AI4AI Survey: From Long-Horizon Agents to Recursive Self-Improvement — Definitions, Reliable Horizons, and Open Problems

Abstract

Can AI improve AI? AI agents can now run experiments, modify code and training pipelines, and iteratively improve AI artifacts. Yet the literature remains fragmented across long-horizon agents, AI for AI (AI4AI), self-improvement, and recursive self-improvement, making it difficult to tell how much progress has actually been made. This survey synthesizes evidence from 223 papers. We organize the AI4AI landscape around one question: how far can an AI system reliably carry an improvement process from idea to verified result? Across model design, agent harnesses, benchmarks, automated research, and self-modifying systems, we examine what current systems can do, how progress should be evaluated, and where claims of self-improvement remain unsupported. AI systems increasingly excel at the work of improvement — planning, coding, experimentation, optimization, and repair — but humans still largely determine the goals, evaluation criteria, and what counts as progress. Moreover, strong performance on individual components rarely translates into reliable end-to-end AI improvement. We call this the composition gap. By separating demonstrated capability from extrapolated autonomy, this survey maps what AI4AI can do today and what must change before AI can reliably improve AI itself. 🔗Project site: kaiwu5.github.io/Awesome-AI4AI 🛠️ RSIHub Harness: https://github.com/simple-agent-lab/RSIHub 📝 Blog: https://simpleagentlab.com/ai4ai