To See is Not to Master: Teaching LLMs to Use Private Libraries for Code Generation

Researchers introduced PriCoder, a new approach that improves Large Language Models' ability to generate code using private library APIs by over 20%. The method uses automatically synthesized training data through graph-based operators to teach LLMs private library usage, addressing a key limitation in current AI coding capabilities.

• →Current LLMs struggle with private-library-oriented code generation even when provided with accurate API documentation.
• →PriCoder uses Progressive Graph Evolution and Multidimensional Graph Pruning to synthesize diverse, high-quality training data.
• →The approach achieved over 20% improvement in pass@1 rates across three mainstream LLMs without affecting general coding performance.
• →Two new benchmarks based on recently released libraries were created to evaluate private-library code generation capabilities.
• →The research addresses a significant gap in AI code generation for enterprise and specialized development environments.