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..
Enhanced Partial Expansion A*
Authors: M. Goldenberg, A. Felner, R. Stern, G. Sharon, N. Sturtevant, R. C. Holte, J. Schaeffer
JAIR 2014 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental studies show significant improvements in run-time and memory performance for several standard benchmark applications. |
| Researcher Affiliation | Academia | Meir Goldenberg EMAIL Ariel Felner EMAIL Roni Stern EMAIL Guni Sharon EMAIL Ben-Gurion University of the Negev Beer-Sheva, Israel Nathan Sturtevant EMAIL The University of Denver, Denver, USA Robert C. Holte EMAIL Jonathan Schaeffer EMAIL The University of Alberta Edmonton, Canada |
| Pseudocode | Yes | Procedure 1 A*, PEA* and EPEA*. Procedure 2 IDA* and EPEIDA*. Procedure 3 Algorithmic component of an additive PDBs-based OSF for MAPF. |
| Open Source Code | No | The paper does not provide concrete access to source code for the methodology described by the authors. While it mentions using Korf's IDA* code, it does not offer the authors' implementation of EPEA* or EPEIDA*. |
| Open Datasets | Yes | Optimal solutions to random instances of the 15-puzzle were first found by Korf (1985) using IDA* and the MD heuristic. Korf has graciously made this code available to the public. The pancake puzzle (Dweighter, 1975). Rubik s Cube was invented in 1974 by Erno Rubik of Hungary. |
| Dataset Splits | No | The paper focuses on combinatorial search problems, where experiments are run on problem 'instances' (e.g., 100 random instances for the 15-puzzle, pancake puzzle, Rubik's cube, or 1,000 generated instances for MAPF). The concept of training/test/validation splits, as typically found in machine learning contexts, is not applicable or explicitly provided for these search problem instances. |
| Hardware Specification | Yes | The timing results were obtained on Dell Optiplex 760. |
| Software Dependencies | No | The paper mentions using "Korf's IDA* code" for the 15-puzzle experiments, but it does not specify any version numbers for this or any other software components used in their methodology. |
| Experiment Setup | Yes | All algorithmic variants were run under the ID framework as described in the footnote. All variants were given up to two minutes and two gigabytes of memory per instance. |