FedMABench: Benchmarking Mobile Agents on Decentralized Heterogeneous User Data
Paper
•
2503.05143
•
Published
Error code: TooManyColumnsError
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
Accepted by EMNLP 2025 Main Conference ⭐️
FedMABench is an open-source benchmark for federated training and evaluation of mobile agents, specifically designed for heterogeneous scenarios.
FedMABench includes the following key features:
Clone the repo, submodules and install the required packages.
git clone --recursive --shallow-submodules https://github.com/wwh0411/FedMABench.git
cd FedMABench
conda create -n fedma python=3.10
conda activate fedma
pip install -r requirements.txt
The six datasets of FedMABench are organized in seperate directories. We also provide test sets and temporary split files for reference.
.
├── App-Level
│ ├── App Half-Skew.jsonl
│ ├── App IID.jsonl
│ ├── App Non-Uniform.jsonl
│ ├── App Skew.jsonl
│ ├── App-Level Val.jsonl
│ ├── train_split.json
│ └── val_split.json
├── Basic-AC
│ ├── Basic-AC Entertainment.jsonl
│ ├── Basic-AC Lives.jsonl
│ ├── Basic-AC Office.jsonl
│ ├── Basic-AC Shopping.jsonl
│ ├── Basic-AC Traveling.jsonl
│ ├── Basic-AC c10n1000.jsonl
│ ├── Basic-AC c10n200.jsonl
│ ├── Basic-AC c10n3000.jsonl
│ ├── Basic-AC c10n500.jsonl
│ ├── Basic-AC c10n5000.jsonl
│ ├── Basic-AC c10n7000.jsonl
│ ├── Basic-AC c30n3000.jsonl
│ ├── Basic-AC c50n5000.jsonl
│ ├── Basic-AC c70n7000.jsonl
│ ├── Val_100.jsonl
│ └── val_100.json
├── Basic-AitW
│ ├── Basic-AitW G-Apps.jsonl
│ ├── Basic-AitW General.jsonl
│ ├── Basic-AitW Install.jsonl
│ ├── Basic-AitW Single.jsonl
│ └── Basic-AitW WebShopping.jsonl
├── Category-Level
│ ├── App Random.jsonl
│ ├── App Skew.jsonl
│ ├── Category Half-Skew.jsonl
│ ├── Category IID.jsonl
│ ├── Category Non-Unifrom.jsonl
│ ├── Category Skew.jsonl
│ ├── Category Val.jsonl
│ ├── train_split.json
│ └── val_split.json
├── ScaleApp
│ ├── ScaleApp IID.jsonl
│ ├── ScaleApp Random.jsonl
│ ├── ScaleApp Skew.jsonl
│ ├── ScaleApp Val_250.jsonl
│ ├── train_split.json
│ └── val_split.json
├── Step-Episode
│ ├── Both Skew.jsonl
│ ├── Episode Skew.jsonl
│ ├── Step Skew.jsonl
│ ├── Step-Episode IID.jsonl
│ ├── Step-Episode Val.jsonl
│ └── val_split.json
└── README.md
@misc{wang2025fedmabenchbenchmarkingmobileagents,
title={FedMABench: Benchmarking Mobile Agents on Decentralized Heterogeneous User Data},
author={Wenhao Wang and Zijie Yu and Rui Ye and Jianqing Zhang and Siheng Chen and Yanfeng Wang},
year={2025},
eprint={2503.05143},
archivePrefix={arXiv},
primaryClass={cs.AI},
url={https://arxiv.org/abs/2503.05143},
}