Lune

ICLR2025Top-tier venue

A Graph Enhanced Symbolic Discovery Framework For Efficient Logic Optimization

Yinqi Bai, Jie Wang, Lei Chen, Zhihai Wang, Yufei Kuang, Mingxuan Yuan, Jianye Hao, Feng Wu

Abstract

The efficiency of Logic Optimization (LO) has become one of the key bottlenecks in chip design. To prompt efficient LO, previous studies propose using a key scoring function to predict and prune a large number of ineffective nodes of the LO heuristics. However, the existing scoring functions struggle to balance inference efficiency, interpretability, and generalization performance, which severely hinders their application to modern LO tools. To address this challenge, we propose a novel data-driven circuit symbolic learning framework, namely CMO, to learn lightweight, interpretable, and generalizable scoring functions. The major challenge of developing CMO is to discover symbolic functions that can well generalize to unseen circuits, i.e., the circuit symbolic generalization problem. Thus, the major technical contribution of CMO is the novel Graph Enhanced Symbolic Discovery framework, which distills dark knowledge from a well-designed graph neural network (GNN) to enhance the generalization capability of the learned symbolic functions. To the best of our knowledge, CMO is the first graph-enhanced approach for discovering lightweight and interpretable symbolic functions that can well generalize to unseen circuits in LO. Experiments on three challenging circuit benchmarks show that the interpretable symbolic functions learned by CMO outperform previous state-of-the-art (SOTA) GPU-based and human-designed approaches in terms of inference efficiency and generalization capability. Moreover, we integrate CMO with the Mfs2 heuristic-one of the most time-consuming LO heuristics. The empirical results demonstrate that CMO significantly improves its efficiency while keeping comparable optimization performance when executed on a CPU-based machine, achieving up to 2.5× faster runtime.

Ask Lune

Know this paper in minutes, not an afternoon.

Lune can read every page of it and the top-tier papers that cite it, and gives you its claims, its evidence and its limits, each one cited to the paper it comes from. No skimming, no AI slop.

4 top-tier papers have cited it since. Know what they found before your reviewers do.

Questions to start from

Lune can draw on

the full paper · the top-tier papers that cite it · related work

Get the answer

Free to start. No credit card required.

Lineage

  • Evolving Graph Structured Programs for Circuit Generation with Large Language Models cites Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients
  • Evolving Graph Structured Programs for Circuit Generation with Large Language Models cites A Hierarchical Circuit Symbolic Discovery Framework for Efficient Logic Optimization
  • A Hierarchical Circuit Symbolic Discovery Framework for Efficient Logic Optimization cites Evolving Graph Structured Programs for Circuit Generation with Large Language Models
  • End-to-end Symbolic Regression with Transformers cites Discovering Symbolic Models from Deep Learning with Inductive Biases
  • End-to-end Symbolic Regression with Transformers cites Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients
  • RulePlanner: All-in-One Reinforcement Learner for Unifying Design Rules in 3D Floorplanning cites Evolving Graph Structured Programs for Circuit Generation with Large Language Models
  • A Hierarchical Circuit Symbolic Discovery Framework for Efficient Logic Optimization cites Discovering Symbolic Models from Deep Learning with Inductive Biases
  • A Hierarchical Circuit Symbolic Discovery Framework for Efficient Logic Optimization cites Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradients
  • A Hierarchical Circuit Symbolic Discovery Framework for Efficient Logic Optimization cites End-to-end Symbolic Regression with Transformers
Builds onCited byLeft to right by year; the more cited, the higher.

Cited by top-tier papers4

  1. Opt-Miner: Empowering Information-Seeking Agent with Tree-Guided Data Synthesis for Optimization ModelingHaoyang Liu, Yuyang Cai, Jie Wang, Xiongwei Han et al.ICML 2026
  2. Evolving Graph Structured Programs for Circuit Generation with Large Language ModelsYinqi Bai, Jie Wang, Lei Chen, Zhihai Wang et al.ICLR 2026
  3. RulePlanner: All-in-One Reinforcement Learner for Unifying Design Rules in 3D FloorplanningRuizhe Zhong, Xingbo Du, Junchi YanICML 2026
  4. A Hierarchical Circuit Symbolic Discovery Framework for Efficient Logic OptimizationYinqi Bai, Jie Wang, Xialiang Tong, Longdi Pan et al.ICLR 2026

Builds on25

  1. Discovering Symbolic Models from Deep Learning with Inductive BiasesMiles D. Cranmer, Alvaro Sanchez-Gonzalez, Peter W. Battaglia, Rui Xu et al.NeurIPS 2020 · 736 citations
  2. Deep symbolic regression: Recovering mathematical expressions from data via risk-seeking policy gradientsBrenden K. Petersen, Mikel Landajuela, T. Nathan Mundhenk, Cláudio Prata Santiago et al.ICLR 2021 · 444 citations
  3. End-to-end Symbolic Regression with TransformersPierre-Alexandre Kamienny, Stéphane d'Ascoli, Guillaume Lample, François ChartonNeurIPS 2022 · 320 citations
  4. Graph-less Neural Networks: Teaching Old MLPs New Tricks Via DistillationShichang Zhang, Yozen Liu, Yizhou Sun, Neil ShahICLR 2022 · 234 citations
  5. Sample-Efficient Reinforcement Learning via Conservative Model-Based Actor-CriticZhihai Wang, Jie Wang, Qi Zhou, Bin Li et al.AAAI 2022 · 38 citations

Related papers

  1. Towards General Algorithm Discovery for Combinatorial Optimization: Learning Symbolic Branching Policy from Bipartite GraphYufei Kuang, Jie Wang, Yuyan Zhou, Xijun Li et al.ICML 2024 · 4 citations
  2. Functionality matters in netlist representation learningZiyi Wang, Chen Bai, Zhuolun He, Guangliang Zhang et al.DAC 2022 · 45 citations
  3. Rethinking Branching on Exact Combinatorial Optimization Solver: The First Deep Symbolic Discovery FrameworkYufei Kuang, Jie Wang, Haoyang Liu, Fangzhou Zhu et al.ICLR 2024 · 15 citations
  4. A Circuit Domain Generalization Framework for Efficient Logic Synthesis in Chip DesignZhihai Wang, Lei Chen, Jie Wang, Yinqi Bai et al.ICML 2024 · 13 citations
  5. Gamora: Graph Learning based Symbolic Reasoning for Large-Scale Boolean NetworksNan Wu, Yingjie Li, Cong Hao, Steve Dai et al.DAC 2023 · 35 citations
Dusk over the sea between two cliffs drawn in fine vertical lines