Elamur Research⌕
☰
home / papers / 2026 / Crossroads of Optimization under Uncertainty: How to Choose the Optimal Model

Crossroads of Optimization under Uncertainty: How to Choose the Optimal Model

Paper · 2026 · Chengxi She, Zhiqiang Chen, Xingyu Lu et al.

Summary

Optimization under Uncertainty (OuU) is hard to use without expertise in optimization and statistics. Practitioners also lack guidance on whether SAA, Robust Optimization (RO) or Distributionally Robust Optimization (DRO) suits a given problem. The authors present LLM4OuU, an LLM-driven multi-agent system. Separate agents handle initial formulation, model transformation, data loading, coding and repair, guided by domain-expert prompts and a retry loop. Six specialized sub-agents reformulate problems into specific uncertainty models.

To test it, they built a benchmark from two sources:

  • a RAG knowledge base of more than 2,000 arXiv OR papers;
  • numerical data from Kaggle and industry partners.

The system reaches up to 99% accuracy on some model types, and removing the expert knowledge from prompts clearly hurts results.

They then ran all six models on each problem and recorded which performed best. These feature-to-model pairs show that dataset size is the strongest predictor of the best model, while industry and scenario also matter. This offers practical, data-backed advice for choosing an OuU model.

Key points

  • The six sub-agents cover:
    • SAA;
    • RO with box, budget and ellipsoidal uncertainty sets;
    • DRO with KL-divergence and Wasserstein ambiguity sets.
  • The benchmark contains 100 data samples. Problem descriptions are generated via RAG and split into five standardized modeling elements. Experiments use DeepSeek-R1, among other models.
  • Ablation: dropping modeling-related expertise from prompts reduces the completion rate by 7–32% and accuracy by 6–33%.
  • Chi-square tests rank Data Scale first (χ² = 104.85, p < 0.001); Industry and Scenario are also significant. A multi-class logistic regression predicts the best model with 76.67% test accuracy (5-fold CV: 77.14% ± 7.0%).
  • Model preference shifts with data size:
    • small data favors DRO (odds ratio 3.87);
    • medium data favors RO (3.38);
    • large data favors SAA (4.19).
  • Code and data are released at github.com/ChengxiSHE/LLM4OuU.

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

🎓Nanjing UniversityNanjing · China74 works · 7,814 citations🏢Ant GroupHangzhou · China42 works · 1,131 citations

 Fields, topics and concepts

📚 Shelves 8The field of science this work belongs to (OpenAlex fields).Computer Science · 16,667 worksArtificial Intelligence · 16,194 worksArts and Humanities · 16,039 worksLanguage and Linguistics · 15,891 worksMathematics · 177 worksDecision Sciences · 209 worksManagement Science and Operations Research · 108 worksNumerical Analysis · 3 works

Cite

Cite as Chengxi She, Zhiqiang Chen, Xingyu Lu, Caihua Chen, Piaoyang Zhao and Xuedong Wang. 2026. Crossroads of Optimization under Uncertainty: How to Choose the Optimal Model. Findings of the Association for Computational Linguistics: ACL 2026.

The paper itself, not this page. BibTeX from the library's record of the work. DOI 10.18653/v1/2026.findings-acl.172

@inproceedings{she2026crossroads,
  title = {{Crossroads of Optimization under Uncertainty: How to Choose the Optimal Model}},
  author = {Chengxi She, Zhiqiang Chen, Xingyu Lu, Caihua Chen, Piaoyang Zhao and Xuedong Wang},
  booktitle = {Findings of the Association for Computational Linguistics: ACL 2026},
  year = {2026},
  doi = {10.18653/v1/2026.findings-acl.172},
  url = {https://doi.org/10.18653/v1/2026.findings-acl.172}
}

References

25 references, 4 of them with a card of ours. ↗ leads to the work, «find» searches for one we have no id for yet.

