arxivcs.ARcs.LG2026-07-31
RTLCurator: Label-Efficient Data Curation for RTL Generation
Siyang Cai, Cangyuan Li, Wenjing Chang, Kun Wang, Haoyu Gao, Yinhe Han, et al.
Training large language models (LLMs) to write register-transfer level (RTL) requires large corpora of paired specifications and code, and such data is scarce enough that most public corpora are now synthesized. Synthesis provides scale but not correctness, and in two widely used…