Pith. sign in

REVIEW 4 major objections 5 minor 25 references

LLM-Guided Graph Generation for Structure-Based Local Improvement Methods

T0 review · 4 major / 5 minor · reviewed 2026-08-14 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper claims that an LLM, guided by semantic rules, can turn any MiniZinc constraint model into a uniform weighted graph that lets a generic local-improvement framework beat a one-shot exact solver on 44.0% of problem-weighted…

desk verdict A genuinely new LLM-to-SLIM pipeline with clear writing and honest evaluation, but the central semantic-weight claim lacks a random-selection control and should be tightened before acceptance. read the letter →

arxiv 2608.13333 v1 pith:OJFGO247 submitted 2026-08-13 cs.AI

Hai Xia , Vaidyanathan Peruvemba Ramaswamy , Stefan Szeider This is my paper · ORCID
classification cs.AI
keywords largeneighborhoodsearchstructure-basedlocalimprovementLLM-guidedgraphgenerationweightedconstraintMiniZincalgorithmselectionvariableoptimization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Large neighborhood search normally needs problem-specific expert knowledge to decide which variables to optimize together. This paper claims that a large language model, prompted with a few semantic guidelines, can read a MiniZinc constraint model and emit a graph generator that maps every instance of that problem to one uniform weighted graph, with nodes for decision variables and edges for constraint couplings. That single graph lets a generic structure-based local-improvement framework (SLIM) select neighborhoods without knowing the problem, and lets one algorithm selector choose among 30 SLIM configurations using the same 54 graph features across all problems. On 20 MiniZinc competition problems, the reported average problem-weighted win rate against a one-shot exact solver is 39.5%, rising to 44.0% after configuration and feature ablation, versus 19.3% for the best fixed configuration. If the claim holds, it replaces hand-built per-problem neighborhood design with a one-time automated semantic compilation step.

What carries the argument

The load-bearing object is the uniform weighted graph generated by the LLM's Python program. Nodes represent decision variables with a weight $w\in[0,1]$ (objective importance) and domain size $d$; edges represent constraint relations with a coupling weight $w\in[0,1]$, combined across parallel constraints with $W=1-\prod_i(1-w_i)$ so no weight exceeds 1. Semantic guidelines push objective-related nodes to weights at or above 0.6, assign large global constraints the weight $\max(0.1,1/n)$, forbid isolated nodes, and bound all weights. This object carries problem structure into two generic consumers: BFS and weighted-random variable extraction for SLIM, and 54 topology/statistics features for configuration selection.

What would settle it

On one problem type, compute a ground-truth variable-importance ranking by perturbing each decision variable individually and measuring the objective change, then compare it with the LLM-assigned node weights and with SLIM's selection frequency; if these do not correlate, the semantic weights are not the active mechanism.

Watch

Extended reading notes

Core claim

The paper's central claim is that LLM-produced graph generators, guided by semantic weighting rules, capture enough constraint structure to drive generic local improvement across heterogeneous optimization problems. Each node in the produced graph carries an importance weight in [0,1] and a domain size; each edge carries a coupling strength, aggregated by $W=1-\prod_i(1-w_i)$ and bounded at 1, with large global constraints down-weighted to $\max(0.1,1/n)$ to avoid clique domination. Because the representation is uniform, BFS-based and weighted-random extraction methods operate identically on every instance, and the same 54 graph features feed a cross-problem configuration selector trained with problem-weighted samples. The decisive evidence is the win-rate comparison against a one-shot Gurobi baseline: the best single SLIM configuration reaches 19.3%, algorithm selection reaches 39.5%, and ablations raise it to 44.0%.

Load-bearing premise

Everything downstream depends on the LLM's graph weights and edge strengths faithfully reflecting which variables and constraint relationships actually matter for improving the objective; if the generator misreads the model, the graph drives variable selection toward the wrong parts of the instance.

