mlsanityData to 7 Oct 2026

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

arXiv 2609.11873 ↗First posted 10 Sep 2026cs.LG cs.AI cs.CLcited by 5

On the map: Synthetic data and self-improvement for LLMs → Self-evolving LLM agents

Abstract (arXiv)

Recursive self-improvement (RSI) enables AI systems to turn experience and feedback into persistent changes that improve both their capabilities and the process of future improvement. We first use the Headroom-Closed Index (HCI) to reveal the problems of existing LLMs, then introduce the RSI concept and its development roadmap: from improvement-execution autonomy, improvement-strategy autonomy, experience-acquisition autonomy, and environment-adaptation autonomy, to recursive meta-improvement. Next we examine RSI across scenarios (e.g., scientific discovery, embodied intelligence, software engineering), highlighting their distinct requirements and development speeds. Drawing on diverse industry practices and preliminary empirical evidence, we connect RSI research with practical systems and identify key challenges to achieving genuine RSI.

Key facts. Cited by 5 papers in the map (data through 7 Oct 2026). Placed in Self-evolving LLM agents, one of 13 directions in Synthetic data and self-improvement for LLMs. Builds on 130 of its 217 references that are in the map. First posted on arXiv 10 Sep 2026.

Builds on

130 of 217 references.

Closest ancestors · 27

Same direction, newest first

  1. 2026Aspire: Can Models Self-Evolve from Vague Goals?31 Aug
  2. 2026PRACTICE: From Experience to Expertise in Self-Evolving Embodied Agents31 Aug
  3. 2026Evo-Harness: Context-to-Harness Skill Compilation for Self-Evolving Agents15 Aug
  4. 2026Self-Evolving Embodied Agents via Skill-Harness Evolution11 Aug
  5. 2026Self-Evolving Coding Agents4 Aug
  6. 2026Rethinking Self-Evolving Agent Skills: Feedback Dynamics over Multiple Rounds31 Jul
  7. 2026Self-Improvements in Modern Agentic Systems: A Survey14 Jul
  8. 2026The Red Queen G"odel Machine: Co-Evolving Agents and Their Evaluators24 Jun
  9. 2026SEAGym: An Evaluation Environment for Self-Evolving LLM Agents16 Jun
  10. 2026Self-Evolving Agents as Dynamic Graph Transformation: A Survey and New Perspective10 Jun
  11. 2026Socratic-SWE: Self-Evolving Coding Agents via Trace-Derived Agent Skills5 Jun
  12. 2026Harness Updating Is Not Harness Benefit: Disentangling Evolution Capabilities in Self-Evolving LLM Agents28 May
15 more
  1. 2026CODESKILL: Learning Self-Evolving Skills for Coding Agents25 May
  2. 2026APEX: Autonomous Policy Exploration for Self-Evolving LLM Agents20 May
  3. 2026Library Drift: Diagnosing and Fixing a Silent Failure Mode in Self-Evolving LLM Skill Libraries19 May
  4. 2026DrugSAGE:Self-evolving Agent Experience for Efficient State-of-the-Art Drug Discovery14 May
  5. 2026EmbodiSkill: Skill-Aware Reflection for Self-Evolving Embodied Agents11 May
  6. 2026SEA-Eval: A Benchmark for Evaluating Self-Evolving Agents Beyond Episodic Assessment10 Apr
  7. 2026PsychAgent: An Experience-Driven Lifelong Learning Agent for Self-Evolving Psychological Counselor1 Apr
  8. 2026Group-Evolving Agents: Open-Ended Self-Improvement via Experience Sharing4 Feb
  9. 2026S1-NexusAgent: a Self-Evolving Agent Framework for Multidisciplinary Scientific Research2 Feb
  10. 2026A Survey of Agent Memory in the Second Half: Towards Self-Evolving and Long-Horizon Agents14 Jan
  11. 2025Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory25 Nov
  12. 2025Self-Improving Embodied Foundation Models18 Sep
  13. 2025A Comprehensive Survey of Self-Evolving AI Agents: A New Paradigm Bridging Foundation Models and Lifelong Agentic Systems10 Aug
  14. 2025A Self-Improving Coding Agent21 Apr
  15. 2024Self-Evolving Multi-Agent Collaboration Networks for Software Development22 Oct

