VeriSelector improves verilog code generation using diverse quality data
Quality over Quantity: Diversity-Aware Data Selection for Efficient Verilog Code Generation
Hardware Architecture
Summary
Verilog code is used to design electronic circuits, but training AI to write Verilog code is hard because many example codes are repetitive or contain errors. The authors created VeriSelector, a tool that picks the best and most varied Verilog examples for training. It checks if code works correctly, measures how hard instructions are to follow, and uses detailed features of the code to ensure variety. This careful selection helps AI models learn better and faster, even from less data.
What this means in practice
- •For hardware design teams: Improve AI-assisted Verilog code generation quality and speed by training models on carefully selected diverse datasets.
- •For software engineering teams: Reduce AI model training time and data requirements while maintaining code generation performance for hardware description languages.
Authors
Yiheng Shen, Wei Zheng, Xiao Wei, Hao Shen, Xiang Chen, Guang Yang
Abstract
Large Language Models (LLMs) have shown remarkable potential in Verilog code generation, yet existing datasets contain con siderable noise and redundancy. Prior data selection methods address only isolated quality aspects, neglect the global diversity of the training set, and cannot capture Verilog-specific structural semantics. To bridge this gap, we propose VeriSelector, the first data selection framework for Verilog code generation that jointly optimizes quality and diversity. We formulate the selec tion problem as a constrained bi-objective subset selection problem and solve it via a three-stage approximation. For quality, a multi-granularity pipeline first verifies functional correctness through testbench simulation and then filters misaligned samples via Instruction-Following Difficulty (IFD) scoring. For diversity, 109-dimensional Verilog-specific structural features (AST, CFG, and Netlist) are fused with textual embeddings for clustering-based diversity modeling. A proportional adaptive sampling strat egy then allocates per-cluster quotas guided by IFD ranks, with a provable distribution preservation guarantee. Experiments on three LLMs and three benchmarks show that VeriSelector outperforms full-dataset training and state-of-the-art baselines us ing only 20%-25% of the data, achieving Performance Retention Rates above 118% and reducing training time by over 80%. Notably, VeriSelector improves average Pass@1 by 18.49%-29.43% over full-dataset training and by 1.36%-7.95% over the best-performing baseline across all evaluated models.