A sequential recommendation dataset drawn from Amazon Reviews 2023, covering Books, CDs_and_Vinyl, Movies_and_TV, Video_Games. Each row of interactions.parquet is one user buying or reviewing one item at one time. Users are sampled so that every one of them is active in all domains, their interactions are ordered chronologically and cut into train/valid/test, and each interaction carries a fixed set of 10 candidate items for ranking… See the full description on the dataset page: https://huggingface.co/datasets/sungjin-code/amazon-reviews-for-llm.