Rethinking Self-Evolving Agents: Do We Still Need Prescribed Optimization Pipelines?
2026-08-10 • Artificial Intelligence
Artificial Intelligence
AI summaryⓘ
The authors studied how agents improve themselves when guided either by fixed step-by-step rules or by a powerful AI model called an optimizer. They created a method called Open-Ended Optimization (OEO) where the optimizer decides how to improve things on its own, compared to traditional preset methods like SkillOpt and GEPA. Their experiments showed that the smart optimizer-driven OEO usually performs better and uses fewer resources than SkillOpt, but only when the optimizer is very capable. The authors conclude that fixed procedures help when the optimizer isn’t strong enough, but a strong optimizer can figure out its own way to improve.
self-evolving agentsoptimization pipelineOpen-Ended Optimization (OEO)SkillOptGEPAoptimizerevolutionary searchAI capabilitypersistent improvementresource budget
Authors
Hui Xue, Fan Yang
Abstract
Self-evolving agents are usually built around prescribed optimization pipelines: the framework decides how to gather evidence, revise a persistent artifact, select candidates, and stop. We ask whether this task-specific procedure remains necessary when a frontier model acts as the optimizer. We introduce Open-Ended Optimization (OEO), which keeps the objective, permitted interactions, resource budget, data boundary, and evaluation fixed while allowing the optimizer to compose the improvement process online. We compare OEO with two complementary prescribed approaches: SkillOpt, a staged pipeline with bounded edits, and GEPA, a reflective evolutionary search. Across 14 head-to-head comparisons over 8 benchmark-target-model settings, GPT-5.5-driven OEO records 12 wins, 1 tie, and 1 narrow loss of 0.21 percentage points. It uses a median 34.3 percent of SkillOpt's configured target-interaction token budget. A one-shot, zero-interaction control shows that the gains are not explained by a single prior-driven rewrite. However, delegation has a capability boundary: SkillOpt outperforms OEO with a medium optimizer, and a weak optimizer cannot operate through the unchanged OEO interface. In the fully instrumented OEO-SkillOpt pair, trajectory analysis further shows that prescription changes how optimization proceeds more consistently than it changes final behavior. Together, these findings recast prescribed pipelines as capability-dependent scaffolding: essential constraints remain external, but a sufficiently capable optimizer can compose the route from measurable feedback to persistent improvement.