Temel Kavramlar
The authors propose a Bi-level Learnable Large Language Model Planning framework to enhance long-term recommendation by incorporating planning capabilities into the decision-making process. This approach combines macro-learning and micro-learning mechanisms to improve personalized recommendations.
Özet
The content discusses the importance of incorporating planning capabilities into recommendation systems to address long-term engagement issues. The proposed Bi-level Learnable LLM Planner framework leverages large language models for personalized recommendations, showcasing superior performance in learning to plan for long-term recommendations.
Traditional recommendation systems tend to focus on immediate interests, neglecting long-term engagement. Reinforcement Learning struggles with sparse data, leading to suboptimal performance in planning. Large Language Models offer powerful planning capabilities through pre-training on diverse textual data.
The proposed Bi-level Learnable LLM Planner framework integrates macro-learning and micro-learning mechanisms for enhanced planning in personalized recommendations. Extensive experiments validate the framework's superiority in learning to plan for long-term recommendations.
İstatistikler
Traditional recommendation systems tend to focus on immediate responses (e.g., clicks) [5].
Reinforcement Learning can struggle with sparse data, affecting planning ability [15, 16, 37].
Large Language Models have powerful planning capabilities through pre-training on textual data [1, 35, 41].