Elamur Research⌕
☰
home / papers / 2026 / AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse

AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse

Paper · 2026 · Zhang Zhang, Shuqi Lu, Hongjin Qian et al.

Summary

Most self-evolving LLM agents store past successes as text, such as reflections, prompts or reasoning traces. That gives no guarantee that a complex task can later be redone efficiently. AgentFactory instead keeps successful solutions as runnable Python subagents. A Meta-Agent breaks each request into subproblems and builds a dedicated subagent for each one, giving it only a selected subset of tools. Working subagents are saved with a SKILL.md description. When related tasks arrive, the system reuses stored subagents and revises them based on execution feedback, so they become more general and robust over time. Mature subagents can be exported to any Python-capable environment. The authors frame this as an Install → Self-Evolve → Deploy lifecycle, with isolated workspaces protecting the shared library. On a second batch of tasks, reusing saved subagents roughly halved the orchestrator's output tokens or better, compared with ReAct and a text-experience baseline. The work shows that agent capabilities can accumulate as portable, auditable code rather than as notes.

Key points

  • Skill levels. The skill system has three levels:

    • fixed meta skills (e.g., create_subagent, modify_subagent, run_subagent);
    • fixed tool skills: web_search (Serper), web_reading (Jina), browser_automation (Playwright) and shell_command;
    • evolving subagent skills, stored as Python scripts.

  • Evaluation setup. Two batches of 15 real-world tasks each were used, spanning information gathering from the web, chart and data plotting, automated browser interaction and audio handling. Batch 2 tasks resemble Batch 1 but have different requirements. Backbones were Claude Opus 4.6 and Claude Sonnet 4.6.

  • Batch 2 results (average orchestrator output tokens per task, with saved subagents):

    MethodOpusSonnet
    AgentFactory29713862
    Self-evolving text baseline62108223
    ReAct70227029

  • Batch 1 results. Even within Batch 1, Opus reused subagents across tasks and used 4324 tokens versus 8298 for ReAct. Sonnet did not benefit in Batch 1, using 9199 tokens versus 6893 for ReAct.

  • Caveat on the metric. Token counts exclude LLM usage inside subagents.

  • Availability. The code is open-sourced at github.com/zzatpku/AgentFactory. Saved subagents are human-readable for auditing, and shell commands pass through checks that flag destructive operations.

Our summary of the paper, in our own words — not its text. The work belongs to its authors; read it at the links above.

 Institutions

🎓Peking UniversityBeijing · China315 works · 43,268 citations🎓Hong Kong Polytechnic UniversityHong Kong · China106 works · 5,190 citations🔬Beijing Academy of Artificial IntelligenceBeijing · China46 works · 8,682 citations

 Fields, topics and concepts

📚 Shelves 2The field of science this work belongs to (OpenAlex fields).Computer Science · 18,614 worksArtificial Intelligence · 17,965 works

Cite

Cite as Zhang Zhang, Shuqi Lu, Hongjin Qian, Di He and Zheng Liu. 2026. AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.acl-demo.81

@inproceedings{zhang2026agentfactory,
  title = {{AgentFactory: A Self-Evolving Framework Through Executable Subagent Accumulation and Reuse}},
  author = {Zhang Zhang, Shuqi Lu, Hongjin Qian, Di He and Zheng Liu},
  booktitle = {Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations)},
  year = {2026},
  doi = {10.18653/v1/2026.acl-demo.81},
  url = {https://doi.org/10.18653/v1/2026.acl-demo.81}
}

