Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Controllability of Control Argumentation Frameworks
Authors: Andreas Niskanen, Daniel Neugebauer, Matti Järvisalo
IJCAI 2020 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We establish a complete computational complexity map of the central computational problem of controllability in CAFs for five key semantics. We also develop Boolean satisfiability based counterexample-guided abstraction refinement algorithms and direct encodings of controllability as quantified Boolean formulas, and empirically evaluate their scalability on a range of NPhard variants of controllability. |
| Researcher Affiliation | Academia | 1HIIT, Department of Computer Science, University of Helsinki, Finland 2Institut f ur Informatik, Heinrich-Heine Universit at D usseldorf, Germany |
| Pseudocode | Yes | Algorithm 1 CEGAR for skeptical controllability Input: CAF C = (F, C, U), target T AF , σ {com, stb}... Algorithm 2 CEGAR for credulous controllability Input: CAF C = (F, C, U), target T AF , σ {adm, stb}. |
| Open Source Code | Yes | The implementation, benchmarks, and runtime data are available online. |
| Open Datasets | Yes | We generated CAFs from the 2019 ICCMA competition AFs (http://argumentationcompetition.org/2019/iccma-instances.tar.gz) as follows. |
| Dataset Splits | No | The paper describes how the CAFs were generated from existing AFs by introducing probabilistic parameters (p C, p U) for control arguments and uncertain parts. However, it does not specify a division of these generated instances into distinct training, validation, and test sets. The generated instances are used directly for evaluation. |
| Hardware Specification | Yes | The experiments were run on Intel Xeon E5-2680 v4 2.4-GHz, 256-GB nodes with Cent OS 7 under a per-instance 900-s time and 64-GB memory limit. |
| Software Dependencies | Yes | We used the QBF solver CAQE 4.0.0 [Tentrup, 2019] with the Bloqqer [Heule et al., 2015] preprocessor and the flag --qdo to obtain assignments corresponding to control configurations. For CEGAR we used the Glucose 4.1 SAT solver [Audemard and Simon, 2018]. |
| Experiment Setup | Yes | For each p C {0.05, 0.1, 0.15, 0.2}, each non-query argument is a control argument with probability p C. For each p U {0, 0.05, 0.1, 0.15, 0.2}, each argument (apart from control and query arguments) is uncertain with probability p U. Each attack is uncertain with probability p U/2 unless the source or the target is a control argument, and has uncertain direction with probability p U/2 unless the reverse direction is already a fixed or an uncertain attack. The rest of the arguments and attacks remain fixed. This yielded a total of 6520 CAFs, out of which 1304 are simplified CAFs with no uncertain part. The experiments were run... under a per-instance 900-s time and 64-GB memory limit. |