Chenyu Huang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Research Interests
Chenyu Huang has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
Ruoqing Jiang
2 shared publicationsZhengyang Tang
2 shared publicationsBenyou Wang
1 shared publication- Benyou Wang
The Chinese University of Hong Kong, Shenzhen, China
1 shared publication Bo Jiang
1 shared publication- Dongdong Ge
Shanghai Jiao Tong University
1 shared publication
Latest Publications
- Open Access
- 25 June 2026
- Published by Elsevier in SSRN Electronic Journal
InvEvolve: Evolving White-box Inventory Policies via Large Language Models with Performance Guarantees
We study how large language models can be used to generate inventory policies in online settings with non-stationary demand. Our work is motivated by recent advances in LLM-based evolutionary search, such as AlphaEvolve, which demonstrates strong performance on static and highly structured problems such as mathematical discovery, but is not directly suited to dynamic inventory settings with online updates. We propose InvEvolve, an end-to-end inventory policy evolution and inference framework grounded in confidenceinterval-based certification. Built on a large language model trained via reinforcement learning, InvEvolve can process demand data together with additional numerical and textual features, and generates white-box inventory policies with statistical safety guarantees for future deployment. We further introduce a unified framework with theoretical guarantees that connects training, inference, and deployment. This allows us to derive a lower bound on the probability that InvEvolve evolves a statistically safe and improved policy, and to characterize the multi-period performance gap relative to the oracle-safe benchmark. Tested on both synthetic data and real-world retail data, InvEvolve outperforms classical inventory policies and deep-learning-based methods. In canonical inventory settings, it generates new policies that outperform existing benchmarks.
- 25 June 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessInvEvolve: Evolving White-box Inventory Policies via Large Language Models with Performance Guarantees
We study how large language models can be used to generate inventory policies in online settings with non-stationary demand. Our work is motivated by recent advances in LLM-based evolutionary search, such as AlphaEvolve, which demonstrates strong performance on static and highly structured problems such as mathematical discovery, but is not directly suited to dynamic inventory settings with online updates. We propose InvEvolve, an end-to-end inventory policy evolution and inference framework grounded in confidenceinterval-based certification. Built on a large language model trained via reinforcement learning, InvEvolve can process demand data together with additional numerical and textual features, and generates white-box inventory policies with statistical safety guarantees for future deployment. We further introduce a unified framework with theoretical guarantees that connects training, inference, and deployment. This allows us to derive a lower bound on the probability that InvEvolve evolves a statistically safe and improved policy, and to characterize the multi-period performance gap relative to the oracle-safe benchmark. Tested on both synthetic data and real-world retail data, InvEvolve outperforms classical inventory policies and deep-learning-based methods. In canonical inventory settings, it generates new policies that outperform existing benchmarks.
- Top Cited Papers
- 1 November 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 73 (6) , 2986-3009
ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling
ORLM: Pioneering Open-Source Framework for Automated Optimization Modeling A study titled "ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling" has been published, introducing the first open-source framework designed to automate optimization modeling using large language models (LLMs). This innovative approach addresses critical challenges in the field of operations research (OR), particularly the overreliance on closed-source LLMs like GPT-4, which raises privacy concerns and limits customization in industrial applications. The research team proposed OR-Instruct, a semiautomated data synthesis framework that generates high-quality training data tailored to specific optimization modeling requirements. They also introduced IndustryOR, the first benchmark for evaluating LLMs’ performance on real-world OR problems. By training several 7B-scale open-source LLMs with the synthesized data, the team achieved state-of-the-art results across multiple benchmarks, including NL4Opt, MAMO, and IndustryOR. This advancement not only enhances the accessibility and applicability of optimization modeling, but also paves the way for more efficient and privacy-conscious solutions in various industrial sectors. The ORLM framework exemplifies the potential of open-source initiatives in driving innovation and democratizing advanced analytical tools for operations research.
- 1 November 2025
- Published by Institute for Operations Research and the Management Sciences (INFORMS) in Operations Research
- Vol. 73 (6) , 2986-3009
Top Cited PapersORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling
ORLM: Pioneering Open-Source Framework for Automated Optimization Modeling A study titled "ORLM: A Customizable Framework in Training Large Models for Automated Optimization Modeling" has been published, introducing the first open-source framework designed to automate optimization modeling using large language models (LLMs). This innovative approach addresses critical challenges in the field of operations research (OR), particularly the overreliance on closed-source LLMs like GPT-4, which raises privacy concerns and limits customization in industrial applications. The research team proposed OR-Instruct, a semiautomated data synthesis framework that generates high-quality training data tailored to specific optimization modeling requirements. They also introduced IndustryOR, the first benchmark for evaluating LLMs’ performance on real-world OR problems. By training several 7B-scale open-source LLMs with the synthesized data, the team achieved state-of-the-art results across multiple benchmarks, including NL4Opt, MAMO, and IndustryOR. This advancement not only enhances the accessibility and applicability of optimization modeling, but also paves the way for more efficient and privacy-conscious solutions in various industrial sectors. The ORLM framework exemplifies the potential of open-source initiatives in driving innovation and democratizing advanced analytical tools for operations research.