Type-Constrained Code Generation with Language Models

Niels Mündler, Jingxuan He, Hao Wang +3

Large language models (LLMs) have achieved notable success in code generation. However, they still frequently produce uncompilable output because their next-token inference procedu…