A Survey on Modern Self-Improving AI Agents
Zhe Ren · hf · 2026-07-16
This survey discusses the evolution of modern self-improving agents: transitioning from research prototypes to deployable systems, with the goal of enabling systems to continuously improve through experience with minimal human intervention.
The article proposes a system-level framework that treats an agent as a composite of foundation models, prompts, memory, tools, and control logic. It formalizes "self-improvement" as a self-induced update operation capable of modifying either model parameters or scaffold components.
The review further categorizes existing work along two main dimensions:
- Update Target: Which components are modified
- Driving Signal: What experiences/feedback drive the modification
Finally, it reviews application scenarios, evaluation methods, and summarizes open questions for the future.
Related event: Survey Maps the Rise of Self-Improving Agents(2 posts)→
More from coding & agent
- Dev turns DJI Mavic into local AI security drone built with Opus 5.5 — alexcovo_eth · 2026-10-27
- Microsoft Research: ForkPilot cuts long-horizon agent compute by up to 59% — EliasEskin · 2026-10-27
- Cursor launches Agent SDK: agents as code with Markdown prompts and TypeScript tools — soleio · 2026-10-27
- Vibecheck, the 'usable formal methods' hackathon, sells out two floors in SF — spikedoanz · 2026-10-27
- How do you verify AI agent outputs beyond unit tests and LLM-as-judge? — Practical_Green5775 · 2026-10-27
- AI agents may soon become the only way to run code, not just write it — IgorCarron · 2026-10-27