This research evaluates strategies to enhance Large Language Models' performance when generating code in programming languages with limited training data.

• Fine-tuning with limited data shows modest improvements but remains insufficient
• Learning patterns from similar high-resource languages yields inconsistent results
• No single technique provides a comprehensive solution for all low-resource languages
• Multi-pronged approaches are needed to address this engineering challenge

For software engineering teams working with specialized languages, this research highlights the continued challenges in automating code generation for niche programming contexts, emphasizing the need for targeted solutions.

Enhancing Code Generation for Low-Resource Languages: No Silver Bullet