RuC: Grammar-Driven Benchmark for RTL Code Completion

A new framework called RuC (Rule Completion) generates benchmarks for evaluating large language models on Register Transfer Level code completion tasks. Unlike existing benchmarks that assess entire modules or single lines, RuC uses grammar rules to mask syntactically defined code regions, enabling control over completion granularity and syntactic range. The system automatically produces tasks from hardware description sources, targeting any hardware description language. This addresses limitations in current LLM evaluation for hardware design, where precise code completion is critical.