任旭滨-LLMs Enhanced Rec 6.22.pdf

《任旭滨-LLMs Enhanced Rec 6.22.pdf》由会员分享，可在线阅读，更多相关《任旭滨-LLMs Enhanced Rec 6.22.pdf（25页珍藏版）》请在三个皮匠报告上搜索。

1、Enhancing Graph-based Recommendation with Large Language ModelsXubin RenThe University of Hong Kong2LLMs+Graph in Recommendation1 Ren,Xubin,et al.A Survey of Large Language Models for Graphs.KDD 2024(a)GNNs as Prefix(c)LLMs-Graphs Interaction(b)LLMs as Prefix(d)LLMs-Only3LLMs+Graph in Recommendation

2、1 Ren,Xubin,et al.A Survey of Large Language Models for Graphs.KDD 20244Representation Learning withLarge Language Models for Recommendation1 Ren,Xubin,et al.Representation Learning with Large Language Models for Recommendation.WWW 20245Background用户-商品交互图推荐算法用户/商品协同过滤特征表示噪音有噪的表征学习用户/商品协同过滤特征表示用户/商品文

3、本模态特征表示促进融合表征中对推荐有益的部分1 Ren,Xubin,et al.Representation Learning with Large Language Models for Recommendation.WWW 20246理论方法最大化(;)critic function1.如何有效地获得高质量的文本模态的特征表示2.如何有效地建模critic function1 Ren,Xubin,et al.Representation Learning with Large Language Models for Recommendation.WWW 20242 Poole,Ben,et

4、 al.On variational bounds of mutual information.ICML 20193 Oord,Aaron van den,et al.Representation learning with contrastive predictive coding.arXiv preprint7 基于大模型的文本特征获取首先要有高质量的文本内容商品画像：描述其会吸引哪一类的用户群体用户画像：描述其会喜欢什么类别的商品Item-to-User 生成范式商品画像生成,Prompts 构建基于 商品描述 生成基于 商品属性&用户反馈 生成用户画像生成,Prompts 构建基于 商

5、品画像&用户反馈 生成核心是要描述出用户/商品的True preference1 Ren,Xubin,et al.Representation Learning with Large Language Models for Recommendation.WWW 20248 基于大模型的文本特征获取其次要有高质量的特征表示文本描述Embedder用户/商品文本模态特征表示ContrieverInstructorText-embedding-ada-0021 Ren,Xubin,et al.Representation Learning with Large Language Models for

6、 Recommendation.”WWW 20242 Izacard,Gautier,et al.Unsupervised dense information retrieval with contrastive learning.TMLR 20223 Su,Hongjin,et al.One embedder,any task:Instruction-finetuned text embeddings.ACL 20239建模critic function对比式对齐（Contrastive Alignment，RLMRec-Con）生成式对齐（Generative Alignment，RLMR