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..
Combinatorial Ski Rental Problem: Robust and Learning-Augmented Algorithms
Authors: Ziwei Li, Bo Sun, Zhiqiu Zhang, Mohammad Hajiesmaili, Binghan Wu, Lin F. Yang, Yang Gao
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental results validate our theoretical analysis and demonstrate the advantages of our algorithms over baseline methods for ski rental problems. Numerical experiments validate our theoretical results and demonstrate the performance advantage of LA-SOAC. |
| Researcher Affiliation | Collaboration | Ziwei Li Nanjing University EMAIL Bo Sun University of Ottawa, Vector Institute EMAIL Zhiqiu Zhang Nanjing University EMAIL Mohammad Hajiesmaili University of Massachusetts, Amherst EMAIL Binghan Wu Asia Info Technologies Limited EMAIL Lin Yang Nanjing University EMAIL Yang Gao Nanjing University EMAIL |
| Pseudocode | Yes | Algorithm 1 OAC: Optimal Amortized Cost Algorithm 2 SOAC: Sorted Optimal Amortized Cost Algorithm 3 LA-SOAC: Learning-Augmented SOAC |
| Open Source Code | Yes | The source code for reproducing all experiments is available at https://github. com/guodongsanjianke/Combinatorial_Ski_Rental_Problems. |
| Open Datasets | Yes | The evaluation utilizes the dataset from [25], comprising 6 shops with purchase prices defined as b1 = 100, b2 = 95, b3 = 90, b4 = 85, b5 = 80, b6 = 75 and rental prices given by r1 = 1.00, r2 = 1.05, r3 = 1.10, r4 = 1.15, r5 = 1.20, r6 = 1.25. Consistent with [25], the actual number of days is modeled as uniformly distributed within the interval [1, 3b1]. |
| Dataset Splits | Yes | We let the actual number of days, T, be uniformly distributed within the region [1, 4TOPT]. The predicted number of days, y, is set to y = T + ϵ, where the simulated error ϵ follows a normal distribution with mean δ and standard deviation η, i.e., ϵ N(δ, η2). Here, δ controls the bias of the prediction, while η (referred to as the error parameter) characterizes the variability or uncertainty in the prediction. For each standard deviation, 10,000 samples are randomly sampled in this study, and their average competitive ratios are calculated. |
| Hardware Specification | Yes | The experimental platform is an AMAX TR40-X4 server, configured with dual Intel Xeon Gold 6448H processors. The system is equipped with 512 GB of DDR5-4800 ECC memory, two 16 TB hard disk drives, two 960 GB solid-state drives, and four graphics processing units. |
| Software Dependencies | No | No specific software dependencies with version numbers are mentioned in the paper. |
| Experiment Setup | Yes | Consider three items, with item prices b1 = 80, b2 = 110, and b3 = 130, and the same rental price of 1. The discount factor for purchasing any two items together is set to 0.95, and 0.9 for three items. We let the actual number of days, T, be uniformly distributed within the region [1, 4TOPT]. The predicted number of days, y, is set to y = T + ϵ, where the simulated error ϵ follows a normal distribution with mean δ and standard deviation η, i.e., ϵ N(δ, η2). Here, δ controls the bias of the prediction, while η (referred to as the error parameter) characterizes the variability or uncertainty in the prediction. For each standard deviation, 10,000 samples are randomly sampled in this study, and their average competitive ratios are calculated. |