Research map / Reinforcement learning / RL from feedback
SLIC: Reinforcement Fine-Tuning Small LMs for Multi-Turn Analog Circuit Optimization
Yan Xu, Zhengqi Gao, Ching-Yun Ko, Zhang-Wei Hong, Wenjie Lu, Tao Yu, Xin Zhang, Duane Boning, Ruonan Han
Read the paperCheck your idea against 137,217 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Proposes SLIC, a reinforcement fine-tuning framework that trains a locally deployable 7B-scale small LM to iteratively optimize analog circuits in a simulator-grounded multi-turn loop.
Problem
Analog IC design is a multi-turn, simulator-grounded task requiring iterative reasoning over coupled performance trade-offs, but proprietary PDKs make hosted LLM workflows hard to deploy.
Approach
Reinforcement fine-tuning of a locally deployable 7B-scale language model for multi-turn analog circuit optimization, so that the model can run on-premises without sending proprietary design data to hosted LLMs.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add domain-validity reward gates (ablation) | SLIC-Random · ID, Turn 8 | Pass@1 (%): 82.8 -> 87.1 | Improved |
| add domain-validity reward gates (ablation) | SLIC-Random · held-out, Turn 8 | Pass@1 (%): 63.6 -> 70.6 | Improved |
| add success-conditional slack bonus (ablation) | SLIC-Random + filter · ID, Turn 8 | Pass@1 (%): 76.4 -> 84.9 | Improved |
| add success-conditional slack bonus (ablation) | SLIC-Random + filter · held-out, Turn 8 | Pass@1 (%): 56.7 -> 68.8 | Improved |
| normalize validity and performance advantages separately (ablation) | SLIC-Random · ID, Turn 8 | Pass@1 (%): 79.0 -> 85.2 | Improved |
| use random or softmax continuation instead of greedy (ablation) | SLIC RGRS · held-out, Turn 8 | Pass@1 (%): 43.9 -> 70.6 (Random), 72.6 (Softmax) | Improved |
| use softmax instead of random continuation (ablation) | SLIC RGRS · held-out, Turn 8 | Pass@1 (%): 70.6 -> 72.6 | No clear effect |
| use online RGRS construction instead of fixed-base TL-GRPO (ablation) | SLIC-Random + filter · ID and held-out | Pass@1 (%): ID 73.8 -> 76.4; OOD 57.6 -> 56.7 | Mixed |
| train multi-turn (RGRS) instead of single-turn GRPO (ablation) | SLIC · ID, Turn 8 | Pass@1 (%): 61.6 -> 87.1 | Improved |
| use RGRS instead of Tree MT-GRPO | SLIC-Random · held-out, Turn 8 | Pass@1 (%): 58.3 -> 70.6 | Improved |
| train with four-turn horizon instead of eight (ablation) | SLIC-Random · ID and held-out, Turn 8 | Pass@1 (%): ID 87.1 -> 89.5; OOD 70.6 -> 69.3 | No clear effect |
| keep raw model outputs in history instead of structured history (ablation) | SLIC-Random · ID, Turn 8 | Pass@1 (%): 87.1 -> 76.8 | Hurt |
Closest papers by meaning
- AUTOCIRCUIT-RL: Reinforcement Learning-Driven LLM for Automated Circuit Topology Generation (ICML 2025 · similarity 0.88)
- LaMAGIC: Language-Model-based Topology Generation for Analog Integrated Circuits (ICML 2024 · similarity 0.87)
- FALCON: An ML Framework for Fully Automated Layout-Constrained Analog Circuit Design (NeurIPS 2025 · similarity 0.87)
- AnalogCoder: Analog Circuit Design via Training-Free Code Generation (AAAI 2025 · similarity 0.87)
- AnalogGenie: A Generative Engine for Automatic Discovery of Analog Circuit Topologies (ICLR 2025 · similarity 0.86)
- Retrieval-Guided Reinforcement Learning for Boolean Circuit Minimization (ICLR 2024 · 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.
