Learning Temporal Plan Preferences from Examples: An Empirical Study
Authors: Valentin Seimetz, Rebecca Eifler, Jörg Hoffmann
IJCAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct an empirical study of this approach in an oversubscription planning context, using hidden target formulas to emulate the user preferences. |
| Researcher Affiliation | Academia | 1German Research Center for Artificial Intelligence (DFKI), Saarbr ucken, Germany 2Saarland University, Saarland Informatics Campus, Saarbr ucken, Germany |
| Pseudocode | No | The paper describes the methods and processes in narrative text but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks or figures. |
| Open Source Code | No | The paper states it used 'publicly available implementations' for plan generation and a 'modified re-implementation' of an existing tool for LTLf learning, but it does not explicitly provide a link or statement confirming that the authors' own assembled code or modifications are open-source or publicly available. |
| Open Datasets | Yes | The planning instances we consider are based on the resource-constrained planning instances used by Eifler et al. [Eifler et al., 2020a]. |
| Dataset Splits | No | The paper discusses generating example plans and using benchmark instances (task-formula pairs) but does not specify explicit training, validation, and test dataset splits with percentages or sample counts for model evaluation. |
| Hardware Specification | Yes | All experiments were run on Intel E5-2660 machines running at 2.20 GHz with a memory limit of 4GB. |
| Software Dependencies | No | The paper mentions software like Sym K, forbiditerative, Fast Downward, VAL, and a re-implementation of Camacho & Mc Ilraith's approach, but it does not provide specific version numbers for any of these software dependencies. |
| Experiment Setup | Yes | The paper details the benchmark design, including how hidden target formulas were generated using LTL templates and random facts, and describes the plan generation process, including the number of example plans generated (e.g., 'up to 50 example plans' for Genapp and 'up to 25 positive and 25 negative examples' for Genideal). |