Multi-Attribute RFANNS Evaluation Datasets (MSMarco and Laion)

This upload contains the processed multi-attribute RFANNS evaluation datasets MSMarco and Laion, derived from public corpora (MS MARCO web passages and the LAION-400M image-text dataset). Each dataset includes dense vector embeddings for all objects and multiple numeric attribute columns (e.g., text length and content statistics for MSMarco, image dimensions and similarity scores for Laion). All files are provided in the binary format used by our implementation (*.meta.bin, *.vectors.bin), making them directly compatible with the APG codebase. The exact file formats are documented in the APG repository: https://github.com/yuyuanhang/APG. Users are encouraged to cite the associated paper when using this data.