ICLR 2026poster0 citations

Boolean Satisfiability via Imitation Learning

Zewei Zhang, Huan Liu, YUANHAO YU, Jun Chen, Xiangyu Xu

Abstract

We propose ImitSAT, a branching policy for conflict-driven clause learning (CDCL) solvers based on imitation learning for the Boolean satisfiability problem (SAT). Unlike previous methods that predict instance-level signals to improve CDCL branching indirectly, or rely on reinforcement learning and insufficient CDCL information to enhance branching, ImitSAT learns from expert KeyTrace that collapses a full run into the sequence of surviving decisions. Replaying a KeyTrace on the same instance is nearly conflict-free, providing dense decision- level supervision and directly reducing propagations—the dominant contributor to wall-clock time. This prefix-conditioned supervision enables ImitSAT to reproduce high-quality branches without exploration, yielding faster convergence, stable training, and seamless integration into CDCL. Extensive experiments demonstrate that ImitSAT reduces propagation counts and runtime, outperforming state-of-the-art learned approaches. We released the source code and trained model at https://github.com/zewei-Zhang/ImitSAT.

Boolean SatisfiabilityImitation LearningAutoregressive ModelingBranching Heuristics
BibTeX
@inproceedings{
zhang2026boolean,
title={Boolean Satisfiability via Imitation Learning},
author={Zewei Zhang and Huan Liu and YUANHAO YU and Jun Chen and Xiangyu Xu},
booktitle={The Fourteenth International Conference on Learning Representations},
year={2026},
url={https://openreview.net/forum?id=LNqWbY5iIf}
}