Elamur Research⌕
☰
home / papers / 2026 / EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discovery

EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discovery

Paper · 2026 · Yougang Lyu, Xi Zhang, Xinhao Yi et al.

Summary

Most current LLM-based "AI scientist" systems follow fixed, hand-built pipelines. They do not learn from what happened in earlier runs, so they may retry experiments that already failed, miss promising directions, or chase ideas that cannot be implemented. EvoScientist addresses this with a multi-agent framework that improves itself over time. It has three agents. A Researcher Agent proposes ideas, an Engineer Agent writes and runs experiment code, and an Evolution Manager Agent turns past interaction histories into reusable knowledge. That knowledge lives in two persistent stores. An ideation memory keeps strong directions from top-ranked ideas and logs directions that failed validation. An experimentation memory keeps useful data-processing and training strategies taken from code-search trajectories and the best implementations. For each new task, the relevant entries are retrieved and added to the agents' prompts. Against 7 open-source and commercial baselines, EvoScientist produced better ideas under both automatic and human judgment, and its code ran successfully more often. This suggests that accumulated experience is a valuable resource for automated research.

Key points

  • The architecture has three agents (Researcher, Engineer, Evolution Manager) and three evolution mechanisms, which respectively refine the directions ideas pursue, the way ideas are validated, and the strategies used in experiments.
  • Idea quality was rated on novelty, feasibility, relevance and clarity, where the system beat 7 state-of-the-art systems. Code execution success rates also improved through evolution.
  • In end-to-end tests, all six generated full papers were accepted to the ICAIS 2025 AI Scientist Track. Two of them won awards: Best Paper and the AI Reviewer's Appraisal Award.
  • The ablations measured how often each reduced variant lost to the full system in pairwise comparisons:
    • Without direction evolution, it lost 66.67% of novelty comparisons and 50.00% of feasibility comparisons.
    • Without validation evolution, it lost 63.33% of feasibility comparisons.
    • Without any idea evolution, it lost 80.00% (novelty) and 83.33% (feasibility).
  • Limitations: the evaluation covers only computational tasks, not physical lab science. Reviewers also noted weak theoretical formalization in the generated papers.

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

🎓Vrije Universiteit AmsterdamAmsterdam · Netherlands42 works · 3,372 citations

Also on the byline: Huawei Technologies (United Kingdom).

 Fields, topics and concepts

📚 Shelves 4The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksDecision Sciences · 209 worksInformation Systems and Management · 41 works

Cite

Cite as Yougang Lyu, Xi Zhang, Xinhao Yi, Yuyue Zhao, Shuyu Guo, Wenxiang Hu, Jan Piotrowski, Jakub Kaliski, Jacopo Urbani, Zaiqiao Meng, Lun Zhou and Xiaohui Yan. 2026. EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discovery. arXiv:2603.08127.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.48550/arxiv.2603.08127 · arXiv 2603.08127

@misc{lyu2026evoscientist,
  title = {{EvoScientist: Towards Multi-Agent Evolving AI Scientists for End-to-End Scientific Discovery}},
  author = {Yougang Lyu, Xi Zhang, Xinhao Yi, Yuyue Zhao, Shuyu Guo, Wenxiang Hu, Jan Piotrowski, Jakub Kaliski, Jacopo Urbani, Zaiqiao Meng, Lun Zhou and Xiaohui Yan},
  year = {2026},
  doi = {10.48550/arxiv.2603.08127},
  eprint = {2603.08127},
  archivePrefix = {arXiv},
  url = {https://doi.org/10.48550/arxiv.2603.08127}
}

