This paper presents a comprehensive survey of self-evolution approaches in large language models (LLMs). The authors propose a conceptual framework for self-evolution, which consists of four phases: experience acquisition, experience refinement, updating, and evaluation. They categorize the evolution objectives of LLMs and LLM-based agents, providing a taxonomy and insights for each module. The survey also highlights existing challenges and suggests future directions to improve self-evolution frameworks. The paper is structured into several sections, covering the background, conceptual framework, evolution objectives, experience acquisition, experience refinement, updating, and evaluation. It discusses various methods for task evolution, solution evolution, feedback acquisition, experience refinement, and updating, including in-weight and in-context learning. The authors emphasize the importance of self-evolution in enabling LLMs to adapt, learn, and improve autonomously, similar to human evolution. The survey aims to deepen the understanding of self-evolving LLMs and guide future research in this field.This paper presents a comprehensive survey of self-evolution approaches in large language models (LLMs). The authors propose a conceptual framework for self-evolution, which consists of four phases: experience acquisition, experience refinement, updating, and evaluation. They categorize the evolution objectives of LLMs and LLM-based agents, providing a taxonomy and insights for each module. The survey also highlights existing challenges and suggests future directions to improve self-evolution frameworks. The paper is structured into several sections, covering the background, conceptual framework, evolution objectives, experience acquisition, experience refinement, updating, and evaluation. It discusses various methods for task evolution, solution evolution, feedback acquisition, experience refinement, and updating, including in-weight and in-context learning. The authors emphasize the importance of self-evolution in enabling LLMs to adapt, learn, and improve autonomously, similar to human evolution. The survey aims to deepen the understanding of self-evolving LLMs and guide future research in this field.