# A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems

**Authors:** Jinyuan Fang, Yanwen Peng, Xi Zhang, Yingxu Wang, Xinhao Yi, Guibin Zhang, Yi Xu, Bin Wu, Siwei Liu, Zihao Li, Zhaochun Ren, Nikos Aletras, Xi Wang, Han Zhou, Zaiqiao Meng  
**arXiv:** [2508.07407](https://arxiv.org/abs/2508.07407)  
**Date:** 2025-08-10  
**Summary page:** https://www.summarizepaper.com/en/paper/2508.07407v1/

## Key points

- Recent advances in large language models (LLMs) have sparked interest in developing AI agents capable of solving complex real-world tasks.
- Existing agent systems often rely on manually crafted configurations that remain static after deployment, limiting their adaptability to dynamic environments.
- Research has explored agent evolution techniques to enhance agent systems automatically based on interaction data and environmental feedback.
- Self-evolving agentic systems bridge static capabilities with continuous adaptability required by lifelong agentic systems.
- The survey introduces a unified conceptual framework highlighting four key components: System Inputs, Agent System, Environment, and Optimisers.
- Various self-evolving techniques targeting different components of the agent system are systematically reviewed.
- Domain-specific evolution strategies have been developed for specialized fields such as biomedicine, programming, and finance.
- Memory optimization techniques support coherent reasoning and decision-making over time.
- Tool use is discussed as a key factor for AI agents to effectively operate in real-world scenarios by extending their functional scope through external tools like web search engines and code interpreters.
- Evaluation, safety considerations, and ethical implications for self-evolving agentic systems are addressed to ensure effectiveness and reliability.

## Summary

Recent advances in large language models (LLMs) have sparked interest in developing AI agents capable of solving complex real-world tasks. However, existing agent systems often rely on manually crafted configurations that remain static after deployment, limiting their adaptability to dynamic environments. To address this challenge, recent research has explored agent evolution techniques that aim to enhance agent systems automatically based on interaction data and environmental feedback. This survey focuses on self-evolving agentic systems, which bridge the static capabilities of foundation models with the continuous adaptability required by lifelong agentic systems. The survey introduces a unified conceptual framework that abstracts the feedback loop underlying the design of self-evolving agentic systems, highlighting four key components: System Inputs, Agent System, Environment, and Optimisers. The survey systematically reviews a wide range of self-evolving techniques targeting different components of the agent system. It also explores domain-specific evolution strategies developed for specialized fields such as biomedicine, programming, and finance. Additionally, the survey delves into memory optimization techniques for long-term memory storage to support coherent reasoning and decision-making over time. Furthermore, the survey discusses tool use as a key factor for AI agents to effectively operate in real-world scenarios by extending their functional scope through external tools like web search engines and code interpreters. Evaluation, safety considerations, and ethical implications for self-evolving agentic systems are also addressed to ensure their effectiveness and reliability. Overall, this comprehensive review aims to provide researchers and practitioners with a systematic understanding of self-evolving AI agents. By laying the foundation for more adaptive, autonomous, and lifelong agentic systems across various domains,this survey contributes to advancing the field of artificial intelligence.