Research map / Graph learning / Knowledge and temporal graphs
FusionCell: Cross-Attentive Fusion of Layout Geometry and Netlist Topology for Standard-Cell Performance Prediction
Haoyi Zhang, Kairong Guo, Bojie Zhang, Yibo Lin, Runsheng Wang
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 correlation edges between related nets (ablation) | FusionCell · ASAP7 standard-cell dataset | average MAPE: 1.45% -> 0.92% | Improved |
| add correlation edges between related nets (ablation) | FusionCell · ASAP7 standard-cell dataset | average Spearman rho: 0.82 -> 0.86 | Improved |
| add correlation edges between related nets (ablation) | FusionCell · ASAP7 standard-cell dataset | average Kendall tau: 0.73 -> 0.76 | Improved |
| add correlation edges between related nets (ablation) | FusionCell · ASAP7 standard-cell dataset | average R2: 0.975 -> 0.977 | Improved |
| replace symmetrical bidirectional cross-attention with graph-query layout-key cross-attention (ablation) | dual-stream fusion · ASAP7 standard-cell dataset | average MAPE: 2.19% -> 0.92% | Improved |
| replace symmetrical bidirectional cross-attention with graph-query layout-key cross-attention (ablation) | dual-stream fusion · ASAP7 standard-cell dataset | average Spearman rho: 0.81 -> 0.86 | Improved |
| replace late fusion (concatenated pooled vectors) with graph-query layout-key cross-attention (ablation) | dual-stream model · ASAP7 standard-cell dataset | average MAPE: 2.33% -> 0.92% | Improved |
| use symmetrical fusion instead of late fusion (ablation) | dual-stream model · ASAP7 standard-cell dataset | average MAPE: 2.33% -> 2.19% | No clear effect |
| add netlist graph input to layout-only model with topology-guided fusion (ablation) | DeiT vision-only · ASAP7 standard-cell dataset | average MAPE: 3.16% -> 0.92% | Improved |
| add netlist graph input to layout-only model with topology-guided fusion (ablation) | DeiT vision-only · ASAP7 standard-cell dataset | average Spearman rho: 0.56 -> 0.86 | Improved |
| replace coarse layout-augmented graph (ProtoCellLayout) with explicit DeiT layout encoding and cross-attention | layout representation · ASAP7 standard-cell dataset | average MAPE: 4.16% -> 0.92% | Improved |
| replace coarse layout-augmented graph (ProtoCellLayout) with explicit DeiT layout encoding and cross-attention | layout representation · ASAP7 standard-cell dataset | average Spearman rho: 0.30 -> 0.86 | Improved |
Closest papers by meaning
- PreRoutGNN for Timing Prediction with Order Preserving Partition: Global Circuit Pre-training, Local Delay Learning and Attentional Cell Modeling (AAAI 2024 · similarity 0.87)
- DeepLayout: Learning Neural Representations of Circuit Placement Layout (ICML 2025 · similarity 0.86)
- Bridging Layout and RTL: Knowledge Distillation based Timing Prediction (ICML 2025 · similarity 0.85)
- AnalogGenie-Lite: Enhancing Scalability and Precision in Circuit Topology Discovery through Lightweight Graph Modeling (ICML 2025 · similarity 0.84)
- CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design (ICLR 2025 · similarity 0.84)
- NeuralSteiner: Learning Steiner Tree for Overflow-avoiding Global Routing in Chip Design (NeurIPS 2024 · similarity 0.84)
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.
