GEPA-Curator: Curated Learning for Self Improving Agents

Hugging Face Daily Papers Papers
prompt-optimization agentic-ai self-improving-agents llm task-curation trajectory-retrieval

Summary

GEPA-Curator is a framework that improves agentic prompt optimization by filtering noisy failure signals and retrieving prior successful trajectories, achieving gains of up to 20.7% over GEPA and 21.2% over ACE on benchmarks like GoBrowse, ALFWorld, AppWorld, and τ-bench.

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 τ-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.
Original Article
View Cached Full Text

Cached at: 11/02/26, 07:31 PM

Paper page - GEPA-Curator: Curated Learning for Self Improving Agents

Source: https://huggingface.co/papers/2610.37651 Authors:

,

,

,

,

,

,

,

,

,

Abstract

GEPA-Curator improves agentic prompt optimization by filtering unreliable failure signals and retrieving past successes, boosting performance across multiple benchmarks.

Large Language Models (LLMs) are increasingly used in multi-turnagentic workloadsthat require sustained interaction with complex environments. Existingreflective prompt optimizerssuch asGEPAlearn from failedrollouts, 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 introduceGEPA-Curator, a framework that augments the prompt optimizer to select which agent experiences should shapeprompt 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, andprior success retrieval, which selectively retrieves past successful trajectories for a task and provides them as additional context to the prompt optimizer. We evaluateGEPA-Curator on GoBrowse, ALFWorld, AppWorld, and τ-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% overGEPAand 21.2% over ACE.

View arXiv pageView PDFAdd to collection

Get this paper in your agent:

hf papers read 2610\.37651

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2610.37651 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2610.37651 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2610.37651 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.

Similar Articles

CurateEvo: Data-Curation Evolving for Agentic Post-Training

arXiv cs.CL

CurateEvo is a failure-driven dynamic evolution framework for agentic post-training data curation. It iteratively rewrites curation strategies using failed trajectories, improving effectiveness and efficiency on benchmarks like ACEBench-Agent and BFCL-V4.

Capability-Grounded Agentic Planning

arXiv cs.AI

The paper presents CGAP, a neuro-symbolic framework that compiles tool catalogs into a PDDL model and uses a classical planner with LLM-based goal parsing to guarantee executable multi-tool plans, outperforming standard tool-calling agents by up to 28.9 points on four benchmarks.