This repository provides a comprehensive survey of self-evolving agents, focusing on their development towards artificial superintelligence. It serves as a curated collection of research papers, methodologies, and challenges in the field, targeting AI researchers and practitioners interested in advanced agent architectures.

How It Works

The survey categorizes self-evolving agents based on what aspects evolve (models, context, tools, architecture), when evolution occurs (intra-test-time, inter-test-time), how evolution is achieved (reward-based, imitation learning, evolutionary methods), and where evolution is applied (general vs. specialized domains). It details various techniques like prompt optimization, memory augmentation, and multi-agent system optimization.

Quick Start & Requirements

This repository is a survey and does not have a direct installation or execution command. It links to numerous research papers and GitHub repositories for specific agent implementations.

Highlighted Details

• Comprehensive categorization of self-evolving agent research.
• Extensive lists of relevant papers and projects for each evolutionary dimension.
• Discussion of evaluation metrics and paradigms for self-evolving agents.
• Exploration of future directions including personalization, generalization, safety, and multi-agent ecosystems.

Maintenance & Community

The project is actively maintained, with a call for contributions via pull requests. It cites a primary research paper with arXiv details.

Licensing & Compatibility

The repository itself does not specify a license. The linked research papers and code repositories will have their own respective licenses.

Limitations & Caveats

This is a survey and does not provide executable code for self-evolving agents. Users must refer to the cited external repositories for implementation details and functionality. The content is research-oriented and may represent early-stage or theoretical concepts.