  1. Ali AhmadiTeshnizi, Wenzhi Gao, and Madeleine Udell. 2023. Optimus: Optimization modeling using mip solvers and large language models. arXiv preprint arXiv:2310.06116. — [arXiv:2310.06116](https://arxiv.org/abs/2310.06116) find
  2. Ali AhmadiTeshnizi, Wenzhi Gao, and Madeleine Udell. 2024. Optimus: Scalable optimization modeling with (mi) lp solvers and large language models. arXiv preprint arXiv:2402.10172. — [arXiv:2402.10172](https://arxiv.org/abs/2402.10172) find
  3. Cynthia Barnhart, Peter Belobaba, and Amedeo R Odoni. 2003. Applications of operations research in the air transport industry. Transportation science, 37(4):368–391. find
  4. Dimitris Bertsimas, Margrét V Bjarnadóttir, Michael A Kane, J Christian Kryder, Rudra Pandey, Santosh Vempala, and Grant Wang. 2008. Algorithmic prediction of health-care costs. Operations Research, 56(6):1382–1392. find
  5. Dimitris Bertsimas, David B Brown, and Constantine Caramanis. 2011. Theory and applications of robust optimization. SIAM review, 53(3):464–501. find
  6. Frank J Fabozzi, Dashan Huang, and Guofu Zhou. 2010. Robust portfolios: contributions from operations research and finance. Annals of operations research, 176(1):191–220. find
  7. Chenyu Huang, Zhengyang Tang, Dongdong Ge, Shixi Hu, Ruoqing Jiang, Benyou Wang, Zizhuo Wang, and Xin Zheng. 2024. Orlm: A customizable framework in training large models for automated optimization modeling. arXiv e-prints, pages arXiv–2405. find
  8. Caigao Jiang, Xiang Shu, Hong Qian, Xingyu Lu, Jun Zhou, Aimin Zhou, and Yang Yu. 2024. Llmopt: Learning to define and solve general optimization problems from scratch. arXiv preprint arXiv:2410.13213. — [arXiv:2410.13213](https://arxiv.org/abs/2410.13213) find
  9. Andrew J Keith and Darryl K Ahner. 2021. A survey of decision making and optimization under uncertainty. Annals of Operations Research, 300(2):319–353. find
  10. Anton J Kleywegt, Alexander Shapiro, and Tito Homem-de Mello. 2002. The sample average approximation method for stochastic discrete optimization. SIAM Journal on optimization, 12(2):479–502. find
  11. Douglas M Lambert and Martha C Cooper. 2000. Issues in supply chain management. Industrial marketing management, 29(1):65–83. find
  12. Peyman Mohajerin Esfahani and Daniel Kuhn. 2018. Data-driven distributionally robust optimization using the wasserstein metric: Performance guarantees and tractable reformulations. Mathematical Programming, 171(1):115–166. find
  13. Mahdi Mostajabdaveh, Timothy T Yu, Rindranirina Ramamonjison, Giuseppe Carenini, Zirui Zhou, and Yong Zhang. 2024. Optimization modeling and verification from problem specifications using a multi-agent multi-stage llm framework. INFOR: Information Systems and Operational Research, 62(4):599– 617. find
  14. Daye Nam, Andrew Macvean, Vincent Hellendoorn, Bogdan Vasilescu, and Brad Myers. 2024. Using an llm to help with code understanding. In Proceedings of the IEEE/ACM 46th International Conference on Software Engineering, pages 1–13. find
  15. Hamed Rahimian and Sanjay Mehrotra. 2019. Distributionally robust optimization: A review. arXiv preprint arXiv:1908.05659. — [arXiv:1908.05659](https://arxiv.org/abs/1908.05659) find
  16. Rindranirina Ramamonjison, Haley Li, Timothy T Yu, Shiqi He, Vishnu Rengan, Amin Banitalebi-Dehkordi, Zirui Zhou, and Yong Zhang. 2022. Augmenting operations research with auto-formulation of optimization models from problem descriptions. arXiv preprint arXiv:2209.15565. find
  17. Rindranirina Ramamonjison, Timothy Yu, Raymond Li, Haley Li, Giuseppe Carenini, Bissan Ghaddar, Shiqi He, Mahdi Mostajabdaveh, Amin Banitalebi-Dehkordi, Zirui Zhou, and 1 others. 2023. Nl4opt competition: Formulating optimization problems based on their natural language descriptions. In NeurIPS 2022… find
  18. Bernardino Romera-Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, M Pawan Kumar, Emilien Dupont, Francisco JR Ruiz, Jordan S Ellenberg, Pengming Wang, Omar Fawzi, and 1 others. 2024. Mathematical discoveries from program search with large language models. Nature, 625(7995):468–475.
  19. Utsav Sadana, Abhilash Chenreddy, Erick Delage, Alexandre Forel, Emma Frejinger, and Thibaut Vidal. 2025. A survey of contextual optimization methods for decision-making under uncertainty. European Journal of Operational Research, 320(2):271–289. find
  20. Xuezhi Wang, Jason Wei, Dale Schuurmans, Quoc Le, Ed Chi, Sharan Narang, Aakanksha Chowdhery, and Denny Zhou. 2022. Self-consistency improves chain of thought reasoning in language models. arXiv preprint arXiv:2203.11171.
  21. Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Fei Xia, Ed Chi, Quoc V Le, Denny Zhou, and 1 others. 2022. Chain-of-thought prompting elicits reasoning in large language models. Advances in neural information processing systems, 35:24824– 24837.
  22. Ziyang Xiao, Dongxiang Zhang, Yangjun Wu, Lilin Xu, Yuan Jessica Wang, Xiongwei Han, Xiaojin Fu, Tao Zhong, Jia Zeng, Mingli Song, and 1 others. 2023. Chain-of-experts: When llms meet complex operations research problems. In The Twelfth International Conference on Learning Representations. find
  23. 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.
  24. Yang Zhang, Hanlei Jin, Dan Meng, Jun Wang, and Jinghua Tan. 2024. A comprehensive survey on process-oriented automatic text summarization with exploration of llm-based methods. arXiv preprint arXiv:2403.02901. — [arXiv:2403.02901](https://arxiv.org/abs/2403.02901) find
  25. Chuanyang Zheng, Zhengying Liu, Enze Xie, Zhenguo Li, and Yu Li. 2023. Progressive-hint prompting improves reasoning in large language models. arXiv preprint arXiv:2304.09797. find

Updated

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/papers/she-2026-crossroads-optimization-under-uncertainty-how/