نسخة أولية وصول مفتوح
SAGE: A Statistical Acceptance Gate for Self-Evolving Agents
Large Language Model (LLM)-based agents increasingly self-evolve by editing a persistent skill document that encodes their workflow, tool-use rules, and decision logic. This loop has two steps, an optimizer that proposes a candidate edit and a gate that accepts or rejects it. Prior work has concentrated on the optimize …