Agentic ESOpt：在极低GPU需求下微调长时程LLM智能体

摘要

强化学习（RL）在单轮大语言模型微调中已展现出良好前景。然而，长视界智能体推理引入了日益增多的分支交互和稀疏奖励，暴露了RL的若干局限：其基于反向传播的重型训练栈使得微调更大的大语言模型不切实际，且更长视界的轨迹使RL中的信用分配变得更加困难。本文论证进化策略（ES）可以成为微调长视界大语言模型智能体的更优选择。与智能体RL相比，ES具有三个关键优势：1）模型可扩展性：ES仅需极小的推理级GPU显存即可实现全参数优化，使得微调大语言模型成为可能；2）灵活性：其轻量级的黑盒反馈接口使ES微调易于与提示空间进化（如技能优化与测试时计算）组合使用；3）长视界可扩展性：ES执行轨迹级参数归因而无需跨视界分解奖励，随着视界长度增长，其可扩展性优于智能体RL。基于这一见解，我们提出Agentic ESOpt，一个专为灵活的参数-上下文协同进化而设计的全参数智能体微调框架。在每一步中，Agentic ESOpt在当前大语言模型参数周围采样扰动，以奖励评估所得智能体，并应用在线奖励加权更新。为改善探索-适应权衡，Agentic ESOpt进一步引入了扰动尺度σ的余弦衰减调度。在WebArena-Lite上，对Qwen-3.5-27B的全参数优化将无技能基线提升了6.69%。在测试时自动启发式设计中，Agentic ESOpt执行在线提示-参数协同进化，在36个设置中的28个中优于其匹配基线。

English

Reinforcement Learning (RL) has been promising in single-turn LLM fine-tuning. However, long-horizon agentic reasoning introduces increasingly branching interactions and sparse rewards, exposing several limitations of RL: its heavyweight backpropagation-based training stack makes it impractical to fine-tune larger LLMs, and longer-horizon trajectories make credit assignment in RL substantially harder. This paper argues that evolution strategies (ES) can be a better choice for fine-tuning long-horizon LLM agents. Compared with agentic RL, ES offers three key advantages: 1) Model Scalability: ES enables full-parameter optimization with only minimal, inference-level GPU memory, making it possible to fine-tune large LLMs. 2) Flexibility: its lightweight, black-box feedback interface makes ES fine-tuning easy to compose with prompt-space evolution (e.g., skill optimization & test-time compute); and 3) Long-Horizon Scalability: ES performs trajectory-level parameter attribution without decomposing rewards across horizons, yielding better scalability than Agentic RL as the horizon length grows. Based on this insight, we propose Agentic ESOpt, a full-parameter agentic fine-tuning framework tailored to flexible parameter--context co-evolution. At each step, Agentic ESOpt samples perturbations around the current LLM parameters, evaluates the resulting agents with rewards, and applies an online reward-weighted update. To improve the exploration--adaptation trade-off, Agentic ESOpt further introduces a cosine decay schedule of the perturbation scale σ. On WebArena-Lite, full-parameter optimization of Qwen-3.5-27B improves the No Skill baseline by 6.69%. In test-time automatic heuristic design, Agentic ESOpt performs online prompt--parameter co-evolution, improving its matched baseline in 28 of 36 settings.