Self-Improvements in Modern Agentic Systems: A Survey
Quick Answer
This paper shows that This survey discusses the evolution of self-improving autonomous agents, emphasizing their transition from prototypes to deployed systems.
Quick Take
This survey discusses the evolution of self-improving autonomous agents, emphasizing their transition from prototypes to deployed systems. It presents a framework for adaptive systems that leverage experience for capability gains, formalizing self-improvement through a self-induced update operator. The paper organizes existing research by update targets and driving signals, while also addressing applications, evaluation methods, and future challenges.
Key Points
- Self-improving agents evolve with minimal human input, enhancing their capabilities.
- The framework couples foundation models with operational components like memory and control logic.
- Self-improvement is formalized as a self-induced update operator for model parameters.
- The survey categorizes prior work by update targets and signals driving change.
- Open problems and future directions for research are discussed.
Paper Resources
📖 Reader Mode
~2 min readAuthors:Zhe Ren, Yimeng Chen, Dandan Guo, Guowei Rong, Tonghui Li, R. B. Xiong, Qingfeng Lan, Wenyi Wang, Li Nanbo, Yibo Yang, Mingchen Zhuge, Jürgen Schmidhuber
Abstract:Self-improving autonomous agents are moving from research prototypes to deployed systems. The primary goal is controllable evolution, or adaptation, from experience with minimal or even no human input. This survey frames modern self-improving agents as adaptive systems that convert experience into accumulated capability gains. We offer a system-level framework that represents a modern agent as a configuration coupling a foundation model with an operational scaffold of prompts, memory, tools, and control logic. Within this framework, self-improvement is formalized as a self-induced update operator that obtains and commits updates to model parameters or scaffold components. We organize prior work by update target and by the signals that drive change, then review applications and discuss evaluation, before closing with open problems and future directions. For convenience, we track technical updates on this https URL.
| Comments: | 97 pages, 12 figures. Project page: this https URL Repository: this https URL |
| Subjects: | Artificial Intelligence (cs.AI); Computation and Language (cs.CL); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.13104 [cs.AI] |
| (or arXiv:2607.13104v1 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.13104 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Zhe Ren [view email]
[v1]
Tue, 14 Jul 2026 09:12:57 UTC (4,958 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.AI
See more →Automatic Ordinary Differential Equations Discovery For Biological Systems Using Large Language Model Powered Agentic System
The MEDA system utilizes large language models and symbolic regression to autonomously discover ordinary differential equations for biological systems, achieving strong structural recovery and biologically plausible models. It outperforms existing methods by integrating domain knowledge and mechanistic constraints, demonstrating effective retrieval and extrapolation capabilities.