yandex/yambda

Yambda-5B是一个大规模的多模态数据集，用于排序和检索任务。它包含来自100万用户的4.79亿个用户-音乐交互，跨越939万首曲目。数据集包括隐式反馈（如收听事件）和显式反馈（如喜欢和不喜欢）。此外，它还提供了区分有机发现和推荐驱动交互的独特标记，以及预计算的音频嵌入，以促进内容感知推荐系统的发展。

Yambda-5B is a large-scale multi-modal dataset for ranking and retrieval tasks. It contains 479 million user-music interactions from 1 million users spanning 9.39 million tracks. The dataset includes implicit feedback such as listening events and explicit feedback like likes and dislikes. Additionally, it provides distinctive markers for organic versus recommendation-driven interactions, along with precomputed audio embeddings to facilitate content-aware recommendation systems.