Research map / Graph learning / GNN architectures
Si-GT: Fast Interconnect Signal Integrity Analysis for Integrated Circuit Design via Graph Transformers
Yuting Hu, Tarek Mohamed, Chenhui Xu, Hua Xiang, Hussam Amrouch, Gi-Joon Nam, Jinjun Xiong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add virtual NET tokens (ablation) | Si-GT · crosstalk glitch V Segment peak voltage | mean relative accuracy: 89.49 -> 97.70 | Improved |
| add virtual NET tokens (ablation) | Si-GT · crosstalk glitch V Segment noise width | mean relative accuracy: 94.97 -> 98.12 | Improved |
| add virtual NET tokens (ablation) | Si-GT · crosstalk delay AV Segment aggressor | mean relative accuracy: 72.58 -> 73.30 | Improved |
| add mesh pattern encoding on top of NET tokens (ablation) | Si-GT · crosstalk delay AV Segment aggressor | mean relative accuracy: 73.30 -> 73.48 | Improved |
| add intra-net attention bias on top of NET and MPE (ablation) | Si-GT · crosstalk delay AV Segment aggressor | mean relative accuracy: 73.48 -> 73.40 | No clear effect |
| add inter-net attention bias (full IIN) on top of intra-net bias (ablation) | Si-GT · crosstalk delay AV Segment aggressor | mean relative accuracy: 73.40 -> 73.67 | Improved |
| add full IIN attention on top of NET and MPE (ablation) | Si-GT · crosstalk glitch V Segment peak voltage | mean relative accuracy: 97.85 -> 97.78 | Hurt |
| add spatial and edge encoding attention biases (ablation) | Si-GT · crosstalk delay AV Segment victim | mean relative accuracy: 87.25 -> 88.32 | Improved |
| replace mesh pattern encoding with RWSE16 in Si-GT (i.e. MPE vs RWSE16) (ablation) | Si-GT · crosstalk glitch Segment noise width | mean relative accuracy: 97.44 -> 98.36 | Improved |
| use directed instead of undirected interconnect graphs (ablation) | DeepGCN · crosstalk delay AV Segment aggressor | accuracy difference: +11.48 | Improved |
| use directed instead of undirected interconnect graphs (ablation) | Si-GT-GCN · segment delay and glitch | accuracy difference: -0.13 | No clear effect |
| train on segment data instead of sink data for sink-level victim delay (ablation) | Si-GT-GCN · crosstalk delay sink victim | accuracy difference: +0.08 | No clear effect |
Closest papers by meaning
- AnalogGenie-Lite: Enhancing Scalability and Precision in Circuit Topology Discovery through Lightweight Graph Modeling (ICML 2025 · similarity 0.85)
- AnalogToBi: Device-Level Analog Circuit Topology Generation via Bipartite Graph and Grammar Guided Decoding (NeurIPS 2026 · similarity 0.83)
- DeepGate4: Efficient and Effective Representation Learning for Circuit Design at Scale (ICLR 2025 · similarity 0.82)
- Circuit Representation Learning with Masked Gate Modeling and Verilog-AIG Alignment (ICLR 2025 · similarity 0.82)
- PreRoutGNN for Timing Prediction with Order Preserving Partition: Global Circuit Pre-training, Local Delay Learning and Attentional Cell Modeling (AAAI 2024 · similarity 0.82)
- EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification (NeurIPS 2025 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
