AL-GR/Item-SID

AL-GR-Item-SID是一个包含来自匿名电商平台产品的语义ID（SID）的数据集。这些SID通过多模态模型生成，专为生成式推荐系统设计，作为密集、有意义的特征。数据集使用残差量化VAE模型创建分层表示，并提供两种配置：基础模型和微调模型生成的SID。数据集以JSON对象形式存储，包含产品唯一标识符和对应的三级分层语义ID。

The AL-GR-Item-SID dataset contains Semantic IDs (SIDs) for products from an anonymized e-commerce platform. These SIDs are generated by a multi-modal model and are designed for generative recommendation systems as dense, meaningful features. The dataset uses a Residual-Quantized VAE model to create a hierarchical representation and is available in two configurations: those generated by the base model and the fine-tuned model. The dataset is stored as JSON objects containing a unique identifier for each product and corresponding three-level hierarchical semantic IDs.