Awesome-Self-Evolving-Agents

Why we included this project

This reading list collects recent work on AI agents that improve themselves over time instead of staying fixed after deployment. The entries are grouped into a taxonomy that separates single-agent optimization, multi-agent optimization, and domain-specific approaches, and each one links to its paper and, when it exists, its code. That makes it easy to follow a technique from the idea in a paper to an actual implementation, whether the method leans on reinforcement learning, test-time search, or feedback-driven refinement. The companion arXiv survey supplies the conceptual framework behind the taxonomy, so the list works as a quick index and as a map for deeper reading. It is a curated index rather than runnable software, but for anyone trying to get a sense of what lifelong agentic systems can do, it saves a lot of time digging through the literature.