Editorial extensions

If this is right

  • A new MiniZinc problem type only needs one LLM synthesis pass over its model, typically minutes, before the same generic SLIM framework and selector can be applied to its instances.
  • Because the graph format is uniform, the same 54 features support algorithm selection across different problem types, so portfolio design no longer requires per-domain feature engineering.
  • The node and edge weights give a principled, automatically derived ranking of which variables and couplings matter, making structure-aware neighborhood selection available for any declarative optimization model.
  • Configuration choice matters: the paper's 19.3%-versus-44.0% gap shows that the same generic SLIM with a learned configuration is much stronger than any fixed strategy, making selection a first-class component of the method.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the central claim holds, the semantic-compilation step could generalize beyond MiniZinc: any constraint language an LLM can parse could be mapped to the same weighted-graph format, extending structure-aware large neighborhood search to satisfaction and hybrid problems.
  • A direct mechanism test would compare LLM-assigned node and edge weights with an empirical sensitivity ranking obtained by perturbing each variable and measuring objective change; a strong correlation would confirm the semantic weights, not just graph topology, carry the signal.
  • The paper's closing suggestion of adaptive weight updates during search is an untested extension: if weights change over iterations, the 54 selector features would need to be recomputed or made time-aware, turning the pipeline into a closed-loop optimizer.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 5 minor

Summary. The paper proposes an automated pipeline in which an LLM, prompted with human-written semantic guidelines, generates a Python graph generator for each MiniZinc problem type. The generator maps any instance of that type to a uniform weighted graph: nodes are decision variables with objective-importance weights and domain sizes, and edges are constraint relations with coupling strengths. A generic structure-based local improvement framework (SLIM) uses this graph for variable selection via BFS-based and LNS-based extraction, and the same graph format supplies 54 features for cross-problem algorithm configuration selection among 30 SLIM configurations. The authors evaluate on 20 MiniZinc competition problem types satisfying an instance filter, reporting that the best single configuration achieves a 19.3% average problem-weighted win rate against a one-shot Gurobi baseline, algorithm selection achieves 39.5%, and configuration/feature ablation boosts this to 44.0%.

Significance. If the central causal claim is supported, this is a significant contribution: it would replace months of problem-specific SLIM engineering with a one-time LLM compilation step and would enable portfolio-style algorithm selection across heterogeneous problem types under a single feature representation. The paper has clear strengths: a held-out evaluation split stratified by problem, problem-weighted metrics that prevent the RCPSP majority from dominating, multiple random seeds, and a Zenodo artifact for reproducibility. The approach also produces deterministic, auditable generators rather than treating the LLM as a runtime oracle. However, the significance depends on whether the LLM-generated semantic weights are actually responsible for the reported gains, and the current experiments do not isolate that mechanism.

