RFBench: A Simulator-Grounded Multimodal Benchmark for Passive RF Layout Design and Understanding
Summary
- Researchers introduce RFBench, a simulator-grounded multimodal dataset and benchmark for passive radio-frequency layout design and understanding.
- The dataset features 10,000 two-port planar layouts simulated with ADS Momentum, including geometric primitives, raster images, occupancy grids, and S-parameters.
- RFBench defines seven tasks spanning layout generation, translation, response prediction, validation, diagnosis, and classification.
- Baseline evaluations across seven multimodal language models show systematic weaknesses in structural validity and electromagnetic response prediction.
- RF-specific fine-tuning significantly improves structural validity and primitive-input accuracy, highlighting the effectiveness of domain adaptation.
Why it matters
RFBench bridges the gap between multimodal language models and specialized physical engineering by providing a rigorous testbed for connecting RF design intent with electromagnetic behavior.