Revisiting SAT-based Solvers: MaxSAT Rules and Core Sequences

2026-02-20 · Source: Journal of Artificial Intelligence Research · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Mathematics & Computational Sciences · Depth: Advanced, quick

Summary

This paper re-examines the current landscape of MaxSAT solving, specifically focusing on SAT-based Core-guided MaxSAT algorithms. It introduces a novel approach to describe Core-guided solvers using Non-CNF MaxSAT rules combined with an Extension rule. By applying these rules alternatively, the research demonstrates how to derive new Core-guided MaxSAT solvers. The study also explores how these solvers navigate the search space of SAT instance sequences, identifying the potential for exponentially harder sequences and proposing methods to circumvent them. Experimental results indicate that the proposed techniques achieve performance comparable to and complementary with existing state-of-the-art solvers.

Key takeaway

For AI Researchers developing or optimizing MaxSAT solvers, understanding the application of Non-CNF MaxSAT and Extension rules can lead to the creation of more efficient Core-guided algorithms. You should investigate how these rules can be integrated into your current solver architectures to potentially mitigate exponentially harder SAT instance sequences and improve overall performance.

Key insights

New Core-guided MaxSAT solvers can be derived using Non-CNF MaxSAT and Extension rules.

Principles

• Core-guided solvers solve SAT instance sequences.
• Exponentially harder sequences exist and can be avoided.

Method

Describe Core-guided solvers with Non-CNF MaxSAT rules plus the Extension rule, then apply them alternatively to obtain new solvers.

In practice

• Apply Non-CNF MaxSAT rules.
• Utilize the Extension rule.

Topics

• MaxSAT Solving
• Core-guided Solvers
• SAT Instances
• Non-CNF MaxSAT Rules
• Search Space Traversal

Best for: AI Researcher, AI Scientist, Research Scientist

Related on AIssential

• Modal CEGAR-tableaux with RECAR and resolution-based SAT-shortcuts — Integrating KSP as a SAT-shortcut oracle significantly enhances CEGAR-tableaux for modal satisfiability.
• Transfer Learning from Foundational Optimization Embeddings to Unsupervised SAT Representations — Foundational optimization embeddings can transfer to Boolean Satisfiability problems, enabling unsupervised SAT representation learning.
• Automaton Hum Drum — `add_automaton` in OR-Tools CP-SAT can be a "blunt instrument" for complex scheduling, often outperformed by explicit constraints.
• Revisiting a crazy global NLP problem — Non-convex quadratic optimization problems can be intractable even for small instances.
• Scaling Neuro-symbolic Problem Solving: Solver-Free Learning of Constraints and Objectives — A neuro-symbolic architecture with a probabilistic loss efficiently learns to solve NP-hard problems from natural inputs.
• Evaluating Robustness of Reasoning Models on Parameterized Logical Problems — A new 2-SAT benchmark reveals LLM reasoning brittleness under targeted structural changes, not just surface difficulty.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Journal of Artificial Intelligence Research.