AI
RFBench: A Simulator-Grounded Multimodal Benchmark for Passive RF Layout Design and Understanding
Publication: cs.AI updates on arXiv.org
Author: Jiarui Cui, Boxun Yan
Source published: 19/11/2026, 05:00:00 UTC
Briefing created: 19/11/2026, 06:23:12 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 19/11/2026, 05:03:24 UTC.
Briefings
RFBench is a simulator-grounded benchmark dataset for passive RF layout design, featuring 10,000 two-port planar layouts simulated with ADS Momentum. The dataset includes 16x16 occupancy grids, complex S-parameters at 201 frequency points, and seven tasks evaluating multimodal model capabilities. Extensive testing across seven multimodal language models revealed systematic weaknesses in generating structurally valid layouts and predicting electromagnetically faithful responses, with RF-specific fine-tuning improving structural validity from 0.0% to 83.2% and response prediction rates from 17.7% to 62.8%. This work establishes a common testbed for evaluating multimodal RF capabilities.
Key points
- RFBench contains 10,000 two-port planar layouts simulated using ADS Momentum with 16x16 occupancy grids and 201 frequency points of S-parameters.
- Seven tasks were defined to evaluate multimodal capabilities, including specification-conditioned layout generation and response prediction.
Context
This benchmark addresses the gap where multimodal language models can interpret diagrams but lack the capacity to connect passive RF design intent, layout geometry, and electromagnetic behavior. The simulator-grounded nature of RFBench provides a standardized testbed for evaluating model performance in realistic RF design scenarios.
Limitations
The dataset is limited to 10,000 two-port planar layouts simulated with ADS Momentum. Results are based on extensive baselines across seven multimodal language models, and the study does not include additional background context beyond the dataset description and task definitions.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 19/11/2026, 06:23:12 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeidljzz2gb4f5qiw52w26undsmwvtcufgnkbztmbtsezpi7csgh4hu