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..

Tight Asymptotics of Extreme Order Statistics

Authors: Jose Correa, Frederik Mallmann-Trenn, Matias Romero

NeurIPS 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Beyond the theoretical analysis, we support our findings with extensive simulations. These empirical results highlight a notable phenomenon: although the multiplicative gap between the maximum and the second maximum grows quickly with n, the ratio remains approximately constant in 99% of trials.
Researcher Affiliation Academia José R. Correa Universidad de Chile EMAIL Frederik Mallmann-Trenn King s College London EMAIL Matías Romero Columbia University EMAIL
Pseudocode No The paper describes mathematical proofs and theoretical derivations in sections like "Proof of Proposition 1", "Proof of Lemma 3", etc., but does not contain any structured pseudocode or algorithm blocks.
Open Source Code Yes The code is included in the supplementary material.
Open Datasets No The experiments with this paper are numerical simulations. No datasets are being used.
Dataset Splits No The experiments with this paper are numerical simulations. No datasets are being used.
Hardware Specification No All experiments were conducted on the Columbia Business School (CBS) Research Grid, a high-performance computing cluster running a Linux environment (Debian 4.19).
Software Dependencies No All experiments were conducted on the Columbia Business School (CBS) Research Grid, a high-performance computing cluster running a Linux environment (Debian 4.19). We fixed the random seed to 42 using Python s default pseudo-random number generator to ensure reproducibility.
Experiment Setup Yes For each distribution, we generate 1,000,000 independent trials for values of n ranging from 10 to 10,000. In each trial, we computed the top order statistics Xn:n, Xn 1:n, Xn 2:n, Xn 3:n and averaged across trials to estimate Mℓ. We fixed the random seed to 42 using Python s default pseudo-random number generator to ensure reproducibility.