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..

On the Existence and Complexity of Core-Stable Data Exchanges

Authors: Jiaxin Song, Pooja Kulkarni, Parnian Shahkar, Bhaskar Ray Chaudhury

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We show that the pivoting algorithm works well in practice through our empirical results. ... In Section 4, we validate the practical efficacy of our algorithm through simulations on a mean estimation task similar to [BGI+24a].
Researcher Affiliation Academia Jiaxin Song University of Illinois, Urbana-Champaign EMAIL Pooja Ravi Kulkarni Northwestern University EMAIL Parnian Shahkar University of California, Irvine EMAIL Bhaskar Ray Chaudhury University of Illinois, Urbana-Champaign EMAIL
Pseudocode Yes Algorithm 2: Pivoting algorithm for finding ฯต-core-stable exchange
Open Source Code Yes We include all the codes in the Supplementary files, including codes and graph data.
Open Datasets Yes We construct a data exchange instance using the same road map dataset [roa] as the previous work [BGI+24b]. ... [roa] Street network of new york in Graph ML. https://www.kaggle.com/datasets/ crailtap/street-network-of-new-york-in-graphml. Accessed: 2023-09-20.
Dataset Splits Yes We adopt the same method as [BGI+24b] to sample an agent i from the road map: Sample a random node u and then sample a length t uniformly at random between 5 and the depth of the BFS tree rooted at u. Then we sample another node uniformly at random from all nodes within layer t, choose the shortest path from u to v, and assign it to agent i.
Hardware Specification Yes All experiments were run on a Mac Book Pro with an Apple M3 Pro CPU and 18 GB RAM.
Software Dependencies Yes The implementation uses Python 3.12 and Sci Py [VGO+20] (v1.13.0) for concave optimization.
Experiment Setup Yes The variance of each edge, ฯƒe, is a random number in range [0, 1], and ยตe is set as (1 ฯƒe) 10 3. In addition, every agent i starts with zi e random data samples for every edge in her path, where zi e is chosen uniformly at random in the range [4, 9].