Structured Testbench Generation for LLM-Driven HDL Design and Verification-Oriented Data Curation

In three linesSTG, a Structured Testbench Generation framework, accelerates verification of LLM-generated HDL designs. 720x faster than iterative LLM approaches, it reduces false positives, improves coverage, and identifies errors in existing benchmarks. As a data curation engine, it is 11x faster than LLM-based filtering with 127x less energy consumption.

Summary generated by Claude — human-verified