AI 换了个新环境就变笨？这个框架让它自己摸索、自己记笔记
现在的 AI 智能体一进陌生软件界面就露怯：按钮不认识、工具不会用、出错方式也猜不到，因为训练时没见过。这篇提出一个不用重新训练的多智能体框架，让 AI 自己当「探险家」:一个智能体负责出题、一个负责动手试、一个负责验收，把试出来的规律——比如「什么操作在什么条件下会引发什么后果」——记成一张可复用的因果笔记，下次直接翻笔记干活，不用改模型参数。它先用广度探索摸清环境全貌，再往深处挖那些刁钻的边界情况和隐藏规则。在 OSWorld-v2 和 Agent's Last Exam 两个测试上，它让开源模型 Kimi-K3 和 GLM-5.3 反超了闭源旗舰 GPT-6。这不是你明天能装上的插件，但它指向一个趋势：AI 适应新环境，可能不再靠更大的预训练，而是靠「到了现场自己学」的机制。
📄 原文摘要(英文)
Digital agents must often adapt to new environments whose interfaces, tools, and failure modes are not fully captured by pretrained models. We introduce RSIAgent, a training-free multi-agent framework for recursive self-improvement through autonomous memory construction. RSIAgent coordinates curriculum, actor, and verifier agents to continually explore the environment, validate outcomes, and retain environment-specific knowledge, including reusable causal relationships between actions, conditions, and consequences. It further adopts a broad-then-deep exploration strategy, combining parallel broad recursive self-exploration for discovering diverse environment structures with focused deep self-exploration for uncovering hard cases, hidden constraints, boundary conditions, and previously unknown causal dependencies. The resulting memory is frozen and can be directly reused for downstream tasks without updating model parameters. Experiments on OSWorld-v2 and Agent's Last Exam show that RSIAgent substantially improves strong open-source models, enabling Kimi-K3 and GLM-5.3 to outperform frontier closed-source models including GPT-6.