Paper Summary: Do LLMs Understand User Preferences? Evaluating LLMs on User Rating Prediction
This is a summary post of this paper: https://arxiv.org/abs/2305.06474 Why this paper/ Goal of this paper? As LLMs have the following properties: Large-scale knowledge and real-world information Strong generalization ability through effective few-shot learning Strong reasons capability with chain-of-thought, self-consistency, etc. Key question answered by this paper: Can we use LLMs for recommender systems? RQ1: Do off-the-shelf LLMs perform well for zero-shot and few-shot recommendations? RQ2: How do LLMs compare with the traditional recommenders in a fair setting?...