References

70 references, 15 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Zhongguancun Academy. 2025. The 1st International Conference on AI Scientists (ICAIS 2025). Website. https://icais.ai/ Tagline: Exploring the frontiers of automated scientific discovery with AI Scientists and autonomous research agents.. ↗
  2. Mohammad Almansoori, Komal Kumar, and Hisham Cholakkal. 2025. MedAgentSim: Self-evolving Multi-agent Simulations for Realistic Clinical Interactions. In International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer, 362–372. find
  3. Ibrahim Al Azher, Miftahul Jannat Mokarrama, Zhishuai Guo, Sagnik Ray Choudhury, and Hamed Alhoori. 2025. Futuregen: Llm-rag approach to generate the future work of scientific article. arXiv preprint arXiv:2503.16561 (2025). — [arXiv:2503.16561](https://arxiv.org/abs/2503.16561) find
  4. Qiguang Chen, Mingda Yang, Libo Qin, Jinhao Liu, Zheng Yan, Jiannan Guan, Dengyun Peng, Yiyan Ji, Hanjing Li, Mengkang Hu, et al. 2025. AI4Research: A Survey of Artificial Intelligence for Scientific Research. arXiv preprint arXiv:2507.01903 (2025).
  5. Prateek Chhikara, Dev Khant, Saket Aryan, Taranjeet Singh, and Deshraj Yadav. 2025. Mem0: Building production-ready ai agents with scalable long-term memory. arXiv preprint arXiv:2504.19413 (2025).
  6. Jinyuan Fang, Yanwen Peng, Xi Zhang, Yingxu Wang, Xinhao Yi, Guibin Zhang, Yi Xu, Bin Wu, Siwei Liu, Zihao Li, et al. 2025. A comprehensive survey of self-evolving ai agents: A new paradigm bridging foundation models and lifelong agentic systems. arXiv preprint arXiv:2508.07407 (2025). — [arXiv:2508.07407](https://arxiv.org/abs/2508.07407) find
  7. Huan-ang Gao, Jiayi Geng, Wenyue Hua, Mengkang Hu, Xinzhe Juan, Hongzhang Liu, Shilong Liu, Jiahao Qiu, Xuan Qi, Yiran Wu, et al. 2025. A survey of self-evolving agents: On path to artificial super intelligence. arXiv preprint arXiv:2507.21046 (2025). find
  8. Xian Gao, Zongyun Zhang, Mingye Xie, Ting Liu, and Yuzhuo Fu. 2025. Graph of AI Ideas: Leveraging Knowledge Graphs and LLMs for AI Research Idea Generation. arXiv preprint arXiv:2503.08549 (2025). — [arXiv:2503.08549](https://arxiv.org/abs/2503.08549) find
  9. Juraj Gottweis, Wei-Hung Weng, Alexander Daryin, Tao Tu, Anil Palepu, Petar Sirkovic, Artiom Myaskovsky, Felix Weissenberger, Keran Rong, Ryutaro Tanno, et al. 2025. Towards an AI co-scientist. arXiv preprint arXiv:2502.18864 (2025).
  10. Mourad Gridach, Jay Nanavati, Khaldoun Zine El Abidine, Lenon Mendes, and Christina Mack. 2025. Agentic AI for Scientific Discovery: A Survey of Progress, Challenges, and Future Directions. (2025). arXiv:2503.08979 [cs.CL] https://arxiv.org/abs/2503.08979 — [arXiv:2503.08979](https://arxiv.org/abs/2503.08979) ↗
  11. Jitai Hao, Qiang Huang, Hao Liu, Xinyan Xiao, Zhaochun Ren, and Jun Yu. 2025. A token is worth over 1,000 tokens: Efficient knowledge distillation through low-rank clone. arXiv preprint arXiv:2505.12781 (2025). find
  12. Jitai Hao, Hao Liu, Xinyan Xiao, Qiang Huang, and Jun Yu. 2025. Uni-X: Mitigating Modality Conflict with a Two-End-Separated Architecture for Unified Multimodal Models. arXiv preprint arXiv:2509.24365 (2025). — [arXiv:2509.24365](https://arxiv.org/abs/2509.24365) find
