cs.CL提交于 2026-10-28已译

GEPA-Curator: 面向自改进智能体的精选学习

GEPA-Curator: Curated Learning for Self Improving Agents

Zack Yu · Nicholas Lee · Minjae Lee · Lakshya Agarwal · Luca Manolache · Wonjun Kang · et al.

中文摘要

大语言模型（LLMs）正被用于需要与复杂环境持续交互的多轮智能体工作负载。现有的反思式提示优化器（如 GEPA）从失败的执行轨迹中学习，但智能体失败可能源于提示弱点、框架限制、随机执行或噪声评估器。这些失败可能无法提供有用的学习信号，甚至有害，因此精选提供给优化器的任务至关重要。GEPA-Curator 框架通过选择哪些智能体经验应塑造提示优化来增强优化器。它包含两个组件：任务精选，过滤可能提供噪声或不可靠信号的任务；以及先前成功检索，选择性检索任务的过往成功轨迹作为额外上下文。在 GoBrowse、ALFWorld、AppWorld 和 τ-bench 上，使用 Gemini 3.1 Pro 和 Claude Opus 5 进行评估。GEPA-Curator 匹配或超过两个基线的平均性能，相比 GEPA 提升高达 20.7%，相比 ACE 提升 21.2%。

关键要点

01指出智能体失败可能源于非提示因素，导致现有优化器从噪声信号中学习甚至受损。
02提出 GEPA-Curator 框架，通过任务精选过滤噪声任务，并检索过往成功轨迹作为优化上下文。
03在 GoBrowse、ALFWorld 等基准上，使用 Gemini 3.1 Pro 和 Claude Opus 5 验证框架有效性。
04实验结果显示，该方法相比 GEPA 和 ACE 基线分别取得最高 20.7% 和 21.2% 的性能提升。

解读

尚无解读。

原始英文摘要

arXiv:2610.37651v1 Announce Type: new Abstract: Large Language Models (LLMs) are increasingly used in multi-turn agentic workloads that require sustained interaction with complex environments. Existing reflective prompt optimizers such as GEPA learn from failed rollouts, but agentic failures can also come from prompt weaknesses, harness limitations, stochastic execution, or noisy judges. We argue that such failures may not provide useful learning signal, and may even be harmful; and thus it is essential to curate which tasks are provided to the prompt optimizer for better performance. We introduce GEPA-Curator, a framework that augments the prompt optimizer to select which agent experiences should shape prompt optimization. In particular, GEPA-Curator consists of two components: task curation, which filters out tasks that are likely to provide noisy or unreliable learning signals, and prior success retrieval, which selectively retrieves past successful trajectories for a task and provides them as additional context to the prompt optimizer. We evaluate GEPA-Curator on GoBrowse, ALFWorld, AppWorld, and $\tau$-bench, using both Gemini 3.1 Pro and Claude Opus 5 as optimizer models. GEPA-Curator matches or exceeds the mean performance of both baselines, with gains of up to 20.7% over GEPA and 21.2% over ACE.