LLaMoCo: Instruction Tuning of Large Language Models for Optimization Code Generation
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2024
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911789037387776 |
|---|---|
| author | Ma, Zeyuan Guo, Hongshu Chen, Jiacheng Peng, Guojun Cao, Zhiguang Ma, Yining Gong, Yue-Jiao |
| author_facet | Ma, Zeyuan Guo, Hongshu Chen, Jiacheng Peng, Guojun Cao, Zhiguang Ma, Yining Gong, Yue-Jiao |
| contents | Recent research explores optimization using large language models (LLMs) by either iteratively seeking next-step solutions from LLMs or directly prompting LLMs for an optimizer. However, these approaches exhibit inherent limitations, including low operational efficiency, high sensitivity to prompt design, and a lack of domain-specific knowledge. We introduce LLaMoCo, the first instruction-tuning framework designed to adapt LLMs for solving optimization problems in a code-to-code manner. Specifically, we establish a comprehensive instruction set containing well-described problem prompts and effective optimization codes. We then develop a novel two-phase learning strategy that incorporates a contrastive learning-based warm-up procedure before the instruction-tuning phase to enhance the convergence behavior during model fine-tuning. The experiment results demonstrate that a CodeGen (350M) model fine-tuned by our LLaMoCo achieves superior optimization performance compared to GPT-4 Turbo and the other competitors across both synthetic and realistic problem sets. The fine-tuned model and the usage instructions are available at https://anonymous.4open.science/r/LLaMoCo-722A. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2403_01131 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | LLaMoCo: Instruction Tuning of Large Language Models for Optimization Code Generation Ma, Zeyuan Guo, Hongshu Chen, Jiacheng Peng, Guojun Cao, Zhiguang Ma, Yining Gong, Yue-Jiao Optimization and Control Artificial Intelligence Computation and Language Machine Learning Neural and Evolutionary Computing Software Engineering Recent research explores optimization using large language models (LLMs) by either iteratively seeking next-step solutions from LLMs or directly prompting LLMs for an optimizer. However, these approaches exhibit inherent limitations, including low operational efficiency, high sensitivity to prompt design, and a lack of domain-specific knowledge. We introduce LLaMoCo, the first instruction-tuning framework designed to adapt LLMs for solving optimization problems in a code-to-code manner. Specifically, we establish a comprehensive instruction set containing well-described problem prompts and effective optimization codes. We then develop a novel two-phase learning strategy that incorporates a contrastive learning-based warm-up procedure before the instruction-tuning phase to enhance the convergence behavior during model fine-tuning. The experiment results demonstrate that a CodeGen (350M) model fine-tuned by our LLaMoCo achieves superior optimization performance compared to GPT-4 Turbo and the other competitors across both synthetic and realistic problem sets. The fine-tuned model and the usage instructions are available at https://anonymous.4open.science/r/LLaMoCo-722A. |
| title | LLaMoCo: Instruction Tuning of Large Language Models for Optimization Code Generation |
| topic | Optimization and Control Artificial Intelligence Computation and Language Machine Learning Neural and Evolutionary Computing Software Engineering |
| url | https://arxiv.org/abs/2403.01131 |