SchGen_dataset

SchGen SchGen** is a dataset of approximately **8K paired natural-language requests and Python-based schematic generation code** for research on LLM-driven PCB schematic generation. The generated Python code can be rendered into **KiCad schematic designs**, enabling research on hardware generation from natural-language descriptions. ➡️ **Paper:** ➡️ **License:** MIT License Overview SchGen contains: 8,420 pairs of: Natural-language hardware design requests Python schematic-generation code The dataset was developed for: Fine-tuning LLMs on schematic generation Benchmarking hardware-generation capabilities Research on AI-assisted PCB design workflows The data was created between **August 2025 and September 2025**. Dataset Construction The dataset was created through a multi-stage pipeline: 1. GPT-5 generated draft schematic designs based on open-source hardware references 2. Human annotation and correction ensured schematic correctness in KiCad 3. GPT-5 synthesized corresponding user requests from finalized schematics Although open-source hardware references (e.g. SparkFun schematic PDFs under CC BY-SA 4.0) were used during drafting, the final dataset does not directly contain those original designs. Features 🔌 **KiCad-Compatible Generation** Python code can be rendered into executable KiCad schematic designs. 🧠 **Natural Language to Hardware** Enables research on generating PCB schematics directly from user requests. 📐 **Structured Schematic APIs** Uses programmatic schematic construction APIs rather than raw schematic files. 🛠️ **Research-Oriented Dataset** Designed for experimentation, benchmarking, and fine-tuning of LLMs. Evaluation The dataset was evaluated using several schematic-generation metrics: **Valid Circuits** Measures whether generated code executes successfully and produces valid schematics. **Spatial Violation** Measures overlaps among schematic objects such as symbols, labels, and wires. **Netlist Accuracy** Compares generated netlists with ground-t