major comments (4)
  1. [Section 4.2 and Section 3.3.1] Section 4.2 explicitly states that SLIM with Gurobi as the subsolver 'becomes standard LNS when the variable extraction is random and does not utilize the uniform weights,' yet no random-selection or unweighted-graph condition is reported in Figure 2, Table 1, or the supplementary ablation. Because the reported win rates are compared only against one-shot Gurobi, they do not isolate the contribution of the LLM-assigned semantic weights; the gains could be produced entirely by the LNS loop, the subsolver choice, or the budget structure. Please add controls that remove the semantic signal—for example, random variable extraction, uniform (constant) node and edge weights, and a syntactic variable–constraint graph without LLM weights—run with the same 30-configuration grid, and report their per-problem and problem-weighted win rates as well as how the algorithm selector behaves on these controls.
  2. [Section 4.1 and Abstract] The instance filter in Section 4.1 requires that Gurobi finds a feasible solution within 10 minutes but does not prove optimality within 60 minutes, and that at least five instances qualify per problem. These criteria select exactly the instances where the one-shot baseline is weakest and exclude entire problem types, so the abstract's claim that the pipeline is 'problem-agnostic to all problems in the MiniZinc format' is not supported by the evaluation. Please either restrict the generalization claim to the 20 selected problem types and their qualifying instances, or report results over a larger, less filtered benchmark together with a sensitivity analysis of the filter thresholds.
  3. [Section 4.3 and Section 3.4] The paper reports 39.5% as the algorithm-selection win rate and 44.0% after configuration and feature ablation, but Section 4.3 states that 'for the best-performing approach on each seed' the win rates are 37.9–40.6%, and the 44.0% comes from greedy backward elimination over configurations and features described in Section 3.4 and the supplementary material. Selecting the best approach per seed and then eliminating configurations/features using cross-validation before reporting a test-set win rate is a repeated model-selection procedure that can inflate the reported average; Table 1 also lacks standard errors or confidence intervals. Please provide nested cross-validation estimates, report per-seed standard errors, and give a precise statement of how the 39.5% and 44.0% headline numbers are aggregated across seeds, problems, and ablation steps.
  4. [Section 5 and Section 3.2] Section 5 concedes that the generated graphs 'remain LLM-produced approximations of the true constraint semantics,' but the paper does not describe how generators are validated or how faithful the graphs are to the original MiniZinc models. Since all downstream extraction and feature computation in Sections 3.3 and 3.4 depends on the correctness of node and edge weights, please add a per-problem validation protocol (for example, checks against the model text, statistics on missed constraints or objective terms, and a sensitivity analysis over the semantic guidelines) and report the resulting error rates.
minor comments (5)
  1. [Section 4.2, Figure 2] Figure 2 says the results are 'averaged across 3 seeds,' but no per-seed spread is shown; please add seed-level error bars or a table with the per-seed values.
  2. [Section 3.2.2] The edge aggregation formula W=1−∏(1−w_i) is given without an equation number or a definition of the index set; please number it and state explicitly that the product runs over all constraints linking the same pair of variables.
  3. [Section 3.3.1] The budget b is described only as 'a threshold bounded by a domain-size-aware metric'; please give the exact formula or point to a numbered equation in the supplementary material.
  4. [Title page] The author block contains formatting errors, including 'V aidyanathan' with an erroneous space and a line break in the name; please fix the LaTeX source.
  5. [Section 1 and Table 1] Section 1 reports the best single configuration as 19.3% while Table 1 lists 17.8, 17.7, and 22.3% for the three seeds; please state explicitly that 19.3% is the seed average and reconcile the 44.0% ablation number with the per-seed values in the supplementary material.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the reported win rates are out-of-sample empirical results, and the LLM graph generators and algorithm selector are not fitted to the final test outcomes by construction.

full rationale

This paper is an empirical pipeline evaluation rather than a formal derivation, and its central claims rest on held-out benchmark comparisons rather than on fitted quantities being re-presented as predictions. The LLM graph generators are produced from human-written semantic guidelines in Section 3.2.2, such as 'Objective-related components should have higher weights (>=0.6)', and are not fitted to the reported win rates against Gurobi. The algorithm selector is trained on per-configuration improvement margins from the training split and evaluated on a 70:30 stratified held-out test split, as stated in Section 4.1, so the 39.5%-44.0% problem-weighted win rates are genuine out-of-sample predictions and are not definitionally equal to the training labels. The paper's prior SLIM papers are cited as background methodology for a general local-improvement framework, not as a self-referential uniqueness theorem and not as the justification that the LLM output is correct. The limitation statement in Section 5 concedes that the generators 'remain LLM-produced approximations of the true constraint semantics', and the paper notes in Section 4.2 that random extraction without the uniform weights would be standard LNS while reporting no such random-extraction control. That missing baseline is a possible threat to the causal attribution of the gains to semantic weighting, but it is an experimental-design and correctness-risk concern, not a circularity, because none of the reported numbers is equivalent to the pipeline's inputs by construction.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

