Research map / Large language models / Agents, code and math
Evolving Graph Structured Programs for Circuit Generation with Large Language Models
Yinqi Bai, Jie Wang, Lei Chen, Zhihai Wang, Yumeng Li, Mingxuan Yuan, Jianye HAO, Defu Lian, Enhong Chen
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 |
|---|---|---|---|
| use an LLM generator instead of a rule-based linear genetic program (ablation) | CircuitEvo (GPT-3.5-turbo) · Espresso4 | init nodes: 940 -> 848 | Improved |
| use the graph-structured circuit program instead of a Boolean function format (ablation) | CircuitEvo (GPT-3.5-turbo) · Espresso4 | init nodes: 940 -> 848 | Improved |
| add exploration and refinement evolutionary strategies (ablation) | CircuitEvo (GPT-3.5-turbo) · Espresso4 | init nodes: 911 -> 848 | Improved |
| add local program search before function completion (ablation) | CircuitEvo (GPT-3.5-turbo) · Espresso4 | init nodes: 911 -> 848 | Improved |
| add function completion (ablation) | CircuitEvo (GPT-3.5-turbo) · Espresso4 | accuracy (%): 86.4 -> 100.0 | Improved |
| add function completion (ablation) | CircuitEvo (GPT-3.5-turbo) · Arithmetic3 | accuracy (%): 67.8 -> 100.0 | Improved |
| add local program search before function completion (ablation) | CircuitEvo (GPT-3.5-turbo) · Arithmetic1 | init nodes: 79 -> 74 | Improved |
| use an LLM generator instead of a rule-based linear genetic program (ablation) | CircuitEvo (GPT-3.5-turbo) · LogicNets2 | init nodes: 90 -> 74 | Improved |
| use the graph-structured circuit program instead of a Boolean function format (ablation) | CircuitEvo (GPT-3.5-turbo) · LogicNets2 | init nodes: 74 -> 74 | No clear effect |
| add evolutionary strategies (ablation) | CircuitEvo (GPT-3.5-turbo) · Arithmetic3 | init nodes: 1233 -> 1206 | Improved |
| use CircuitEvo (Deepseek-V3) instead of DNAS | DNAS · Initial circuit size, average of 16 circuits | init nodes: 532.88 -> 470.13 | Improved |
| use CircuitEvo instead of DNAS | DNAS · Convergence time, average of 4 circuits | time (h): 16.91 -> 1.48 | Improved |
Closest papers by meaning
- Towards Next-Generation Logic Synthesis: A Scalable Neural Circuit Generation Framework (NeurIPS 2024 · similarity 0.91)
- A Circuit Domain Generalization Framework for Efficient Logic Synthesis in Chip Design (ICML 2024 · similarity 0.90)
- LaMAGIC: Language-Model-based Topology Generation for Analog Integrated Circuits (ICML 2024 · similarity 0.88)
- Retrieval-Guided Reinforcement Learning for Boolean Circuit Minimization (ICLR 2024 · similarity 0.87)
- AUTOCIRCUIT-RL: Reinforcement Learning-Driven LLM for Automated Circuit Topology Generation (ICML 2025 · similarity 0.86)
- PCBSchemaGen: Reward-Guided LLM Code Synthesis for Printed Circuit Boards (PCB) Schematic Design with Structured Verification (NeurIPS 2026 · similarity 0.86)
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.
