Awesome-Self-Evolving-Agents: Explore Research on Evolving Agents

Overview

Awesome-Self-Evolving-Agents is a research-oriented catalog of papers, benchmarks, open-source libraries, and applications related to self-evolving agents. It addresses the difficulty of navigating a fast-growing field by grouping resources around how agents improve their models, their interaction environments, or both together.

The collection accompanies the survey A Systematic Survey of Self-Evolving Agents: From Model-Centric to Environment-Driven Co-Evolution. It is useful as a literature map and discovery tool, not as an agent implementation or ready-to-run software package.

Key Features

• Organizes research into model-centric, environment-centric, and model-environment co-evolution.
• Breaks down research papers into themes such as self-correction, online learning, memory, skills, tools, and agent architecture.
• Catalogs benchmarks across self-evolution, general knowledge, scientific and mathematical reasoning, coding, and agentic tasks.
• Lists agent libraries and infrastructure for orchestration, training, fine-tuning, and inference.
• Surveys application examples in scientific discovery, software engineering, and open-world simulation.
• Links entries to papers, code repositories, datasets, and project pages where listed.
• Invites community contributions through issues and pull requests.

Use Cases

• Researchers entering the field can use the taxonomy to identify relevant subtopics and related work.
• Students preparing a literature review can locate papers and benchmarks across multiple agent-evolution approaches.
• Agent developers can scan listed libraries and applications for related tools or design ideas before choosing an implementation.
• Benchmark designers can compare evaluation resources for agent capabilities and self-improvement.

Getting Started

No installation is required. Browse the repository README to explore the categorized resources and survey links.

Alternatives

• agentskills: Covers papers, tools, and learning resources across agent skills, MCP, tool use, and computer use, rather than focusing on self-evolving agents.
• awesome-ai-agents-2026: Surveys agents, frameworks, protocols, and models broadly, instead of centering research on how agents evolve.

Considerations

• This is a curated bibliography and resource list, not a software framework or executable agent.
• The breadth of the catalog means individual entries may vary in maturity, availability, and level of detail. Check linked papers and projects for current status.
• The input does not specify a license, programming language, or installation instructions.