The quantitative claims rest on a chain of empirical choices rather than a formal derivation: hand-set graph-weight guidelines, a fixed grid of 30 SLIM configurations, an instance filter, a hand-designed 54-feature set, and the unverified-in-the-text assumption that LLM-generated graphs approximate true semantics. None of these is fitted to the target win rates in a closed-form sense, but each shapes the reported numbers.

free parameters (6)
  • Objective node weight threshold = w >= 0.6
    Guideline 1 in Section 3.2.2; hand-chosen threshold controlling how strongly objective-related variables are prioritized during extraction.
  • Large constraint edge weight rule = max(0.1, 1/n) for n > 10
    Guideline 2 in Section 3.2.2; hand-chosen to prevent clique domination, affects every edge weight in the graph.
  • Small constraint edge weight = w >= 0.8 for n <= 5
    Guideline 2 in Section 3.2.2; hand-set coupling strength for small constraints.
  • Edge weight aggregation rule = W = 1 - product_i (1 - w_i)
    Guideline 4 in Section 3.2.2; bounded aggregation of multiple edges between the same variables.
  • SLIM configuration grid = b in {10,20,50,70,100,200}; t in {20,30,45,60}s; BFS/LNS with freeze modes
    Section 3.3; the 30 configurations are the search space for both the best-single-config baseline and the algorithm selector.
  • Instance selection thresholds = feasible within 10 min; not optimal within 60 min; at least 5 qualifying instances
    Section 4.1; this filter defines the benchmark and removes problems or instances where the comparison is undefined or dominated by easy cases.
assumptions (5)
  • domain assumption MiniZinc models and instances are faithful descriptions of the target optimization problems.
    The entire pipeline consumes .mzn and .dzn files; if a model misrepresents the problem, the resulting graph and local search are solving the wrong thing.
  • ad hoc to paper LLM-produced graph generators approximate true constraint semantics well enough to guide effective search.
    This is the central enabling premise, explicitly flagged in Section 5: 'the generators, while validated, remain LLM-produced approximations of the true constraint semantics.' If this fails, the pipeline degrades to unweighted random selection.
  • domain assumption Gurobi as subsolver and as one-shot baseline is an appropriate reference for measuring improvement.
    All win rates are relative to one-shot Gurobi under the same 60-minute budget; a different baseline (e.g., a tuned MIP focus or another solver) would change every headline number.
  • domain assumption The 54 hand-designed generic graph features carry enough signal for cross-problem algorithm selection.
    Section 3.4 relies on these features for the random forest selector; if they were uninformative, the selector would not beat the best single configuration.
  • domain assumption Problem-weighted evaluation is the correct aggregation across imbalanced problem types.
    RCPSP contributes 38% of instances; the paper equalizes per-problem weight, and this weighting directly determines the reported win rates.

how reviews work

0 comments
Cite this review

Pith. "Pith review of LLM-Guided Graph Generation for Structure-Based Local Improvement Methods." pith.science (2026). https://pith.science/paper/OJFGO247

@misc{pith2026260813333,
  author       = {Pith},
  title        = {Pith review of: LLM-Guided Graph Generation for Structure-Based Local Improvement Methods},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OJFGO247}},
  note         = {Machine review of arXiv:2608.13333}
}
read the original abstract

Large neighborhood search normally selects a random subset of decision variables for iterative optimization. For efficiently solving different problems, researchers tend to design variable selection strategies by taking into account structural features from different domains. In this paper, we build an automatic pipeline that is problem-agnostic to all problems in the MiniZinc format. By prompting an LLM with our semantic guidelines, we guide the LLM to produce a graph generator that maps any instance of a problem type to a uniform weighted graph, where nodes represent decision variables and edges represent constraint relationships. These problem-agnostic graphs guide our structure-based local improvement framework (SLIM) in variable selection. Meanwhile, the weighted graph enables all problem instances to share the same generic graph representation, from which the same graph features can be extracted and used for configuration selection. We evaluated our pipeline on instances across 20 MiniZinc competition problems, finding that algorithm selection achieves a 39.5% average problem-weighted win rate against a one-shot Gurobi baseline, more than doubling the best single configuration (19.3%). Configuration and feature ablation boost the performance further to 44.0%, demonstrating that LLM-based semantic generation enables effective automated structure extraction and feature extraction for constraint optimization.

