# SAT-ACT: Learning CDCL Branching from Outcome-Derived Action Preferences

DevFeed: [SAT-ACT: Learning CDCL Branching from Outcome-Derived Action Preferences](<https://devfeed.tech/articles/sat-act-learning-cdcl-branching-from-outcome-derived-action-preferences-111566.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2610.41290>)

Author: Zhenchao Sun, Chongyang Tao, Ping Lu, Tongkai Lu, Jianling Gao, Shuai Ma

Published: 2026-11-02T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>), [arXiv: Machine Learning](<https://devfeed.tech/sources/arxiv-machine-learning.md>)

Topics: [Machine learning](<https://devfeed.tech/topics/machine-learning.md>), [Neural Network](<https://devfeed.tech/topics/neural-network.md>), [Reinforcement learning](<https://devfeed.tech/topics/reinforcement-learning.md>), [decision-making](<https://devfeed.tech/topics/decision-making.md>)

Tags: [benchmarks](<https://devfeed.tech/tags/benchmarks.md>), [comparison](<https://devfeed.tech/tags/comparison.md>), [competition](<https://devfeed.tech/tags/competition.md>), [costs](<https://devfeed.tech/tags/costs.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [cs-lg](<https://devfeed.tech/tags/cs-lg.md>), [experiments](<https://devfeed.tech/tags/experiments.md>), [feedback](<https://devfeed.tech/tags/feedback.md>), [learning](<https://devfeed.tech/tags/learning.md>), [network](<https://devfeed.tech/tags/network.md>), [search](<https://devfeed.tech/tags/search.md>), [state](<https://devfeed.tech/tags/state.md>), [static](<https://devfeed.tech/tags/static.md>), [supervision](<https://devfeed.tech/tags/supervision.md>), [training](<https://devfeed.tech/tags/training.md>)

## AI overview

SAT-ACT learns CDCL branching priorities from preferences derived by comparing independent action rollouts from the same solver state. Its three outcome-comparison criteria preserve unresolved comparisons when evidence is insufficient or conflicting, rather than imposing a ranking through a single predefined metric. A state-conditioned network then guides search without online rollouts. The method also keeps solver-selected actions as training targets only when they remain non-dominated under the learned preferences. The abstract reports reduced CDCL search costs on three SAT families and improved PAR-2 on SAT Competition benchmarks; it provides no experimental details for assessing the size or scope of those gains.

## Source excerpt

arXiv:2610.41290v1 Announce Type: cross Abstract: Learning-based approaches have increasingly been explored for SAT solving, with particular attention to improving branching decisions in conflict-driven clause learning (CDCL) solvers. These approaches derive branching guidance either from static structural properties of SAT formulas or from dynamic signals obtained through solver feedback and execution traces. However, these forms of supervision provide limited evidence about state-dependent branching effects, while different outcome measures may favor different branching choices. To address these limitations, we propose SAT-ACT, an outcome-derived learning framework for CDCL branching. We design three complementary outcome comparison criteria to extract action preferences from independent branching action rollouts initiated at the same solver state. The criteria identify preferences supported by the outcomes and leave comparisons unresolved when the evidence is insufficient or conflicting, instead of forcing an ordering using a predefined scalar metric. A state-conditioned network learns branching priorities from the extracted preferences and uses them to guide CDCL search directly without online action rollouts. We further introduce non-dominated heuristic action supervision, which retains the solver-selected action as a training target only when it remains non-dominated under these preferences. Solver-level experiments show that SAT-ACT consistently reduces CDCL search costs on three SAT families and improves PAR-2 on the SAT Competition benchmarks.