Elamur Research⌕
☰
home / papers / 2026 / Learning to Ideate for Machine Learning Engineering Agents

Learning to Ideate for Machine Learning Engineering Agents

Paper · 2026 · Yunxiang Zhang, Kang Zhou, Zhichao Xu et al.

Summary

Agents for machine learning engineering (MLE) often stop once they have one working solution and do not keep refining it. This paper presents MLE-IDEATOR, which splits the job between two agents. An Implementer, built on CodeAct, writes and runs the code. When progress stalls, it calls a new <seek_help> action to ask a separate IDEATOR for a structured suggestion: an analysis, a next action and a rationale.

Even without training, this setup beats implementation-only baselines on MLE-Bench. The authors then train the IDEATOR with GRPO. Each sampled idea is carried out once by a frozen Implementer, and the reward depends on whether the score improves. This avoids expensive multi-turn rollouts. Training used about 1K states from 10 Kaggle tasks. The resulting Qwen3-8B IDEATOR improves on its untrained version by 11.5% (relative) and outperforms Claude Sonnet 3.5 in the same role.

The work suggests a cheaper route to useful MLE agents. A small model can be trained to supply strategy that guides a stronger coding model, with no need to retrain the coder.

Key points

  • Reward design: +1 if performance improves, 0 if it does not improve or execution fails, −1 for format errors. Training states were collected offline from 90 MLE-IDEATOR runs per task with Claude Sonnet 3.5, giving 1,000 training and 100 validation states.
  • Evaluation: 51 held-out MLE-Bench tasks (21 low- and 30 medium-complexity). The 15 high-complexity tasks were excluded.
  • Ablations: The NULL IDEA and VAGUE IDEA templates add little or nothing. This indicates the gains come from specific, contextual ideas rather than from the extra interaction.
  • Effect of RL on idea types: RL moves the IDEATOR toward Feature Engineering and Data Preparation ideas, which were more often effective. It reduces less reliable Hyperparameter Tuning and Model Training suggestions.
  • Costs:
    • With Sonnet 3.5 in both roles, a run costs about 1.4 USD versus 0.9 USD for CodeAct alone.
    • Runs take 17.4 versus 15.0 steps on average.
    • Reward computation for RL ran on 128 A10G GPUs.

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

 Fields, topics and concepts

📚 Shelves 3The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksInformation Systems · 1,437 works

Cite

Cite as Yunxiang Zhang, Kang Zhou, Zhichao Xu, Kiran Ramnath, Yun Zhou, Sangmin Woo, Haibo Ding and Lin Lee Cheong. 2026. Learning to Ideate for Machine Learning Engineering Agents. Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers).

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.eacl-short.32

@inproceedings{zhang2026learning,
  title = {{Learning to Ideate for Machine Learning Engineering Agents}},
  author = {Yunxiang Zhang, Kang Zhou, Zhichao Xu, Kiran Ramnath, Yun Zhou, Sangmin Woo, Haibo Ding and Lin Lee Cheong},
  booktitle = {Proceedings of the 19th Conference of the European Chapter of the Association for Computational Linguistics (Volume 2: Short Papers)},
  year = {2026},
  doi = {10.18653/v1/2026.eacl-short.32},
  url = {https://doi.org/10.18653/v1/2026.eacl-short.32}
}

