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..
Design-Based Bandits Under Network Interference: Trade-Off Between Regret and Statistical Inference
Authors: Zichen Wang, Haoyang Hong, Chuanhao Li, Haoxuan Li, Zhiheng Zhang, Huazheng Wang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we demonstrate the empirical performance of our EXP3-N-CS by some simulation studies. The code is available at: https://github.com/TheoryMagic/Design-based-Bandits. Setup. We consider a network consisting of 101 units. ... Each algorithm is executed 1000 times, and we report the averaged results. Results. The simulation results are shown in Fig. 2(a), 2(b) and 2(c). |
| Researcher Affiliation | Academia | 1Department of ECE and CSL, UIUC 2School of EECS, Oregon State University 3Department of Industrial Engineering, Tsinghua University 4Center for Data Science, Peking University 5School of Statistics and Data Science, Shanghai University of Finance and Economics, Shanghai 200433, P.R. China 6Institute of Data Science and Statistics, Shanghai University of Finance and Economics, Shanghai 200433, P.R. China |
| Pseudocode | Yes | Algorithm 1 EXP3-N-CS 1: Input: arm set A, unit set U, exposure super arm set UE, sequence {Lm} m=1 2: for t = 1, 2, . . . do ... Algorithm 2 Sampling 1: Input: St 2: Derive the set of real super arm {Zl }l [l] such that for all Zl , {S(i, Zl , H)}i U = St 3: Sample At from set {Zl }l [l] based on P(At = Zl | St), pull At, and observe reward Rt(St) = 1 i U Yi,t(At) |
| Open Source Code | Yes | The code is available at: https://github.com/TheoryMagic/Design-based-Bandits. |
| Open Datasets | No | We consider a network consisting of 101 units. Specifically, there is one center cluster C1 = {1} that contains a single unit, which is connected to every unit in the five outer clusters. Each of the outer clusters contains 20 units. We set the action set K = {0, 1}. Additionally, we define the exposure mapping inspired by [Leung, 2022a, Gao and Ding, 2023]... Besides, we let Yt(S) be sampled from a Bernoulli distribution. The mean of this Bernoulli distribution is uniformly resampled from [0, 1] every 1000 rounds. |
| Dataset Splits | No | Each algorithm is executed 1000 times, and we report the averaged results. |
| Hardware Specification | No | The paper does not provide specific hardware details for the experimental setup. |
| Software Dependencies | No | The paper does not explicitly list software dependencies with specific version numbers within its text. While code is provided, the paper does not mention Python, PyTorch, or any other library versions. |
| Experiment Setup | Yes | We set the trade-off parameter of EXP3-N-CS to α {0.1, 0.2, 0.3, 0.4, 0.49} and compare its performance against two baselines: Standard (where δt = 0) and Uniform (where δt = 1). |