RecBench+
收藏资源简介:
RecBench+是由香港理工大学构建的一个新数据集,旨在评估大型语言模型作为个性化推荐助手的能力。该数据集包含大约3万个高质量的复杂用户查询,涵盖了不同难度、用户指定的条件数量以及用户画像,反映了现实世界推荐场景中多样化的用户需求。RecBench+根据用户查询的类型,分为基于条件和基于用户画像的查询,以评估推荐助手在不同场景下的性能。该数据集是首个公开可用于有效评估LLM时代个性化推荐助手性能的数据集。
RecBench+ is a novel dataset constructed by The Hong Kong Polytechnic University, designed to evaluate the capabilities of large language models (LLMs) as personalized recommendation assistants. It contains approximately 30,000 high-quality complex user queries, covering varying difficulty levels, numbers of user-specified conditions and user profiles, which reflect diverse user needs in real-world recommendation scenarios. RecBench+ is categorized into condition-based and profile-based queries according to the type of user queries, to assess the performance of recommendation assistants in different scenarios. This dataset is the first publicly available resource for effectively evaluating the performance of personalized recommendation assistants in the LLM era.

- 1Towards Next-Generation Recommender Systems: A Benchmark for Personalized Recommendation Assistant with LLMs香港理工大学 · 2025年



