BRIDGES introduces a novel framework that integrates graph modality with Large Language Models to significantly improve performance on Electronic Design Automation (EDA) tasks.

• Overcomes limitations of representing graph-structured data (RTL code, netlists) as sequential text
• Leverages dataflow graphs and other graph-structured information that traditional LLMs ignore
• Demonstrates measurable performance improvements in EDA applications
• Provides a more natural approach to handling the inherently graph-based nature of hardware design

This research matters because modern chip design and verification workflows rely heavily on graph representations, and better LLM integration could dramatically accelerate the engineering design process, reduce errors, and enable more complex systems.

BRIDGES: Bridging Graph Modality and Large Language Models within EDA Tasks