constraint-guided model generation

Constraint modeling languages (e.g., MiniZinc, Essence) traditionally perform loop unrolling of quantified expressions and comprehensions at compile time by exhaustively enumerating all combinations of induction variables, then relying on partial evaluation to prune invalid terms (e.g., true conjuncts or zero-valued summands), resulting in substantial redundant computation. This paper proposes a solver-aided loop unrolling method: during compilation, a constraint solver is invoked to precisely infer the set of valid loop body instances that must be instantiated—thereby avoiding full enumeration and post-hoc filtering. The approach integrates constraint solving, partial evaluation, and semantic analysis of quantified expressions, guaranteeing generation of a logically equivalent constraint model while dramatically improving compilation efficiency for high-selectivity cases. Experiments demonstrate significant speedups in compilation time on large-scale problems, achieving, for the first time, semantics-driven selective unrolling at compile time.

Existing fuzzing tools for AI systems exhibit poor generalizability and high rates of invalid inputs when generating test cases for highly structured 3D data (e.g., meshes, point clouds). Method: We propose the first graph-based unified test input generation framework: (1) mapping diverse structured inputs to constraint graphs; (2) designing a neighborhood-similarity-guided graph mutation strategy; and (3) introducing a constraint-driven graph refinement mechanism to jointly enforce structural validity and semantic preservation. The framework supports cross-modal structural modeling, joint structure–semantics verification, and prediction consistency analysis. Results: Evaluated on eight real-world AI systems, our approach achieves up to 2.8× higher structural validity, improves semantic retention by 41.3%, reduces invalid input discard rate by 67.5%, and maintains tractable generation overhead—outperforming baselines including AFL and MeshAttack.

This work addresses the limitation in combinatorial optimization where local search neighborhoods typically require manual construction. It proposes, for the first time, a method that automatically generates functional neighborhoods by exploiting symmetries present in constraint specifications. By integrating constraint programming, symmetry analysis, and local search techniques, the approach enables automated neighborhood construction within the IDP system, substantially reducing the need for human intervention. Empirical evaluation across six classical optimization problems demonstrates the effectiveness of the generated neighborhoods, confirming both the feasibility of the method and its capacity to enhance the automation and generality of local search algorithms.

Streamlining constraints in Constraint Optimization Problems (COPs) rely heavily on manual, problem-specific design, suffering from poor generalizability and scalability. Method: We propose the first LLM-based approach for automatically generating MiniZinc streamlining constraints. Our method integrates prompt engineering, lightweight empirical feedback loops, and test-driven iterative validation to dynamically identify high-performing constraint combinations—mitigating memorization bias and ensuring offline runtime independence and cross-problem generalizability. Contribution/Results: Evaluated on seven representative COP classes, our generated constraints significantly reduce search space size and outperform both human-crafted and systematically constructed baselines in solving speed. Ablation studies—including on adversarial “confused” and “camouflaged” benchmarks—demonstrate robustness and transferability. This work pioneers the use of LLMs for creative, data-efficient streamlining constraint synthesis and establishes a verifiable, reproducible empirical optimization paradigm grounded in rigorous testing and feedback.

Verifying coverage completeness of input generators in property-based testing remains challenging. Method: This paper proposes a static verification approach based on a “must-style” refinement type system, reformulating conventional “may-produce” type semantics into “must-produce” semantics. It formally defines full coverage for higher-order functions and inductive data types, enabling fully automated verification of generator completeness. Contribution/Results: To our knowledge, this is the first refinement type system provably guaranteeing generation of all inputs satisfying both type and constraint specifications. Experimental evaluation demonstrates substantial improvements in detecting coverage gaps across diverse complex generators, while significantly reducing manual verification effort.

This study addresses the high cost and error-proneness of model refactoring caused by paradigm disparities among constraint solvers. We propose a modular automated translation framework based on CPMpy that employs a layered waterfall architecture to uniformly handle sub-expression negation and auxiliary variable generation while optimizing linearization strategies for nonlinear operators. This approach enables seamless translation from high-level models to low-level paradigms, including CP, SMT, and ILP. Experimental results demonstrate that the framework effectively eliminates manual rewriting and that its optimized linearization significantly enhances ILP and PB solving performance. Consequently, this work provides an efficient, flexible, and standardized solution for the automatic benchmarking of multi-paradigm solvers in combinatorial optimization.

This work proposes a novel approach to automatically restructure constraint programming models by leveraging a large language model (LLM)-based autonomous agent operating in an open-ended space. Unlike traditional rule-based reformulation methods, which are constrained by predefined heuristics, the proposed agent iteratively generates, validates, diagnoses, and refines candidate models on training instances, enabling experience-driven, flexible optimization. Integrated with the CPMpy modeling framework and a solution-reinjection validation mechanism, the method demonstrates significant performance gains: across 27 test instances spanning nine combinatorial optimization problems, it outperforms the original formulations on 21 instances, with speedups exceeding two orders of magnitude on certain problems. These results substantially surpass the limitations inherent in conventional rule-driven reformulation techniques.

本文提出CIT-CAD框架，通过构建约束意图树来指导CAD代码生成及验证，解决现有方法忽视设计结构与关系错误的问题。

论文提出约束驱动的上下文工程（CDCE）设计方法，通过识别和操作化约束来改进AI系统在特定领域的适用性和质量。

本文针对PDDL规划中实例生成的难题，提出了一种利用大语言模型自动生成实例生成器的方法，确保了生成实例的有效性和多样性。