Figures

Figures reproduced from arXiv: 2608.13333 by the authors.

Figure 1
Figure 1. Pipeline overview. Programs (solid border) process files (dashed border). The blue region shows the one-time generator creation by the LLM, and the orange region shows per-instance processing. The instance files (.mzn + .dzn) feed the Variable Extractor, the Graph Generator, and SLIM (which solves subproblems on the original model). The dashed arrow indicates that generator.py is instan￾tiated as the Graph Generator… view at source ↗
Figure 2
Figure 2. SLIM (best configuration per problem type) vs one-shot Gurobi baseline, averaged across 3 seeds. Problem types are sorted by win rate. Instance counts per type are shown in parentheses. Even though SLIM failed to get any better solutions on the filters and grid-coloring problems, on all instances of these two problems SLIM can get the same good solutions as one-shot Gurobi. In general, when SLIM is equipped with sui… view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

25 extracted references · 14 canonical work pages

  1. [1]

    SUNNY: a lazy portfolio approach for constraint solving

    Roberto Amadini, Maurizio Gabbrielli, and Jacopo Mauro. SUNNY: a lazy portfolio approach for constraint solving. Theory Pract. Log. Program., 14 0 (4-5): 0 509--524, 2014. doi:10.1017/S1471068414000179. URL https://doi.org/10.1017/S1471068414000179

    work page doi:10.1017/s1471068414000179 2014

  2. [2]

    Large language model meets constraint propagation

    Alexandre Bonlarron, Florian R \' e gin, Elisabetta De Maria, and Jean - Charles R \' e gin. Large language model meets constraint propagation. In Proceedings of the Thirty-Fourth International Joint Conference on Artificial Intelligence, IJCAI 2025, Montreal, Canada, August 16-22, 2025 , pages 10036--10044. ijcai.org, 2025. doi:10.24963/IJCAI.2025/1115. ...

    work page doi:10.24963/ijcai.2025/1115 2025

  3. [3]

    Tree clustering for constraint networks

    Rina Dechter and Judea Pearl. Tree clustering for constraint networks. Artif. Intell., 38 0 (3): 0 353--366, 1989. doi:10.1016/0004-3702(89)90037-4. URL https://doi.org/10.1016/0004-3702(89)90037-4

    work page doi:10.1016/0004-3702(89)90037-4 1989

  4. [4]

    SAT -based local improvement for finding tree decompositions of small width

    Johannes Klaus Fichte, Neha Lodha, and Stefan Szeider. SAT -based local improvement for finding tree decompositions of small width. In Serge Gaspers and Toby Walsh, editors, Theory and Applications of Satisfiability Testing - SAT 2017 - 20th International Conference, Melbourne, VIC, Australia, August 28 - September 1, 2017, Proceedings , volume 10491 of L...

    doi:10.1007/978-3-319-66263-3 2017

  5. [5]

    Exact combinatorial optimization with graph convolutional neural networks

    Maxime Gasse, Didier Ch \' e telat, Nicola Ferroni, Laurent Charlin, and Andrea Lodi. Exact combinatorial optimization with graph convolutional neural networks. In Hanna M. Wallach, Hugo Larochelle, Alina Beygelzimer, Florence d'Alch \' e - Buc, Emily B. Fox, and Roman Garnett, editors, Advances in Neural Information Processing Systems 32: Annual Conferen...

    work page 2019

  6. [6]

