HeLLM: Multi-Modal Hypergraph Enhanced LLM Learning for Recommendation

    To evaluate our proposed method, we conduct experiments on three public benchmark datasets from Amazon (original Datasets are available at http:\/\/jmcauley.ucsd.edu\/data\/amazon\/links.html): Sports (Sports and Outdoors), Beauty, and Toys (Toys and Games), with dataset statistics in Table.    Each dataset includes visual and textual modalities. For the textual modality, we concatenate titles, brands, categories, and descriptions, processing them through BERT (https:\/\/huggingface.co\/google-bert\/bert-base-uncased) to obtain 768-dimensional features. For the visual modality, images are sourced from the original URLs, with 512-dimensional features extracted via VIT(https:\/\/huggingface.co\/openai\/clip-vit-base-patch32).    Following prior work, we split each user sequence by using the last interaction as the test set, the penultimate as the validation set, and all preceding interactions as the training set.