Research map / Large language models / Alignment and preferences
Expert-level Leaf Cell Layout Generation via Preference-Optimized LLM
Yaohui Han, Rongliang Fu, Yanming Liu, Shuo Ren, Shuai Dong, Yunpeng Wang, Tinghuan Chen, Bei Yu, Tsung-Yi Ho
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 SFT stage on BO-generated golden scripts (ablation) | GenLeaf · Leaf cell layout generation (ablation) | tracks used: +30.9% without SFT | Improved |
| add SFT stage on BO-generated golden scripts (ablation) | GenLeaf · Leaf cell layout generation (ablation) | wirelength: +21.7% without SFT | Improved |
| add SFT stage on BO-generated golden scripts (ablation) | GenLeaf · Leaf cell layout generation (ablation) | via count: +36.6% without SFT | Improved |
| add GRPO preference optimization after SFT (ablation) | GenLeaf · Leaf cell layout generation (ablation) | tracks used: +3.3% without GRPO | Improved |
| add GRPO preference optimization after SFT (ablation) | GenLeaf · Leaf cell layout generation (ablation) | wirelength: +13.2% without GRPO | Improved |
| add GRPO preference optimization after SFT (ablation) | GenLeaf · Leaf cell layout generation (ablation) | via count: +20.0% without GRPO | Improved |
| fine-tune Qwen3 (14B) with SFT and GRPO | Qwen3 (14B) · Leaf cell layout generation | tracks used: 5.5 -> 3.9 | Improved |
| use performance-aware layout embedding instead of Circuit GNN | layout representation · downstream prediction | MSE: 1.287 -> 0.849 | Improved |
| GenLeaf vs golden expert design | GenLeaf · 14 industrial leaf cells | tracks used: ratio 1.00 -> 0.52 | Improved |
| GenLeaf vs golden expert design | GenLeaf · 14 industrial leaf cells | via count: ratio 1.00 -> 1.01 | No clear effect |
| GenLeaf vs Claude Sonnet 4 | GenLeaf · Leaf cell layout generation | wirelength: 46.36 -> 39.05 | Improved |
Closest papers by meaning
- DeepLayout: Learning Neural Representations of Circuit Placement Layout (ICML 2025 · similarity 0.83)
- EGPlace: An Efficient Macro Placement Method via Evolutionary Search with Greedy Repositioning Guided Mutation (ICML 2025 · similarity 0.83)
- See it to Place it: Evolving Macro Placements with Vision Language Models (NeurIPS 2026 · similarity 0.82)
- AUTOCIRCUIT-RL: Reinforcement Learning-Driven LLM for Automated Circuit Topology Generation (ICML 2025 · similarity 0.82)
- CircuitFusion: Multimodal Circuit Representation Learning for Agile Chip Design (ICLR 2025 · similarity 0.82)
- Towards Next-Generation Logic Synthesis: A Scalable Neural Circuit Generation Framework (NeurIPS 2024 · 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.
