Collaborative Knowledge Fusion for Multi-task Recommender Systems via LLMs: A Novel Approach
This paper introduces CKF, a novel framework that enhances multi-task recommender systems by integrating collaborative knowledge from traditional models into LLMs through personalized mapping and a multi-task tuning strategy.
A Survey on Leveraging Large Language Models for Next-Generation Recommender Systems
Large language models (LLMs) present a paradigm shift in recommender systems, offering enhanced user and item representation, deeper understanding of user behavior, and the potential to bridge the gap between academic research and industrial application.
RosePO: Enhancing LLM-Based Recommender Systems with Personalized Preference Optimization for Helpfulness and Harmlessness
RosePO is a novel framework that leverages personalized preference optimization to improve the helpfulness and harmlessness of LLM-based recommender systems by constructing diverse preference pairs and dynamically adjusting for uncertainty in user preferences.
Mitigating Propensity Bias of Large Language Models for Recommender Systems
The proposed CLLMR framework mitigates the propensity bias of Large Language Models (LLMs) in recommender systems by incorporating structural information into side information representation and employing counterfactual inference to eliminate the biases introduced by LLMs.
© 2024 by Linnk AI