AutoFyn Technical Report: Non-Parametric Expert Iteration for Long-Horizon Agents

AutoFyn 技术报告：面向长周期智能体的非参数专家迭代

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. 摘要： 我们介绍了 AutoFyn，这是一个受“专家迭代”（Expert Iteration）算法启发的智能体框架。它通过从已验证的奖励信号中更新持久化状态，而非更新模型权重，从而使冻结模型能够适应多轮任务。

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. 每一轮任务都从全新的模型会话开始，持久性信息仅通过明确的接口（如持久化内存文件、报告和代码库状态）重新引入。

Within a round, an orchestrator explores, plans and builds many alternative approaches with specialized agents, while a task-grounded verifier verifies the work and supplies an objective reward for measuring progress. 在每一轮中，编排器（Orchestrator）会利用专用智能体探索、规划并构建多种替代方案，同时由任务导向的验证器对工作成果进行验证，并提供客观的奖励以衡量进展。

This reward is distilled back into the persistent state, which updates the effective policy for the next round. In this technical report, we formalize this loop and describe its persistent state and verification interfaces. 该奖励被提炼回持久化状态中，从而更新下一轮的有效策略。在本技术报告中，我们将这一循环形式化，并描述了其持久化状态和验证接口。

We then demonstrate its use in three domains, namely olympiad mathematics, data science, and cybersecurity. On the six fresh problems of the 2026 International Mathematical Olympiad, every model with room to improve scores higher under AutoFyn than in its provider’s own coding agent. 随后，我们在奥林匹克数学、数据科学和网络安全这三个领域展示了其应用。在 2026 年国际数学奥林匹克竞赛的六道新题上，所有具备提升空间的模型在使用 AutoFyn 后，得分均高于其提供商自有的编码智能体。

AutoFyn also built the top-ranked agent on the Spider 2.0 dbt benchmark, and has produced 16 maintainer-confirmed vulnerability advisories in this http URL, MetaMask, pnpm, Warp, LiteLLM, Langflow, and Open WebUI. AutoFyn 还在 Spider 2.0 dbt 基准测试中构建了排名第一的智能体，并已在 this http URL、MetaMask、pnpm、Warp、LiteLLM、Langflow 和 Open WebUI 中发现了 16 个经维护者确认的漏洞公告。