  13. hypogenic. 2026. hypogenic assistant. hypogenic Website (2026). https://hypogenic.ai/ ↗
  14. Intology. 2025. Zochi Technical Report. arXiv (2025). find
  15. k-dense. 2026. k-dense. k-dense. (2026). https://www.k-dense.ai/ ↗
  16. David Klahr. 2000. Exploring science: The cognition and development of discovery processes. MIT press. find
  17. David Klahr and Herbert A Simon. 1999. Studies of scientific discovery: Complementary approaches and convergent findings. Psychological Bulletin 125, 5 (1999), 524. find
  18. Miyoung Ko, Jinhyuk Lee, Hyunjae Kim, Gangwoo Kim, and Jaewoo Kang. 2020. Look at the First Sentence: Position Bias in Question Answering. In Proceedings of EMNLP. 1109–1121.
  19. Thomas S Kuhn and Ian Hacking. 1970. The structure of scientific revolutions. Vol. 2. University of Chicago press Chicago. find
  20. Pat Langley. 1987. Scientific discovery: Computational explorations of the creative processes. MIT press. find
  21. Sean Lee, Aamir Shakir, Darius Koenig, and Julius Lipp. 2024. Open Source Strikes Bread - New Fluffy Embeddings Model. https://www.mixedbread.ai/blog/mxbai-embed-large-v1 ↗
  22. Ruochen Li, Liqiang Jing, Chi Han, Jiawei Zhou, and Xinya Du. 2024. Learning to generate research idea with dynamic control. arXiv preprint arXiv:2412.14626 (2024). — [arXiv:2412.14626](https://arxiv.org/abs/2412.14626) find
  23. Ben Liu, Jihai Zhang, Fangquan Lin, Xu Jia, and Min Peng. 2025. One size doesn’t fit all: A personalized conversational tutoring agent for mathematics instruction. In Companion Proceedings of the ACM on Web Conference 2025. 2401–2410. find
  24. Yujie Liu, Zonglin Yang, Tong Xie, Jinjie Ni, Ben Gao, Yuqiang Li, Shixiang Tang, Wanli Ouyang, Erik Cambria, and Dongzhan Zhou. 2025. Researchbench: Benchmarking llms in scientific discovery via inspiration-based task decomposition. arXiv preprint arXiv:2503.21248 (2025). find
  25. Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob Foerster, Jeff Clune, and David Ha. 2024. The ai scientist: Towards fully automated open-ended scientific discovery. arXiv preprint arXiv:2408.06292 (2024).
  26. Yougang Lyu, Jitai Hao, Zihan Wang, Kai Zhao, Shen Gao, Pengjie Ren, Zhumin Chen, Fang Wang, and Zhaochun Ren. 2023. Multi-Defendant Legal Judgment Prediction via Hierarchical Reasoning. In Findings of EMNLP. 2198–2209. find
  27. Yougang Lyu, Piji Li, Yechang Yang, Maarten de Rijke, Pengjie Ren, Yukun Zhao, Dawei Yin, and Zhaochun Ren. 2023. Feature-Level Debiased Natural Language Understanding. In Proceedings of AAAI. 13353–13361. find
  28. Yougang Lyu, Shijie Ren, Yue Feng, Zihan Wang, Zhumin Chen, Zhaochun Ren, and Maarten de Rijke. 2025. Self-Adaptive Cognitive Debiasing for Large Language Models in Decision-Making. arXiv preprint arXiv:2504.04141 (2025). — [arXiv:2504.04141](https://arxiv.org/abs/2504.04141) find
  29. Yougang Lyu, Zihan Wang, Zhaochun Ren, Pengjie Ren, Zhumin Chen, Xiaozhong Liu, Yujun Li, Hongsong Li, and Hongye Song. 2022. Improving Legal Judgment Prediction through Reinforced Criminal Element Extraction. Inf. Process. Manag. 59, 1 (2022), 102780. find