Other references · 103

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2023Reflexion: Language Agents with Verbal Reinforcement Learningcited by 4,064
  3. 2023Voyager: An Open-Ended Embodied Agent with Large Language Modelscited by 2,081
  4. 2023Is Your Code Generated by ChatGPT Really Correct? Rigorous Evaluation of Large Language Models for Code Generationcited by 1,332
  5. 2024τ-bench: A Benchmark for Tool-Agent-User Interaction in Real-World Domainscited by 1,280
  6. 2025DeepSeek-V3.2: Pushing the Frontier of Open Large Language Modelscited by 755
  7. 2024Yi: Open Foundation Models by 01.AIcited by 706
  8. 2024Phi-4 Technical Reportcited by 675
  9. 2025BrowseComp: A Simple Yet Challenging Benchmark for Browsing Agentscited by 616
  10. 2026Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfacescited by 515
  11. 2025Humanity's Last Examcited by 467
  12. 2025HealthBench: Evaluating Large Language Models Towards Improved Human Healthcited by 385
38 more
  1. 2025The AI Scientist-v2: Workshop-Level Automated Scientific Discovery via Agentic Tree Searchcited by 375
  2. 2024Agent Hospital: A Simulacrum of Hospital with Evolvable Medical Agentscited by 210
  3. 2019Paired Open-Ended Trailblazer (POET): Endlessly Generating Increasingly Complex and Diverse Learning Environments and Their Solutionscited by 182
  4. 2025AIMO-2 Winning Solution: Building State-of-the-Art Mathematical Reasoning Models with OpenMathReasoning datasetcited by 165
  5. 2026Trace2Skill: Distill Trajectory-Local Lessons into Transferable Agent Skillscited by 161
  6. 2025Phi-4-reasoning Technical Reportcited by 146
  7. 2024Continual Learning for Large Language Models: A Surveycited by 144
  8. 2024Nemotron-4 340B Technical Reportcited by 143
  9. 2023Self-Taught Optimizer (STOP): Recursively Self-Improving Code Generationcited by 128
  10. 2026Agentic Harness Engineering: Observability-Driven Automatic Evolution of Coding-Agent Harnessescited by 119
  11. 2026Learning to Discover at Test Timecited by 94
  12. 2024A Survey on Self-Evolution of Large Language Modelscited by 86
  13. 2026Self-Harness: Harnesses That Improve Themselvescited by 79
  14. 2025Lifelong Learning of Large Language Model based Agents: A Roadmapcited by 78
  15. 2025CyberGym: Evaluating AI Agents' Real-World Cybersecurity Capabilities at Scalecited by 71
  16. 2026Externalization in LLM Agents: A Unified Review of Memory, Skills, Protocols and Harness Engineeringcited by 68
  17. 2024OMNI-EPIC: Open-endedness via Models of human Notions of Interestingness with Environments Programmed in Codecited by 67
  18. 2025SWE-Exp: Experience-Driven Software Issue Resolutioncited by 64
  19. 2026ASPIRE: Agentic /Skills Discovery for Roboticscited by 44
  20. 2024EnvGen: Generating and Adapting Environments via LLMs for Training Embodied Agentscited by 41
  21. 2026EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discoverycited by 41
  22. 2026DeepSWE: Measuring Frontier Coding Agents on Original, Long-Horizon Engineering Taskscited by 41
  23. 2026Hyperagentscited by 40
  24. 2026VLAW: Iterative Co-Improvement of Vision-Language-Action Policy and World Modelcited by 39
  25. 2026CORAL: Towards Autonomous Multi-Agent Evolution for Open-Ended Discoverycited by 38
  26. 2026SKILLFOUNDRY: Building Self-Evolving Agent Skill Libraries from Heterogeneous Scientific Resourcescited by 38
  27. 2026World-VLA-Loop: Closed-Loop Learning of Video World Model and VLA Policycited by 36
  28. 2025GenEnv: Difficulty-Aligned Co-Evolution Between LLM Agents and Environment Simulatorscited by 34
  29. 2025CASCADE: Cumulative Agentic Skill Creation through Autonomous Development and Evolutioncited by 33
  30. 2026ENPIRE: Agentic Robot Policy Self-Improvement in the Real Worldcited by 31
  31. 2023Self-Improving Robots: End-to-End Autonomous Visuomotor Reinforcement Learningcited by 29
  32. 2024Lifelong Robot Library Learning: Bootstrapping Composable and Generalizable Skills for Embodied Control with Language Modelscited by 26
  33. 2025SIMA 2: A Generalist Embodied Agent for Virtual Worldscited by 26
  34. 2026From Static Templates to Dynamic Runtime Graphs: A Survey of Workflow Optimization for LLM Agentscited by 22
  35. 2026Workspace-Bench 1.0: Benchmarking AI Agents on Workspace Tasks with Large-Scale File Dependenciescited by 20
  36. 2025Self-Correction Bench: Uncovering and Addressing the Self-Correction Blind Spot in Large Language Modelscited by 19
  37. 2026SIA: Self Improving AI with Harness & Weight Updatescited by 18
  38. 2025A co-evolving agentic AI system for medical imaging analysiscited by 16

50 most cited of 103 All 103 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 5.

5 papers.

  1. 2026PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Pluginscited by 0
  2. 2026Reliability Engineering for AI Systems: Challenges, Methods, and Directionscited by 0
  3. 2026MERID: Multimodal Exploration via Recursive Self-Improvement Agents for Major Depression Analysiscited by 0
  4. 2026Verification and Self-Improvement in Agentic AI: Foundations and Limitscited by 0
  5. 2026Agentic-TTT: Training test-time policy for test-time trainingcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026Recursive Self-Improvement in AI: From Bounded Self-Refinement to Autonomous Research Loops
  2. 2026RSI-Forge: From Research Papers to Environments for Recursive Self-Improvement
  3. 2026RSIGym: A Flexible Environment for Recursive Self-Improvement
  4. 2026RSIAgent: Autonomous Exploration for Recursive Self-improvement in New Environments
  5. 2026AI4AI-Bench: Benchmarking LLM Agents in Algorithmic Design for Recursive Self-Improvement
  6. 2026NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness
  7. 2026Dream-RSI: Recursive Self-Improvement through Evolving Worlds
  8. 2026Frontis-MA1: Training an AI4AI Model towards Recursive Self-Improvement in Machine Learning Engineering