Research map / Graph learning / Graph algorithms and clustering
G2SAT: Learning to Generate SAT Formulas
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Learn the distribution of clause-node merges from input SAT formulas | SAT formula generation · Training SAT formulas | average relative error across graph statistics: 24% lower relative error | Helped |
| Generate formulas with G2SAT rather than PS | SAT solver behavior on synthetic formulas · Training SAT formulas | solver ranking accuracy: 33% -> 100% | Helped |
| Tune Glucose on G2SAT-generated formulas rather than the training formulas | Glucose · 22 unseen real-world SAT formulas | runtime gain: 18.25% | Helped |
| Tune Glucose using CA-generated formulas | Glucose · 22 unseen real-world SAT formulas | runtime gain: 2.31% | Helped |
| Tune Glucose using PS (T=0)-generated formulas | Glucose · 22 unseen real-world SAT formulas | runtime gain: 0.41% | Helped |
| Tune Glucose using PS (T=1.5)-generated formulas | Glucose · 22 unseen real-world SAT formulas | runtime gain: 0.07% | Helped |
| Increase the number of GCN layers (ablation) | G2SAT · Generated SAT formulas | modularity similarity to training formulas | Helped |
| Generate formulas larger than the training formulas | G2SAT · Formulas with 13,028 to 27,360 clauses | VCG modularity: 0.81 ± 0.03 | Helped |
Closest papers by meaning
- On the Expressive Power of GNNs for Boolean Satisfiability (ICLR 2026 · similarity 0.91)
- MixSATGEN: Learning Graph Mixing for SAT Instance Generation (ICLR 2024 · similarity 0.91)
- HardCore Generation: Generating Hard UNSAT Problems for Data Augmentation (NeurIPS 2024 · similarity 0.90)
- NeuroBack: Improving CDCL SAT Solving using Graph Neural Networks (ICLR 2024 · similarity 0.87)
- Learning How to Cube (NeurIPS 2026 · similarity 0.87)
- Learning Better Representations From Less Data For Propositional Satisfiability (NeurIPS 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.