AutoFyn：面向长时程智能体的非参数专家迭代技术报告
原标题：AutoFyn Technical Report: Non-Parametric Expert Iteration for Long-Horizon Agents
AI 摘要
AutoFyn 是一种受专家迭代算法启发的智能体框架，通过持久化状态而非模型权重来适应冻结模型，每轮从全新上下文开始，仅通过显式接口引入持久信息。在 2026 年国际数学奥林匹克六道新题上，AutoFyn 使所有可改进模型的得分均高于其提供商的编码智能体；在 Spider 2.0 dbt 基准上构建了排名第一的智能体，并在 Next.js、MetaMask 等开源项目中发现了多个维护者确认的漏洞。该框架旨在解决长时任务中的上下文累积和错误累积问题。
正文节选
AutoFyn Technical Report: Non-Parametric Expert Iteration for Long-Horizon Agents Abstract We introduce AutoFyn, an agent harness inspired by the Expert Iteration algorithm, adapting a frozen model across many rounds by updating persistent state from verified reward signals rather than model weights. Each round begins from a fresh model session, and durable information is reintroduced only through explicit interfaces such as persistent memory files, reports, and repository state. Within a round,