发表机构
Aether AI; University of California San Diego; University of Illinois Chicago（Aether AI; 加州大学圣迭戈分校; 伊利诺伊大学芝加哥分校）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
RSIAgent提出无需训练的多智能体框架，通过自主记忆构建和先广后深探索实现递归自我改进，显著提升开源模型在新环境中的表现。
AI 中文摘要
数字智能体通常必须适应新环境，而预训练模型无法完全涵盖这些环境的界面、工具和故障模式。我们提出了RSIAgent，一个无需训练的多智能体框架，通过自主记忆构建实现递归自我改进。RSIAgent协调课程智能体、执行智能体和验证智能体，持续探索环境、验证结果，并保留环境特定知识，包括动作、条件和后果之间可复用的因果关系。它进一步采用“先广后深”的探索策略，将并行广泛的递归自我探索（用于发现多样的环境结构）与聚焦深入的自我探索（用于发现困难案例、隐藏约束、边界条件和先前未知的因果依赖）相结合。由此产生的记忆被冻结，可直接复用于下游任务，无需更新模型参数。在OSWorld-v2和Agent's Last Exam上的实验表明，RSIAgent显著提升了强开源模型，使Kimi-K3和GLM-5.3能够超越包括GPT-6在内的前沿闭源模型。
英文摘要
Digital agents must often adapt to new environments whose interfaces, tools, and failure modes are not fully captured by pretrained models. We introduce \textbf{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 \textbf{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.
Comments50 pages
