移动端智能数据集

随着移动设备的算力增强，轻量级的排序模型已经可以在端上进行推理，未来随着技术的发展，我们希望做到数据不回收而只回收抽象模型，这提供了更好地保护消费者隐私的能力，同时通过端上模型能力保证优质的服务。

As the computing power of mobile devices continues to increase, lightweight ranking models can already perform inference on the device side. In the future, with technological advancements, we aim to refrain from collecting raw user data and only collect abstract models instead. This approach not only provides better protection for consumers' privacy but also ensures high-quality services through the on-device model capabilities.

搜集汇总

背景与挑战

背景概述

该数据集专注于移动端智能推荐，包含传统排序和时空信息流推荐两大场景，旨在通过端上轻量级模型推理保护用户隐私。数据涵盖加密的用户、商品及时空特征，提供训练集和测试集，适用于排序模型的研究与评估。