PCB-Bench

Published as a Conference Paper at ICLR 2026

PCB-Bench: Benchmarking LLMs for Printed Circuit Board Placement and Routing

Jindong Li*, Lianrong Chen*, Bin Yang*, Jiadong Zhu*, Ying Wang, Yuzhe Ma, Menglin Yang†

*Equal contribution    †Corresponding author

The Hong Kong University of Science and Technology (Guangzhou)

📄 Paper (OpenReview) 💻 Code (GitHub) 🌐 Project Page

PCB-Bench is the first comprehensive benchmark designed to systematically evaluate (multimodal) large language models (LLMs/MLLMs) in the context of PCB placement and routing. It integrates text, images, and real PCB design artifacts into a unified evaluation framework for realistic PCB engineering reasoning assessment.

Introduction

PCB-Bench highlights the gap between existing publicly available PCB-related benchmarks and realistic PCB design evaluation, and provides a unified benchmark suite for PCB placement and routing reasoning.

Table 1

Benchmark Overview

PCB-Bench spans three complementary task settings and corresponding datasets: (1) text-based reasoning (Text-to-Text QA & CQ), (2) multimodal image-text reasoning (Image-and-Text QA/CQ), and (3) real-world PCB design comprehension (PCB Design Understanding).

Figure 1
~3,700 Text QA/CQ
~500 Multimodal Problems
174 Real-world PCB Projects
Figure 2

Models & Results

We benchmark a diverse set of state-of-the-art LLMs/MLLMs under a unified zero-shot protocol. The results reveal substantial performance variation across tasks, reflecting the remaining challenges in PCB placement and routing reasoning.

Figure 5
Table 2
Table 3
Table 4

Data Sources & Licensing

PCB designs are collected from publicly available and legally accessible sources (including OSHWHub operated by JLCPCB), with no proprietary or sensitive industrial data involved. Each real-world project is associated with traceable artifacts for transparency and reproducibility.

Figure 4
Figure 6

Citation

@inproceedings{li2026pcbbench,
  title     = {PCB-Bench: Benchmarking LLMs for Printed Circuit Board Placement and Routing},
  author    = {Jindong Li and Lianrong Chen and Bin Yang and Jiadong Zhu and Ying Wang and Yuzhe Ma and Menglin Yang},
  booktitle = {International Conference on Learning Representations (ICLR)},
  year      = {2026}
}

© PCB-Bench (ICLR 2026). For academic research and benchmarking.