  30. Yougang Lyu, Lingyong Yan, Shuaiqiang Wang, Haibo Shi, Dawei Yin, Pengjie Ren, Zhumin Chen, Maarten de Rijke, and Zhaochun Ren. 2024. KnowTuning: Knowledge-aware Fine-tuning for Large Language Models. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing. 14535– 1… find
  31. Yougang Lyu, Lingyong Yan, Zihan Wang, Dawei Yin, Pengjie Ren, Maarten de Rijke, and Zhaochun Ren. [n. d.]. MACPO: Weak-to-Strong Alignment via Multi-Agent Contrastive Preference Optimization. In The Thirteenth International Conference on Learning Representations. find
  32. Yougang Lyu, Xiaoyu Zhang, Zhaochun Ren, and Maarten de Rijke. 2024. Cognitive biases in large language models for news recommendation. arXiv preprint arXiv:2410.02897 (2024). — [arXiv:2410.02897](https://arxiv.org/abs/2410.02897) find
  33. Yougang Lyu, Xiaoyu Zhang, Lingyong Yan, Maarten de Rijke, Zhaochun Ren, and Xiuying Chen. 2025. Deepshop: A benchmark for deep research shopping agents. arXiv preprint arXiv:2506.02839 (2025). — [arXiv:2506.02839](https://arxiv.org/abs/2506.02839) find
  34. Novix. 2026. Novix. Novix Website (2026). https://novix.science/ ↗
  35. Ollama. 2026. Ollama. Website. https://ollama.com/ Accessed 2026-02-09. ↗
  36. Charles O’Neill, Tirthankar Ghosal, Roberta Răileanu, Mike Walmsley, Thang Bui, Kevin Schawinski, and Ioana Ciucă. 2025. Sparks of science: Hypothesis generation using structured paper data. arXiv preprint arXiv:2504.12976 (2025).
  37. John R Platt. 1964. Strong Inference: Certain systematic methods of scientific thinking may produce much more rapid progress than others. science 146, 3642 (1964), 347–353. find
  38. Karl Popper. 2005. The logic of scientific discovery. Routledge. find
  39. Biqing Qi, Kaiyan Zhang, Kai Tian, Haoxiang Li, Zhang-Ren Chen, Sihang Zeng, Ermo Hua, Hu Jinfang, and Bowen Zhou. 2024. Large language models as biomedical hypothesis generators: a comprehensive evaluation. arXiv preprint arXiv:2407.08940 (2024).
  40. Jiahao Qiu, Xuan Qi, Tongcheng Zhang, Xinzhe Juan, Jiacheng Guo, Yifu Lu, Yimin Wang, Zixin Yao, Qihan Ren, Xun Jiang, et al. 2025. Alita: Generalist agent enabling scalable agentic reasoning with minimal predefinition and maximal self-evolution. arXiv preprint arXiv:2505.20286 (2025). — [arXiv:2505.20286](https://arxiv.org/abs/2505.20286) find
  41. Yansheng Qiu, Haoquan Zhang, Zhaopan Xu, Ming Li, Diping Song, Zheng Wang, and Kaipeng Zhang. 2025. Ai idea bench 2025: Ai research idea generation benchmark. arXiv preprint arXiv:2504.14191 (2025). — [arXiv:2504.14191](https://arxiv.org/abs/2504.14191) find
  42. Changle Qu, Sunhao Dai, Xiaochi Wei, Hengyi Cai, Shuaiqiang Wang, Dawei Yin, Jun Xu, and Ji-Rong Wen. 2024. From exploration to mastery: Enabling llms to master tools via self-driven interactions. arXiv preprint arXiv:2410.08197 (2024). — [arXiv:2410.08197](https://arxiv.org/abs/2410.08197) find
  43. Chandan K Reddy and Parshin Shojaee. 2025. Towards scientific discovery with generative ai: Progress, opportunities, and challenges. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 39. 28601–28609. find
