The Path to Recursive Self-Improving Agents: Foundation, Framework, and Future Directions
Preprint
- 3 August 2026
- preprint
- Published by MDPI AG in Preprints.org
Abstract
Agent systems based on foundation models have advanced rapidly and can now accomplish increasingly complex tasks with greater autonomy. However, manually developing and refining agent systems requires substantial human effort and time, motivating a shift toward agent systems that can improve themselves. In this survey, we study self-improving agent systems that autonomously transform experience and evaluation feedback into persistent updates to their components. To compare the self-improvement capabilities of diverse agent systems, we introduce a five-level grading standard, ranging from manual improvement to general recursive self-improvement. Furthermore, we propose a unified research framework that jointly models the foundation model, agent harness, agent data system, agent trainer, and the mechanism governing their improvement as a coupled system. Guided by this framework, we systematically review the literature on self-improvement and provide a taxonomy covering self-improvement in individual components and co-improvement across components. In addition, this survey identifies open research problems on the path toward recursive self-improving agents. Overall, this survey provides a foundation for studying self-improving agent systems and a roadmap for future research.Keywords
All Related Versions
This publication has 206 references indexed in Scilit:
- Reconstructing Qwen3 into Qwen3-Protein: A Versatile Biophysics-Grounded Protein Language Foundation Model for PTMs and Downstream Tasks2026
- Intelligent surgical workflow recognition-based skill assessment for minimally invasive submandibular gland resectionnpj Digital Medicine, 2026
- Exp-SWE-Agent: A Memory-Enhanced Coding Agent for Sequential Issue Resolution2026
- AI Agents in Autonomous Retail Data Pipelines: Self-evolving ETL for Customer Behavior Data2026
- Agentic skill discoveryRobotics and Autonomous Systems, 2025
- Sia: Heterogeneity-aware, goodput-optimized ML-cluster scheduling2023
- Online evolutionary batch size orchestration for scheduling deep learning workloads in GPU clusters2021
- Optimus2018
- Gödel Machines: Fully Self-referential Optimal Universal Self-improvers2007
- Speculations Concerning the First Ultraintelligent Machine1966