References

31 references, 17 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Anthropic. 2026. Claude opus 4.6: Frontier intelligence for coding and agents. https://www.anthropic.com/news/claude-opus-4-6. Accessed: 2026-03-13. ↗
  2. Harrison Chase. 2022. Langchain. https://github.com/langchain-ai/langchain. evolutionary algorithms yields powerful prompt optimizers. In The Twelfth International Conference on Learning Representations. ↗
  3. Sirui Hong, Mingchen Zhuge, Jonathan Chen, Xiawu Zheng, Yuheng Cheng, Jinlin Wang, Ceyao Zhang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. 2024. MetaGPT: Meta programming for A multi-agent collaborative framework. In The T…
  4. Shengran Hu, Cong Lu, and Jeff Clune. 2024. Automated design of agentic systems. arXiv preprint arXiv:2408.08435.
  5. Zelong Li, Shuyuan Xu, Kai Mei, Wenyue Hua, Balaji Rama, Om Raheja, Hao Wang, He Zhu, and Yongfeng Zhang. 2024. AutoFlow: Automated workflow generation for large language model agents. arXiv preprint arXiv:2407.12821. — [arXiv:2407.12821](https://arxiv.org/abs/2407.12821) find
  6. Zijun Liu, Yanzhe Zhang, Peng Li, Yang Liu, and Diyi Yang. 2023. Dynamic LLM-Agent network: An LLM-agent collaboration framework with agent team optimization. arXiv preprint arXiv:2310.02170.
  7. Aman Madaan, Niket Tandon, Prakhar Gupta, Skyler Hallinan, Luyu Gao, Sarah Wiegreffe, Uri Alon, Nouha Dziri, Shrimai Prabhumoye, Yiming Yang, and 1 others. 2023. Self-refine: Iterative refinement with self-feedback. Advances in neural information processing systems, 36:46534–46594. find
  8. Alexander Novikov, Ngân Vũ, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov, Borislav Kozlovskii, Francisco JR Ruiz, Abbas Mehrabian, and 1 others. 2025. AlphaEvolve: A coding agent for scientific and algorithmic discovery. arXiv preprint arXiv:2506.13131.
  9. OpenAI. 2026. Gpt-5.4 thinking system card. Technical report, OpenAI Deployment Safety Hub. find
  10. Weize Chen, Yusheng Su, Jingwei Zuo, Cheng Yang, Chenfei Yuan, Chi-Min Chan, Heyang Yu, Yaxi Lu, Yi-Hsin Hung, Chen Qian, Yujia Qin, Xin Cong, Ruobing Xie, Zhiyuan Liu, Maosong Sun, and Jie Zhou. 2024. Agentverse: Facilitating multi-agent collaboration and exploring emergent behaviors. In The Twelft…
  11. Chrisantha Fernando, Dylan Banarse, Henryk Michalewski, Simon Osindero, and Tim Rocktäschel. 2024. Promptbreeder: Self-referential self-improvement via prompt evolution. In Forty-first International Conference on Machine Learning.
  12. Chen Qian, Wei Liu, Hongzhang Liu, Nuo Chen, Yufan Dang, Jiahao Li, Cheng Yang, Weize Chen, Yusheng Su, Xin Cong, Juyuan Xu, Dahai Li, Zhiyuan Liu, and Maosong Sun. 2024. ChatDev: Communicative agents for software development. In Proceedings of the 62nd Annual Meeting of the Association for Computat…
  13. Gemini team. 2026. Gemini 3.1 pro: A smarter model for your most complex tasks. find
  14. Noah Shinn, Federico Cassano, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. 2023. Reflexion: Language agents with verbal reinforcement learning. Advances in neural information processing systems, 36:8634–8652.
  15. Qingyan Guo, Rui Wang, Junliang Guo, Bei Li, Kaitao Song, Xu Tan, Guoqing Liu, Jiang Bian, and Yujiu Yang. 2024. EvoPrompt: Connecting llms with Significant Gravitas. 2023. Autogpt: Autonomous ai agents powered by large language models. https://github.com/Significant-Gravitas/AutoGPT. ↗