  44. Maxime Robeyns, Martin Szummer, and Laurence Aitchison. 2025. A self-improving coding agent. arXiv preprint arXiv:2504.15228 (2025). — [arXiv:2504.15228](https://arxiv.org/abs/2504.15228) find
  45. Aishik Sanyal, Samuel Schapiro, Sumuk Shashidhar, Royce Moon, Lav R Varshney, and Dilek Hakkani-Tur. 2025. Spark: A system for scientifically creative idea generation. arXiv preprint arXiv:2504.20090 (2025). find
  46. Samuel Schmidgall and Michael Moor. 2025. Agentrxiv: Towards collaborative autonomous research. arXiv preprint arXiv:2503.18102 (2025). find
  47. Samuel Schmidgall, Yusheng Su, Ze Wang, Ximeng Sun, Jialian Wu, Xiaodong Yu, Jiang Liu, Zicheng Liu, and Emad Barsoum. 2025. Agent laboratory: Using llm agents as research assistants. arXiv preprint arXiv:2501.04227 (2025).
  48. Chenyang Shao, Dehao Huang, Yu Li, Keyu Zhao, Weiquan Lin, Yining Zhang, Qingbin Zeng, Zhiyu Chen, Tianxing Li, Yifei Huang, et al. 2025. OmniScientist: Toward a Co-evolving Ecosystem of Human and AI Scientists. arXiv preprint arXiv:2511.16931 (2025). — [arXiv:2511.16931](https://arxiv.org/abs/2511.16931) find
  49. Chenyang Shao, Dehao Huang, Yu Li, Keyu Zhao, Weiquan Lin, Yining Zhang, Qingbin Zeng, Zhiyu Chen, Tianxing Li, Yifei Huang, et al. 2025. OmniScientist: Toward a Co-evolving Ecosystem of Human and AI Scientists. arXiv preprint arXiv:2511.16931 (2025). — [arXiv:2511.16931](https://arxiv.org/abs/2511.16931) find
  50. Zhengliang Shi, Yiqun Chen, Haitao Li, Weiwei Sun, Shiyu Ni, Yougang Lyu, Run-Ze Fan, Bowen Jin, Yixuan Weng, Minjun Zhu, et al. 2025. Deep Research: A Systematic Survey. arXiv preprint arXiv:2512.02038 (2025). — [arXiv:2512.02038](https://arxiv.org/abs/2512.02038) find
  51. 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 (2023), 8634–8652.
  52. Chenglei Si, Diyi Yang, and Tatsunori Hashimoto. 2024. Can llms generate novel research ideas? a large-scale human study with 100+ nlp researchers. arXiv preprint arXiv:2409.04109 (2024).
  53. Giulio Starace, Oliver Jaffe, Dane Sherburn, James Aung, Jun Shern Chan, Leon Maksin, Rachel Dias, Evan Mays, Benjamin Kinsella, Wyatt Thompson, et al. 2025. PaperBench: Evaluating AI’s Ability to Replicate AI Research. arXiv preprint arXiv:2504.01848 (2025). find
  54. Haoyang Su, Renqi Chen, Shixiang Tang, Zhenfei Yin, Xinzhe Zheng, Jinzhe Li, Biqing Qi, Qi Wu, Hui Li, Wanli Ouyang, et al. 2025. Many heads are better than one: Improved scientific idea generation by a llm-based multi-agent system. In Proceedings of the 63rd Annual Meeting of the Association for Co…
  55. Jiabin Tang, Lianghao Xia, Zhonghang Li, and Chao Huang. 2025. AI-Researcher: Autonomous Scientific Innovation. arXiv preprint arXiv:2505.18705 (2025). find
  56. NovelSeek Team, Bo Zhang, Shiyang Feng, Xiangchao Yan, Jiakang Yuan, Zhiyin Yu, Xiaohan He, Songtao Huang, Shaowei Hou, Zheng Nie, et al. 2025. Novel-Seek: When Agent Becomes the Scientist–Building Closed-Loop System from Hypothesis to Verification. arXiv preprint arXiv:2505.16938 (2025). find
