MAC: Multi-Attribution BenChmark

Paper | Code

Conversion Rate (CVR) prediction is a cornerstone of online advertising systems. However, existing public CVR datasets—such as Criteo and Ali-CCP—provide conversion labels derived from a single attribution mechanism, severely limiting research into more holistic modeling paradigms. To bridge this gap, we introduce MAC (Multi-Attribution BenChmark), the first public CVR dataset featuring labels from multiple attribution mechanisms.

Overview

MAC is the first benchmark featuring labels from multiple attribution mechanisms, specifically designed to foster research in Multi-Attribution Learning (MAL). By learning from conversion labels yielded by multiple attribution mechanisms, models can obtain a more comprehensive and robust understanding of touchpoint value.

Along with the dataset, the authors provide PyMAL, an open-source library covering a wide array of baseline methods (such as MMoE, PLE, and MoAE) for industrial-scale CVR prediction.

Dataset Structure

The files in this repository are organized as follows:

• train/: Training data files.
• test/: Test data files.
• vocabs/: ID mappings and vocabulary files for features.

Usage

To download the dataset directly via git clone:

git clone https://huggingface.co/datasets/alimamaTech/MAC data

Citation

@misc{wu2026macconversionrateprediction,
      title={MAC: A Conversion Rate Prediction Benchmark Featuring Labels Under Multiple Attribution Mechanisms}, 
      author={Jinqi Wu and Sishuo Chen and Zhangming Chan and Yong Bai and Lei Zhang and Sheng Chen and Chenghuan Hou and Xiang-Rong Sheng and Han Zhu and Jian Xu and Bo Zheng and Chaoyou Fu},
      year={2026},
      eprint={2603.02184},
      archivePrefix={arXiv},
      primaryClass={cs.LG},
      url={https://arxiv.org/abs/2603.02184}, 
}