RecBench
收藏资源简介:
RecBench是由香港理工大学等机构提出的推荐系统评估平台,该平台旨在全面评估大型语言模型在推荐任务中的性能。它涵盖了不同的项目表示形式,并评估了点击率预测和序列推荐两个主要推荐任务。该平台评估了17种大型模型,并使用来自时尚、新闻、视频、书籍和音乐领域的五个不同数据集进行实验。RecBench提供了一个深入的评估,以推动推荐系统中大型语言模型的研究与发展。
RecBench is a recommender system evaluation platform proposed by The Hong Kong Polytechnic University and other institutions. This platform aims to comprehensively evaluate the performance of large language models (LLMs) in recommendation tasks. It covers diverse item representation forms, and evaluates two core recommendation tasks: click-through rate (CTR) prediction and sequential recommendation. The platform evaluates 17 large-scale models and conducts experiments using five distinct datasets from the domains of fashion, news, video, books, and music. RecBench provides an in-depth evaluation to advance the research and development of large language models in recommender systems.