  16. Xiangru Tang, Tianrui Qin, Tianhao Peng, Ziyang Zhou, Daniel Shao, Tingting Du, Xinming Wei, Peng Xia, Fang Wu, He Zhu, and 1 others. 2025. Agent kb: Leveraging cross-domain experience for agentic problem solving. arXiv preprint arXiv:2507.06229. — [arXiv:2507.06229](https://arxiv.org/abs/2507.06229) find
  17. Guanzhi Wang, Yuqi Xie, Yunfan Jiang, Ajay Mandlekar, Chaowei Xiao, Yuke Zhu, Linxi Fan, and Anima Anandkumar. 2023. Voyager: An open-ended embodied agent with large language models. arXiv preprint arXiv:2305.16291. find
  18. Lei Wang, Chen Ma, Xueyang Feng, Zeyu Zhang, Hao Yang, Jingsen Zhang, Zhiyuan Chen, Jiakai Tang, Xu Chen, Yankai Lin, and 1 others. 2024a. A survey on large language model based autonomous agents. Frontiers of Computer Science, 18(6):186345.
  19. Xinyuan Wang, Chenxi Li, Zhen Wang, Fan Bai, Haotian Luo, Jiayou Zhang, Nebojsa Jojic, Eric P. Xing, and Zhiting Hu. 2024b. PromptAgent: Strategic planning with language models enables expert-level prompt optimization. In The Twelfth International Conference on Learning Representations.
  20. Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, and 1 others. 2024. Autogen: Enabling next-gen llm applications via multi-agent conversations. In First conference on language modeling. find
  21. Jenny Zhang, Shengran Hu, Cong Lu, Robert Lange, and Jeff Clune. 2025b. Darwin godel machine: Open-ended evolution of self-improving agents. arXiv preprint arXiv:2505.22954. find
  22. Jiayi Zhang, Jinyu Xiang, Zhaoyang Yu, Fengwei Teng, Xiong-Hui Chen, Jiaqi Chen, Mingchen Zhuge, Xin Cheng, Sirui Hong, Jinlin Wang, Bingnan Zheng, Bang Liu, Yuyu Luo, and Chenglin Wu. 2025c. AFlow: Automating agentic workflow generation. In The Thirteenth International Conference on Learning Repres…
  23. Zijian Zhou, Ao Qu, Zhaoxuan Wu, Sunghwan Kim, Alok Prakash, Daniela Rus, Jinhua Zhao, Bryan Kian Hsiang Low, and Paul Pu Liang. 2025. MEM1: Learning to synergize memory and reasoning for efficient long-horizon agents. arXiv preprint arXiv:2506.15841. — [arXiv:2506.15841](https://arxiv.org/abs/2506.15841) find
  24. Mingchen Zhuge, Wenyi Wang, Louis Kirsch, Francesco Faccio, Dmitrii Khizbullin, and Jürgen Schmidhuber. 2024. GPTSwarm: Language agents as optimizable graphs. In Forty-first International Conference on Machine Learning.
  25. Zhiheng Xi, Wenxiang Chen, Xin Guo, Wei He, Yiwen Ding, Boyang Hong, Ming Zhang, Junzhe Wang, Senjie Jin, Enyu Zhou, and 1 others. 2025. The rise and potential of large language model based agents: A survey. Science China Information Sciences, 68(2):121101.
  26. Wujiang Xu, Kai Mei, Hang Gao, Juntao Tan, Zujie Liang, and Yongfeng Zhang. 2025. A-MEM: Agentic memory for llm agents. arXiv preprint arXiv:2502.12110.
  27. Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V Le, Denny Zhou, and Xinyun Chen. 2023. Large language models as optimizers. In The Twelfth International Conference on Learning Representations.
  28. Shunyu Yao, Jeffrey Zhao, Dian Yu, Nan Du, Izhak Shafran, Karthik Narasimhan, and Yuan Cao. 2023. React: Synergizing reasoning and acting in language models. Preprint, arXiv:2210.03629.
  29. Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah Goodman. 2022. Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems, 35:15476–15488.
  30. Barry Zhang, Keith Lazuka, and Mahesh Murag. 2025a. Equipping agents for the real world with agent skills. https://www.anthropic.com/engineering/eq uipping-agents-for-the-real-world-with-a gent-skills. ↗
  31. Anthropic Engineering Blog, Oct 2025. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2026-agentfactory-self-evolving-framework-through/