大模型驱动的推荐前沿进展.pdf

1、Large Language Models for Recommendation:Progresses and Future DirectionFuli Feng USTCThanks to Jizhi Zhang,Keqin Bao,Yang Zhang,Wenjie Wang,Xiangnan He1YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024Background of RecSys2q Information explosion eraE-commerc

2、e:12 million items in Amazon.Social networks:2.8 billion users in Facebook.Content sharing platforms:720,000 hours videos uploaded to Youtube per day；35 million videos posted on TikTok dailyq Recommender systemInformation seekingvia user history feedbackRecommendationImages from:Deep neural networks

3、 for youtube recommendationsYSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024Background of RecSys3RecommendationsInteractionsUser feedbackRecommenderTrainingInferenceSystem sideUser sideItemdatabaseUserq Workflow of Recommender System(1)Train recommender on c

4、ollected interaction data to capture user preferences.(2)Recommender genrates recommendations based on estimated preferences.(3)User engage with the recommended tiems,forming new data,affected by open world.(4)train recommender with new data again,either refining user interests or capturing new ones

5、.Open worldYSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024Background of RecSys4q Core idea of personalized recommendationCollaborative filtering(CF):Making automatic predictions(filtering)about the interests of a user by collecting preferences from many use

6、rs(collaborating).Images from:Neural Collaborative Filtering,Memory-based CFUser CFItem CFModel-based CFMF FISM YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024YSSNLP2024Background of RecSys5q Core idea of personalized recommendationCollaborative filtering(CF):Making