    Hypertree decompositions and tractable queries

    Georg Gottlob, Nicola Leone, and Francesco Scarcello. Hypertree decompositions and tractable queries. J. Comput. Syst. Sci., 64 0 (3): 0 579--627, 2002. doi:10.1006/JCSS.2001.1809. URL https://doi.org/10.1006/jcss.2001.1809

    arXiv 2002

  7. [7]

    Neural large neighborhood search for routing problems

    Andr \' e Hottung and Kevin Tierney. Neural large neighborhood search for routing problems. Artif. Intell., 313: 0 103786, 2022. doi:10.1016/J.ARTINT.2022.103786. URL https://doi.org/10.1016/j.artint.2022.103786

    arXiv 2022

  8. [8]

    GRAPH reinforcement learning for operator selection in the ALNS metaheuristic

    Syu - Ning Johnn, Victor - Alexandru Darvariu, Julia Handl, and J \" o rg Kalcsics. GRAPH reinforcement learning for operator selection in the ALNS metaheuristic. In Bernab \' e Dorronsoro, Francisco Chicano, Gr \' e goire Danoy, and El - Ghazali Talbi, editors, Optimization and Learning - 6th International Conference, OLA 2023, Malaga, Spain, May 3-5, 20...

    work page doi:10.1007/978-3-031-3402 2023

Show all 25 references
  1. [9]

    Hoos, Frank Hutter, and Torsten Schaub

    Marius Lindauer, Holger H. Hoos, Frank Hutter, and Torsten Schaub. Autofolio: An automatically configured algorithm selector. J. Artif. Intell. Res., 53: 0 745--778, 2015. doi:10.1613/JAIR.4726. URL https://doi.org/10.1613/jair.4726

    2015 doi

  2. [10]

    A SAT approach to branchwidth

    Neha Lodha, Sebastian Ordyniak, and Stefan Szeider. A SAT approach to branchwidth. ACM Trans. Comput. Log. , 20 0 (3): 0 15:1--15:24, 2019. doi:10.1145/3326159. URL https://doi.org/10.1145/3326159

    2019 doi

  3. [11]

    Learning large Bayesian networks with expert constraints

    Vaidyanathan Peruvemba Ramaswamy and Stefan Szeider. Learning large Bayesian networks with expert constraints. In James Cussens and Kun Zhang, editors, Proceedings of the Thirty-Eighth Conference on Uncertainty in Artificial Intelligence, UAI 2022 , volume 180 of Proceedings o...

    2022

  4. [12]

    Large neighborhood search

    David Pisinger and Stefan Ropke. Large neighborhood search. In Michel Gendreau and Jean - Yves Potvin, editors, Handbook of Metaheuristics, pages 99--127. Springer International Publishing, Cham, 2019. ISBN 978-3-319-91086-4. doi:10.1007/978-3-319-91086-4_4. URL https://doi.or...

    2019 doi

  5. [13]

    MaxSAT-Based postprocessing for treedepth

    Vaidyanathan Peruvemba Ramaswamy and Stefan Szeider. MaxSAT-Based postprocessing for treedepth. In Helmut Simonis, editor, Principles and Practice of Constraint Programming - 26th International Conference, CP 2020, Louvain-la-Neuve, Belgium, September 7-11, 2020, Proceedings ,...

    2020 doi

  6. [14]

    Turbocharging treewidth-bounded bayesian network structure learning

    Vaidyanathan Peruvemba Ramaswamy and Stefan Szeider. Turbocharging treewidth-bounded bayesian network structure learning. In Thirty-Fifth AAAI Conference on Artificial Intelligence, AAAI 2021, Thirty-Third Conference on Innovative Applications of Artificial Intelligence, IAAI ...

    2021 doi

  7. [15]

    The power of collaboration: Learning large bayesian networks at scale

    Vaidyanathan Peruvemba Ramaswamy, Stefan Szeider, and Hai Xia. The power of collaboration: Learning large bayesian networks at scale. In 36th IEEE International Conference on Tools with Artificial Intelligence, ICTAI 2024, Herndon, VA, USA, October 28-30, 2024 , pages 371--378...