References

33 references, 14 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. David Anugraha, Genta Indra Winata, Chenyue Li, Patrick Amadeus Irawan, and En-Shiun Annie Lee. 2025. Proxylm: Predicting language model performance on multilingual tasks via proxy models. In Findings of the Association for Computational Linguistics: NAACL 2025, Albuquerque, New Mexico, USA, April 2… find
  2. Collin Burns, Pavel Izmailov, Jan Hendrik Kirchner, Bowen Baker, Leo Gao, Leopold Aschenbrenner, Yining Chen, Adrien Ecoffet, Manas Joglekar, Jan Leike, Ilya Sutskever, and Jeffrey Wu. 2024. Weak-to-strong generalization: Eliciting strong capabilities with weak supervision. In Forty-first Internatio…
  3. Jun Shern Chan, Neil Chowdhury, Oliver Jaffe, James Aung, Dane Sherburn, Evan Mays, Giulio Starace, Kevin Liu, Leon Maksin, Tejal Patwardhan, Aleksander Madry, and Lilian Weng. 2025. Mle-bench: Evaluating machine learning agents on machine learning engineering. In The Thirteenth International Confer…
  4. Haoyang Fang, Boran Han, Nick Erickson, Xiyuan Zhang, Su Zhou, Anirudh Dagar, Jiani Zhang, Ali Caner Türkmen, Cuixiong Hu, Huzefa Rangwala, Ying Nian Wu, Yuyang Wang, and George Karypis. 2025. Mlzero: A multi-agent system for end-to-end machine learning automation. CoRR, abs/2505.13941. — [arXiv:2505.13941](https://arxiv.org/abs/2505.13941) find
  5. Edward J. Hu, Yelong Shen, Phillip Wallis, Zeyuan Allen-Zhu, Yuanzhi Li, Shean Wang, Lu Wang, and Weizhu Chen. 2022. Lora: Low-rank adaptation of large language models. In The Tenth International Conference on Learning Representations, ICLR 2022, Virtual Event, April 25-29, 2022. OpenReview.net.
  6. Qian Huang, Jian Vora, Percy Liang, and Jure Leskovec. 2024. Mlagentbench: Evaluating language agents on machine learning experimentation. In Forty-first International Conference on Machine Learning, ICML 2024, Vienna, Austria, July 21-27, 2024. OpenReview.net.
  7. Minqi Jiang, Andrei Lupu, and Yoram Bachrach. 2025a. Bootstrapping task spaces for self-improvement. arXiv preprint arXiv:2509.04575. — [arXiv:2509.04575](https://arxiv.org/abs/2509.04575) find
  8. Zhengyao Jiang, Dominik Schmidt, Dhruv Srikanth, Dixing Xu, Ian Kaplan, Deniss Jacenko, and Yuxiang Wu. 2025b. Aide: Ai-driven exploration in the space of code. ArXiv, abs/2502.13138. find
  9. Carlos E. Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R. Narasimhan. 2024. Swe-bench: Can language models resolve real-world github issues? In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024. OpenRe…
  10. Zexi Liu, Yuzhu Cai, Xinyu Zhu, Yujie Zheng, Runkun Chen, Ying Wen, Yanfeng Wang, Weinan E, and Siheng Chen. 2025a. Ml-master: Towards ai-for-ai via integration of exploration and reasoning. CoRR, abs/2506.16499. — [arXiv:2506.16499](https://arxiv.org/abs/2506.16499) find
  11. Zexi Liu, Jingyi Chai, Xinyu Zhu, Shuo Tang, Rui Ye, Bolun Zhang, Lei Bai, and Siheng Chen. 2025b. Ml-agent: Reinforcing LLM agents for autonomous machine learning engineering. CoRR, abs/2505.23723. — [arXiv:2505.23723](https://arxiv.org/abs/2505.23723) find
  12. Chris Lu, Cong Lu, Robert Tjarko Lange, Jakob N. Foerster, Jeff Clune, and David Ha. 2024. The AI scientist: Towards fully automated open-ended scientific discovery. CoRR, abs/2408.06292.
  13. Lingrui Mei, Jiayu Yao, Yuyao Ge, Yiwei Wang, Baolong Bi, Yujun Cai, Jiazhi Liu, Mingyu Li, Zhong-Zhi Li, Duzhen Zhang, Chenlin Zhou, Jiayi Mao, Tianze Xia, Jiafeng Guo, and Shenghua Liu. 2025. A survey of context engineering for large language models. CoRR, abs/2507.13334. — [arXiv:2507.13334](https://arxiv.org/abs/2507.13334) find
  14. Jungsoo Park, Ethan Mendes, Gabriel Stanovsky, and Alan Ritter. 2025. Look before you leap: Estimating llm benchmark scores from descriptions. Preprint, arXiv:2509.20645. — [arXiv:2509.20645](https://arxiv.org/abs/2509.20645) find
  15. Rushi Qiang, Yuchen Zhuang, Yinghao Li, Dingu Sagar V. K, Rongzhi Zhang, Changhao Li, Ian Shu-Hei Wong, Sherry Yang, Percy Liang, Chao Zhang, and Bo Dai. 2025. Mle-dojo: Interactive environments for empowering LLM agents in machine learning engineering. CoRR, abs/2505.07782. find
  16. Guangming Sheng, Chi Zhang, Zilingfeng Ye, Xibin Wu, Wang Zhang, Ru Zhang, Yanghua Peng, Haibin Lin, and Chuan Wu. 2025. Hybridflow: A flexible and efficient RLHF framework. In Proceedings of the Twentieth European Conference on Computer Systems, EuroSys 2025, Rotterdam, The Netherlands, 30 March 20…
