自我进化RSI被炒得非常火，比如田渊栋的Recursive Superintelligence刚拿了几亿美元的融资，相比于Recursive，OpenRSI开源透露了很多细节。 首发成果是一个 35B 模型：在MLE基准上，单张 RTX 4090（限 12GB…

(5/7) 🧪 With OpenMLE-Evo fixed, Frontis-MA1-35B raises Human Rank across all five modalities and never lowers Medal Rate. Its 14 extra medals span image / text / tabular / audio / multimodal: +2 /…

Recursive self-improvement is the most consequential question in AI. It's being answered behind closed doors. Today we launch OpenRSI — open, executable, reproducible research on AI that improves AI.…