A survey of 239 papers examines how AI agents may be improved through updates to the model or surrounding components such as prompts, memory, and tools.
Published on July 19, 2026, the survey covers 239 papers on self-improvement in agentic systems. Its scope includes changes to the model itself and to the scaffold around it—components such as prompts, memory, and tools—to improve agents.
The survey maps these approaches, but the available record does not detail comparative results or recommend a specific technique. To check its scope and conclusions, consult the original survey and verify the cited papers; do not infer effectiveness beyond the evidence presented.