LLMs generate domain-specific language code from natural language prompts

Researchers have introduced Text2DSL, a method for generating code for domain-specific languages (DSLs) from natural language descriptions. They developed the PolkitBench dataset, containing over 4,000 natural-language-to-Polkit-rule pairs, validated through an AST-based pipeline. Experiments with GigaChat-10B and Nemotron-3-Nano models showed that providing structured context, such as BNF grammar and API specifications, significantly improves code generation quality, increasing syntactic validity to nearly 99% and CodeBLEU scores by up to 95%. This approach enables high-quality DSL code generation without requiring model fine-tuning. AI

IMPACT Enhances the usability of domain-specific languages by enabling code generation from natural language, potentially lowering the barrier to entry for complex policy management.

RANK_REASON The cluster contains an academic paper detailing a new method and dataset for code generation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• Alexander Kozachok
• GigaChat-10B-A1.8B
• Nemotron-3-Nano-30B-A3B
• PolkitBench
• Text2DSL

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →