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..
Online Strategic Classification With Noise and Partial Feedback
Authors: Tianrun Zhao, Xiaojie Mao, Yong Liang
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We also conduct numerical experiments to evaluate our proposed algorithm, with results presented in Appendix A.2. |
| Researcher Affiliation | Academia | Tianrun Zhao, Xiaojie Mao , Yong Liang School of Economics and Management, Tsinghua University, Beijing, China, 100084 EMAIL, EMAIL, EMAIL, |
| Pseudocode | Yes | Algorithm 1: Main-Algorithm Algorithm 2: Initialization Algorithm 3: Refinement Algorithm 4: Batched-Enhancement Algorithm 5: Original Strategic Perceptron with Full Feedback (Ahmadi et al. [2021]) Algorithm 6: Strategic Perceptron with Partial Feedback Algorithm 7: Non-Strategic Learning under Massart Noise Algorithm 8: Non-Strategic-Initialization Algorithm 9: Non-Strategic-Refinement Algorithm 10: Non-Strategic-Batched Enhancement |
| Open Source Code | Yes | The paper provides the code and data required for the experiments in the supplementary material. The code is well-organized and well-documented, which facilitates the reproduction process. |
| Open Datasets | No | Our numerical experiment only uses simulated data, so it does not involve any deprecated datasets or copyright violations. |
| Dataset Splits | No | The paper describes numerical experiments using simulated data but does not specify any training/test/validation splits for this data. It mentions 'Each setting is replicated 30 times', indicating multiple runs but not data partitioning for model training/evaluation. |
| Hardware Specification | Yes | We state in Appendix A.2 that all experiments can be conducted locally using a standard CPU without requiring specialized hardware, making reproduction accessible and straightforward. |
| Software Dependencies | Yes | We use Python 3.9 to conduct our numerical experiments. |
| Experiment Setup | Yes | We test the algorithms under two different settings, with key parameters outlined in Table 1. Each setting is replicated 30 times, and we report the average regret for each algorithm. Our analysis includes a performance comparison of the different algorithms and an investigation of how various problem parameters influence our proposed algorithm. |