Anka: A DSL for Reliable LLM Code Generation

This paper introduces Anka, a domain-specific language (DSL) designed to improve the reliability of code generation by Large Language Models (LLMs). It argues that the flexibility of general-purpose languages leads to errors in complex programming tasks. The paper's significance lies in demonstrating that LLMs can learn novel DSLs from in-context prompts and that constrained syntax can significantly reduce errors, leading to higher accuracy on complex tasks compared to general-purpose languages like Python. The release of the language implementation, benchmark suite, and evaluation framework is also important for future research.