  17. Charlie Snell, Jaehoon Lee, Kelvin Xu, and Aviral Kumar. 2024. Scaling LLM test-time compute optimally can be more effective than scaling model parameters. CoRR, abs/2408.03314.
  18. Kyle Swanson, Wesley Wu, Nash L. Bulaong, John E. Pak, and James Zou. 2024. The virtual lab: Ai agents design new sars-cov-2 nanobodies with experimental validation. bioRxiv. find
  19. Edan Toledo, Karen Hambardzumyan, Martin Josifoski, Rishi Hazra, Nicolas Mario Baldwin, Alexis Audran-Reiss, Michael Kuchnik, Despoina Magka, Minqi Jiang, Alisia Maria Lupidi, Andrei Lupu, Roberta Raileanu, Kelvin Niu, Tatiana Shavrina, Jean-Christophe Gagnon-Audet, Michael Shvartsman, Shagun Sodhan… — [arXiv:2507.02554](https://arxiv.org/abs/2507.02554) find
  20. Huacan Wang, Ziyi Ni, Shuo Zhang, Shuo Lu, Sen Hu, Ziyang He, Chen Hu, Jiaye Lin, Yifu Guo, Yuntao Du, and Pin Lyu. 2025a. Repomaster: Autonomous exploration and understanding of github repositories for complex task solving. CoRR, abs/2505.21577. — [arXiv:2505.21577](https://arxiv.org/abs/2505.21577) find
  21. Xingyao Wang, Boxuan Li, Yufan Song, Frank F. Xu, Xiangru Tang, Mingchen Zhuge, Jiayi Pan, Yueqi Song, Bowen Li, Jaskirat Singh, Hoang H. Tran, Fuqiang Li, Ren Ma, Mingzhang Zheng, Bill Qian, Yanjun Shao, Niklas Muennighoff, Yizhe Zhang, Binyuan Hui, and 2 others. 2025b. Openhands: An open platform…
  22. Jiaxin Wen, Chenglei Si, Yueh-han Chen, He He, and Shi Feng. 2025. Predicting empirical AI research outcomes with language models. CoRR, abs/2506.00794. — [arXiv:2506.00794](https://arxiv.org/abs/2506.00794) find
  23. 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. CoRR, abs/2501.04227.
  24. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Mingchuan Zhang, Y. K. Li, Y. Wu, and Daya Guo. 2024. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. CoRR, abs/2402.03300.
  25. An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, Chujie Zheng, Dayiheng Liu, Fan Zhou, Fei Huang, Feng Hu, Hao Ge, Haoran Wei, Huan Lin, Jialong Tang, and 40 others. 2025a. Qwen3 technical report. CoRR, abs/2505.09388.
  26. Junhong Shen, Hao Bai, Lunjun Zhang, Yifei Zhou, Amrith Setlur, Shengbang Tong, Diego Caples, Nan Jiang, Tong Zhang, Ameet Talwalkar, and Aviral Kumar. 2025. Thinking vs. doing: Agents that reason by scaling test-time interaction. CoRR, abs/2506.07976. — [arXiv:2506.07976](https://arxiv.org/abs/2506.07976) find
  27. John Yang, Carlos E. Jimenez, Alexander Wettig, Kilian Lieret, Shunyu Yao, Karthik Narasimhan, and Ofir Press. 2024. Swe-agent: Agent-computer interfaces enable automated software engineering. In Advances in Neural Information Processing Systems 38: Annual Conference on Neural Information Processing…
  28. Sherry Yang, Joy He-Yueya, and Percy Liang. 2025b. Reinforcement learning for machine learning engineering agents. arXiv preprint arXiv:2509.01684. find
  29. Xu Yang, Xiao Yang, Shikai Fang, Bowen Xian, Yuante Li, Jian Wang, Minrui Xu, Haoran Pan, Xinpeng Hong, Weiqing Liu, Yelong Shen, Weizhu Chen, and Jiang Bian. 2025c. R&d-agent: Automating data-driven AI solution building through llm-powered automated research, development, and evolution. CoRR, abs/2… — [arXiv:2505.14738](https://arxiv.org/abs/2505.14738) find
  30. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, and 16 others. 2025. DAPO: an open-source LLM reinforcement learning sys…
  31. Bo Zhang, Shi Feng, Xiangchao Yan, Jiakang Yuan, Zhiyin Yu, Xiaohan He, Songtao Huang, Shaowei Hou, Zheng Nie, Zhilong Wang, Jinyao Liu, Runmin Ma, Tianshuo Peng, Peng Ye, Dongzhan Zhou, Shufei Zhang, Xiaosong Wang, Yilan Zhang, Meng Li, and 5 others. 2025a. Novelseek: When agent becomes the scienti… find
  32. Yunxiang Zhang, Muhammad Khalifa, Shitanshu Bhushan, Grant D. Murphy, Lajanugen Logeswaran, Jaekyeom Kim, Moontae Lee, Honglak Lee, and Lu Wang. 2025b. MLRC-Bench: Can language agents solve machine learning research challenges? CoRR, abs/2504.09702. find
  33. King Zhu, Hanhao Li, Siwei Wu, Tianshun Xing, Dehua Ma, Xiangru Tang, Minghao Liu, Jian Yang, Jiaheng Liu, Yuchen Eleanor Jiang, Changwang Zhang, Chenghua Lin, Jun Wang, Ge Zhang, and Wangchunshu Zhou. 2025. Scaling test-time compute for LLM agents. CoRR, abs/2506.12928. — [arXiv:2506.12928](https://arxiv.org/abs/2506.12928) find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/zhang-2026-learning-ideate-machine-learning-engineering/