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..
Time-uniform and Asymptotic Confidence Sequence of Quantile under Local Differential Privacy
Authors: Leheng Cai, Qirui Hu, Juntao Sun, Shuyuan Wu
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we evaluate the finite-sample performance of the proposed method. The confidential data are generated from two distributions: standard Normal N(0, 1) and standard Cauchy C(0, 1). Target quantiles are set to Ï = 0.8, 0.5, 0.3. The truthful response rates are chosen as r = 1, 0.9, 0.75, 0.5, 0.25... Our first analysis focuses on the time-uniform convergence performance... Results for the standard normal distribution based on equations (4) and (6) are presented in Figures 3 and 4... In this section, we empirically evaluate the effectiveness of our proposed method on the following two representative real datasets widely used in privacy research: Law school dataset [53]... Government salary dataset [42]. |
| Researcher Affiliation | Academia | 1 Department of Statistics and Data Science Tsinghua University, Beijing, China 2 School of Statistics and Data Science Shanghai University of Finance and Economics, Shanghai, China 3Institute of Big Data Research, Shanghai University of Finance and Economics, Shanghai, China EMAIL EMAIL EMAIL EMAIL |
| Pseudocode | Yes | Algorithm 1 Data allocation for parallel runs |
| Open Source Code | Yes | Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: We have uploaded the code that reproduces the experimental results in the paper. |
| Open Datasets | Yes | In this section, we empirically evaluate the effectiveness of our proposed method on the following two representative real datasets widely used in privacy research: Law school dataset [53]... Government salary dataset [42]. |
| Dataset Splits | No | The paper describes generating data from distributions and using real datasets but does not explicitly detail training, validation, or test splits. It mentions a 'burn-in strategy' but not how the overall dataset was partitioned for evaluation. |
| Hardware Specification | Yes | Each experiment is replicated 2000 times using 110 Intel Xeon Platinum 8352V CPU @ 2.10GHz CPUs with 360GB memory and 1200GB storage. |
| Software Dependencies | No | The paper does not explicitly list any software dependencies with version numbers. |
| Experiment Setup | Yes | Target quantiles are set to Ï = 0.8, 0.5, 0.3. The truthful response rates are chosen as r = 1, 0.9, 0.75, 0.5, 0.25... The algorithm uses random initialization with standard Normal N(0, 1) of all chains and step sizes set to ηÎș,t = 1/ta with a = 0.6 for all chains as well... Following [35], we incorporate a burn-in strategy into the algorithm to reduce the impact of initial parameter bias and enhance the stability of statistical inference, with the number of burn-in samples being about (0.25/r2)% of the total sample size. Each experiment is replicated 2000 times... |