    2024

  8. [16]

    Pawan Kumar, Emilien Dupont, Francisco J

    Bernardino Romera - Paredes, Mohammadamin Barekatain, Alexander Novikov, Matej Balog, M. Pawan Kumar, Emilien Dupont, Francisco J. R. Ruiz, Jordan S. Ellenberg, Pengming Wang, Omar Fawzi, Pushmeet Kohli, and Alhussein Fawzi. Mathematical discoveries from program search with la...

    2024 doi

  9. [17]

    An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows

    Stefan Ropke and David Pisinger. An adaptive large neighborhood search heuristic for the pickup and delivery problem with time windows. Transp. Sci., 40 0 (4): 0 455--472, 2006. doi:10.1287/TRSC.1050.0135. URL https://doi.org/10.1287/trsc.1050.0135

    2006

  10. [18]

    Sat-boosted tabu search for coloring massive graphs

    Andr \' e Schidler and Stefan Szeider. Sat-boosted tabu search for coloring massive graphs. ACM J. Exp. Algorithmics , 28: 0 1.5:1--1.5:19, 2023. doi:10.1145/3603112. URL https://doi.org/10.1145/3603112

    2023 doi

  11. [19]

    Structure-guided local improvement for maximum satisfiability

    Andr \' e Schidler and Stefan Szeider. Structure-guided local improvement for maximum satisfiability. In Paul Shaw, editor, 30th International Conference on Principles and Practice of Constraint Programming, CP 2024, Girona, Spain, September 2-6, 2024 , volume 307 of LIPIcs, p...

    2024 doi

  12. [20]

    SAT -based decision tree learning for large data sets

    Andr \' e Schidler and Stefan Szeider. SAT -based decision tree learning for large data sets. J. Artif. Intell. Res., 80: 0 875--918, 2024 b . doi:10.1613/JAIR.1.15956. URL https://doi.org/10.1613/jair.1.15956

    2024 doi

  13. [21]

    Using constraint programming and local search methods to solve vehicle routing problems

    Paul Shaw. Using constraint programming and local search methods to solve vehicle routing problems. In Michael J. Maher and Jean - Francois Puget, editors, Principles and Practice of Constraint Programming - CP98, 4th International Conference, Pisa, Italy, October 26-30, 1998,...

    1998 doi

  14. [22]

    Text2zinc: A cross-domain dataset for modeling optimization and satisfaction problems in minizinc

    Akash Singirikonda, Serdar Kadioglu, and Karthik Uppuluri. Text2zinc: A cross-domain dataset for modeling optimization and satisfaction problems in minizinc. CoRR, abs/2503.10642, 2025. doi:10.48550/ARXIV.2503.10642. URL https://doi.org/10.48550/arXiv.2503.10642

    2025 doi arXiv

  15. [23]

    SAT-Based tree decomposition with iterative cascading policy selection

    Hai Xia and Stefan Szeider. SAT-Based tree decomposition with iterative cascading policy selection. In Michael J. Wooldridge, Jennifer G. Dy, and Sriraam Natarajan, editors, Thirty-Eighth AAAI Conference on Artificial Intelligence, AAAI 2024, Thirty-Sixth Conference on Innovat...

    2024

  16. [24]

    Hoos, and Kevin Leyton - Brown

    Lin Xu, Frank Hutter, Holger H. Hoos, and Kevin Leyton - Brown. Satzilla: Portfolio-based algorithm selection for SAT . J. Artif. Intell. Res., 32: 0 565--606, 2008. doi:10.1613/JAIR.2490. URL https://doi.org/10.1613/jair.2490

    2008 doi

  17. [25]

    Le, Denny Zhou, and Xinyun Chen

    Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V. Le, Denny Zhou, and Xinyun Chen. Large language models as optimizers. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024 . OpenReview.net, 2024. URL https:/...

    2024

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.