Research map / Graph learning / Graph self-supervised learning
UniRTL: Unifying Code and Graph for Robust RTL Representation Learning
Yi Liu, Hongji Zhang, Lei Chen, Mingxuan Yuan, Qiang Xu
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 graph modality to code-only UniRTL (ablation) | UniRTL · Performance prediction (w/o netlist info), Area | R2: 0.5173 -> 0.7682 | Improved |
| add code modality to graph-only UniRTL (ablation) | UniRTL · Performance prediction (w/o netlist info), Area | R2: 0.7546 -> 0.7682 | Improved |
| add graph modality to code-only UniRTL (ablation) | UniRTL · Performance prediction (w/o netlist info), Delay | R2: 0.3839 -> 0.7832 | Improved |
| add code modality to graph-only UniRTL (ablation) | UniRTL · Performance prediction (w/o netlist info), Delay | R2: 0.7602 -> 0.7832 | Improved |
| add graph modality to code-only UniRTL (ablation) | UniRTL · Performance prediction (w/ netlist info), Area | R2: 0.7547 -> 0.8879 | Improved |
| add graph modality to code-only UniRTL (ablation) | UniRTL · Natural language code search | F1 (main): 0.644 -> 0.662 | Improved |
| add graph modality to code-only UniRTL (ablation) | UniRTL · Functionality equivalence checking | AP (main): 0.712 -> 0.745 | Improved |
| use full alignment strategy instead of direct combination of code and graph (ablation) | UniRTL · Performance prediction (w/o netlist info), Area | R2: 0.7602 -> 0.7682 | Improved |
| use full alignment strategy instead of direct combination of code and graph (ablation) | UniRTL · Natural language code search | F1 (main): 0.650 -> 0.662 | Improved |
| UniRTL code-only pretraining vs plain CodeBERT (ablation) | UniRTL (w/o graph) · Performance prediction (w/o netlist info), Area | R2: 0.3858 -> 0.5173 | Improved |
| increase fine-tuning data from 5% to 100% | UniRTL · Performance prediction (w/o netlist info), Area (few-shot) | R2: -0.0604 -> 0.7682 | Improved |
| increase fine-tuning data from 10% to 20% | UniRTL · Performance prediction (w/o netlist info), Area (few-shot) | R2: 0.4334 -> 0.5630 | Improved |
Closest papers by meaning
- Bridging Layout and RTL: Knowledge Distillation based Timing Prediction (ICML 2025 · similarity 0.87)
- CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design (ICLR 2025 · similarity 0.86)
- DeepRTL: Bridging Verilog Understanding and Generation with a Unified Representation Model (ICLR 2025 · similarity 0.85)
- SymRTLO: Enhancing RTL Code Optimization with LLMs and Neuron-Inspired Symbolic Reasoning (NeurIPS 2025 · similarity 0.85)
- Circuit Representation Learning with Masked Gate Modeling and Verilog-AIG Alignment (ICLR 2025 · similarity 0.83)
- CraftRTL: High-quality Synthetic Data Generation for Verilog Code Models with Correct-by-Construction Non-Textual Representations and Targeted Code Repair (ICLR 2025 · similarity 0.83)
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.
