PDF
Large Language Models in Code Generation: Capabilities, Limits, and Safety
Chen Wei
In: Machine Learning and Artificial Intelligence: Trends and Applications
We systematically evaluate large language models (LLMs) for automated code generation across multiple programming languages and task types. Our study assesses functional correctness, security vulnerabilities, and alignment with developer intent using a novel benchmark suite. We discuss prompt engineering strategies that improve generation quality and safety filtering mechanisms.