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 [1].
The Computational Rise and Fall of Fairness
Authors: John Dickerson, Jonathan Goldman, Jeremy Karp, Ariel Procaccia, Tuomas Sandholm
AAAI 2014 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We support these results experimentally and show that the asymptotic behavior of the theory holds even when the number of goods and agents is quite small. |
| Researcher Affiliation | Academia | John P. Dickerson Carnegie Mellon University EMAIL Jonathan Goldman Carnegie Mellon University EMAIL Jeremy Karp Carnegie Mellon University EMAIL Ariel D. Procaccia Carnegie Mellon University EMAIL Tuomas Sandholm Carnegie Mellon University EMAIL |
| Pseudocode | No | No structured pseudocode or algorithm blocks were found. |
| Open Source Code | Yes | Source code & data: https://github.com/John Dickerson/Envy Free |
| Open Datasets | No | The paper describes generating instances by sampling valuations from two distributions, 'UNIFORM(x, y)' and 'CORRELATED(x, y)', but does not provide concrete access information (link, DOI, repository, or formal citation) to the specific datasets used in their experiments. |
| Dataset Splits | No | The paper describes generating instances with 'n' agents and 'm' goods but does not provide specific details on training, validation, and test dataset splits. The experiments analyze the existence and computation of EF allocations across varying 'm' and 'n', rather than using a train/validation/test split for model evaluation. |
| Hardware Specification | Yes | Runs were conducted on Blacklight,3 a cc NUMA supercomputer with 8GB of RAM per core; each experiment was run at least 160 times with a time limit of 12 hours per run. |
| Software Dependencies | Yes | All experiments were performed in Python using IBM ILOG CPLEX 12.61 in single-threaded mode under its default configuration. |
| Experiment Setup | Yes | We generate instances with n agents and m goods as follows by sampling valuations for each agent and each good from a given distribution over utility functions. In our experimental setup, we draw from two distributions CORRELATED(0.4, 0.6) and UNIFORM(0, 1) defined earlier... All experiments were performed in Python using IBM ILOG CPLEX 12.61 in single-threaded mode under its default configuration. Runs were conducted on Blacklight,3 a cc NUMA supercomputer with 8GB of RAM per core; each experiment was run at least 160 times with a time limit of 12 hours per run. |