RSBench
收藏资源简介:
RSBench是由南洋理工大学开发的基准问题集,专门用于评估基于大型语言模型的进化算法在推荐系统中的应用。该数据集包含九个多目标实例,旨在通过会话推荐任务来发现一组帕累托最优提示,以指导推荐过程,提供准确、多样和公平的推荐。RSBench的创建过程结合了经典的进化算法框架和大型语言模型的操作,旨在解决推荐系统中的多目标优化问题,特别是在会话推荐中的应用。
RSBench is a benchmark problem suite developed by Nanyang Technological University (NTU), specifically designed to evaluate evolutionary algorithms powered by large language models (LLMs) for recommender system applications. This dataset comprises nine multi-objective instances, which are designed to discover a set of Pareto-optimal prompts via session recommendation tasks to guide the recommendation workflow and deliver accurate, diverse, and fair recommendations. The development of RSBench integrates classical evolutionary algorithm frameworks and the operational capabilities of large language models, with the goal of addressing multi-objective optimization problems in recommender systems, particularly in session recommendation scenarios.

- 1Language Model Evolutionary Algorithms for Recommender Systems: Benchmarks and Algorithm Comparisons南洋理工大学 · 2024年



