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..
Federated Multi-armed Bandits with Efficient Bit-Level Communications
Authors: Haoran Zhang, Yang Xu, Xuchuang Wang, Hao-Xu Chen, Hao Qiu, Lin F. Yang, Yang Gao
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Theoretically, we derive tight upper bounds on both individual cumulative regret and group regret, and prove that our method asymptotically matches the lower bound of regret in federated settings. Experimental results on synthetic data validate the effectiveness of Epo Inc-SE in various settings and under heterogeneous feedback. |
| Researcher Affiliation | Academia | Haoran Zhang Nanjing University EMAIL Yang Xu Nanjing University EMAIL Xuchuang Wang University of Massachusetts Amherst EMAIL Hao-Xu Chen Nanjing University EMAIL Hao Qiu Università degli Studi di Milano EMAIL Lin Yang Nanjing University EMAIL Yang Gao Nanjing University EMAIL |
| Pseudocode | Yes | Algorithm 1: Epoch-based Successive Elimination Algorithm (Epo Inc-SE)(for agent j) Algorithm 2: Epoch-based consensus estimation subroutine (EBCES) (for agent j) |
| Open Source Code | Yes | This work does not have any dataset, but we provide all code in the supplemental material. |
| Open Datasets | No | Experimental results on synthetic data validate the effectiveness of Epo Inc-SE in various settings and under heterogeneous feedback. This work does not have any dataset, but we provide all code in the supplemental material. |
| Dataset Splits | No | The paper mentions using "synthetic data" for experiments but does not specify any training/test/validation splits for this data. |
| Hardware Specification | No | The paper does not provide specific hardware details. In the NeurIPS Paper Checklist, question 8 states: "Our work is theoretical, and the experiment is not the core. Besides, our experiment is a small simulation that could work on any computer with VS Code. Hence, we think it is unnecessary." |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers. |
| Experiment Setup | Yes | Setups and Baselines. Unless otherwise stated, the experiment scenario involves a network of N = 8 agents and K = 10 arms, parameters C = N, a = 5 and T = 106. To ensure a fair comparison, we use a ring graph and all agents occupy the same weight in their neighborhoods, which is a common undirected connected graph with the second largest eigenvalue λ2 = 0.5713. |