  57. 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 (2023). find
  58. Peiyi Wang, Lei Li, Liang Chen, Dawei Zhu, Binghuai Lin, Yunbo Cao, Qi Liu, Tianyu Liu, and Zhifang Sui. 2023. Large Language Models are not Fair Evaluators. CoRR abs/2305.17926 (2023). find
  59. Yanlin Wang, Yanli Wang, Daya Guo, Jiachi Chen, Ruikai Zhang, Yuchi Ma, and Zibin Zheng. 2024. Rlcoder: Reinforcement learning for repository-level code completion. arXiv preprint arXiv:2407.19487 (2024). — [arXiv:2407.19487](https://arxiv.org/abs/2407.19487) find
  60. Zihan Wang, Ziqi Zhao, Yougang Lyu, Zhumin Chen, Maarten de Rijke, and Zhaochun Ren. 2025. A cooperative multi-agent framework for zero-shot named entity recognition. In Proceedings of the ACM on Web Conference 2025. 4183–4195. find
  61. Zora Zhiruo Wang, Jiayuan Mao, Daniel Fried, and Graham Neubig. 2024. Agent workflow memory. arXiv preprint arXiv:2409.07429 (2024).
  62. Yixuan Weng, Minjun Zhu, Qiujie Xie, Qiyao Sun, Zhen Lin, Sifan Liu, and Yue Zhang. 2025. Deepscientist: Advancing frontier-pushing scientific findings progressively. arXiv preprint arXiv:2509.26603 (2025). find
  63. Yixuan Weng, Minjun Zhu, Qiujie Xie, Qiyao Sun, Zhen Lin, Sifan Liu, and Yue Zhang. 2025. Deepscientist: Advancing frontier-pushing scientific findings progressively. arXiv preprint arXiv:2509.26603 (2025). find
  64. Tianze Xu, Pengrui Lu, Lyumanshan Ye, Xiangkun Hu, and Pengfei Liu. 2025. Researcherbench: Evaluating deep ai research systems on the frontiers of scientific inquiry. arXiv preprint arXiv:2507.16280 (2025). — [arXiv:2507.16280](https://arxiv.org/abs/2507.16280) find
  65. Yutaro Yamada, Robert Tjarko Lange, Cong Lu, Shengran Hu, Chris Lu, Jakob Foerster, Jeff Clune, and David Ha. 2025. The ai scientist-v2: Workshop-level automated scientific discovery via agentic tree search. arXiv preprint arXiv:2504.08066 (2025).
  66. Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah Goodman. 2022. Star: Bootstrapping reasoning with reasoning. Advances in Neural Information Processing Systems 35 (2022), 15476–15488.
  67. Jenny Zhang, Shengran Hu, Cong Lu, Robert Lange, and Jeff Clune. 2025. Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents. arXiv preprint arXiv:2505.22954 (2025). find
  68. Xiaoyu Zhang, Ruobing Xie, Yougang Lyu, Xin Xin, Pengjie Ren, Mingfei Liang, Bo Zhang, Zhanhui Kang, Maarten de Rijke, and Zhaochun Ren. 2024. Towards empathetic conversational recommender systems. In Proceedings of the 18th ACM Conference on Recommender Systems. 84–93. find
  69. Andrew Zhao, Daniel Huang, Quentin Xu, Matthieu Lin, Yong-Jin Liu, and Gao Huang. 2024. Expel: Llm agents are experiential learners. In Proceedings of the AAAI Conference on Artificial Intelligence, Vol. 38. 19632–19642. find
  70. Lianmin Zheng, Wei-Lin Chiang, Ying Sheng, Siyuan Zhuang, Zhanghao Wu, Yonghao Zhuang, Zi Lin, Zhuohan Li, Dacheng Li, Eric P. Xing, Hao Zhang, Joseph E. Gonzalez, and Ion Stoica. 2024. Judging LLM-as-a-judge with MT-Bench and Chatbot Arena. Proceedings of NeurIPS 36 (2024).

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/lyu-2026-evoscientist-towards-multi-agent-evolving/