SOURCE-LINKED INTELLIGENCE

RobustSGPO: Search-Space Control for Agent Harness Evolution

arXiv · AI, language, vision and robotics · article · Sep 9, 2026 · UTC

Semantic-gradient-based prompt optimization (SGPO) improves agent harnesses using execution feedback, but its local update rule leaves the choice of edit scope and operation unresolved. We introduce RobustSGPO, which specifies the requested edit, constructs and checks the patch, and continues search from either the incumbent or retained snapshots. We evaluate permission scheduling, cumulative controls, and task-family transfer in the AgentX brainstorming workflow using 120 tasks, 95 runs, and 7,350 candidate attempts. Periodic $1\to2\to3$ scheduling exceeds fixed maximum permission by 0.28 tes

Read original source ↗ Open in workspace

recordType

paper

region

Global