DaRec: A Disentangled Alignment Framework for Large Language Model and Recommender System
直接蒸馏语义表征和协同表征两者对齐得不够好。先解耦,直接将llm表征和协同模型的表征分解;之后进行全局和局部对齐
Large Language Model Empowered Embedding Generator for Sequential Recommendation
使用LLM生成表征给传统推荐模型用。还是得用LLM做推荐语义的理解。
Large Language Models meet Collaborative Filtering: An Efficient All-round LLM-based Recommender System
KDD‘24
LLM在warm的场景效果不行
利用预训练好的协同模型的信息。不要训练cf和llm
与CoLLM: Integrating Collaborative Embeddings into Large Language Models for Recommendation发现的问题一致
Large Language Models Enhanced Collaborative Filtering
通过LLM来提供更好的协同过滤的信息
ctr任务,llm表征喂给cf模型
Collaborative Knowledge Fusion: A Novel Approach for Multi-task Recommender Systems via LLMs
重点还是LLM忽视传统的协同信息
Real-Time Personalization for LLM-based Recommendation with Customized In-Context Learning
训练完LLM推荐后,引入ICL适应新的用户兴趣,而不是重新训练LLM。有一个很大的问题在于为啥不直接扩张用户序列,而是利用ICL的方式。
Beyond Utility: Evaluating LLM as Recommender
评估LLM作为推荐系统的适用性。验证了很多共识。
Observation 1. Overall, current LLMs are less accurate than
traditional models, but they can exhibit greater accuracy in
domains where they possess more extensive knowledge.
Observation 2. LLMs are adept at recommending more niche
items correctly.
Observation 3. LLMs require only brief histories to perform
well, while longer histories do not always benefit LLMs. LLMs
can beat the traditional models in the cold-start scenario.
Observation 4. LLMs suffer from a severe candidate position
bias, favoring items at the beginning most of the time.
Observation 5. LLMs can generate user profiles that capture
a majority of the key patterns useful for recommendations,
which can help enhance the recommendation interpretability.
Observation 6. In general, most LLMs tend to generate below 5% of non-existent items, while some LLMs significantly
hallucinate more items.
Observation 7. Compared to ranking tasks, LLMs are better
at re-ranking tasks regarding utility and beyond.
Collaborative Cross-modal Fusion with Large Language Model for Recommendation
与collm相似,序列也加入了cf embedding,改进了融合机制,文本embedding融入cf embedding的信息,提出两阶段训练
Enhancing Content-based Recommendation via Large Language Model
cf和llm 结合的方式
- cf产生的表征给LLM使用进行推荐:A-LLMRec(KDD’24)Large Language Models meet Collaborative Filtering: An Efficient All-round LLM-based Recommender System, CoLLM: Integrating Collaborative Embeddings into Large Language Models for Recommendation,LLaRA: Large Language-Recommendation Assistant,Text-like Encoding of Collaborative Information in Large Language Models for Recommendation,Collaborative Knowledge Fusion: A Novel Approach for Multi-task Recommender Systems via LLMs,Collaborative Cross-modal Fusion with Large Language Model for Recommendation
- LLM产生的表征给cf使用进行推荐: Large Language Model Empowered Embedding Generator for Sequential Recommendation, Large Language Models Enhanced Collaborative Filtering,
- LLM与cf表征对齐后使用cf模型推荐:DaRec: A Disentangled Alignment Framework for Large Language Model and Recommender System,
- LLM直接微调文本推荐:TALLRec: An Effective and Efficient Tuning Framework to Align Large Language Model with Recommendation,CALRec: Contrastive Alignment of Generative LLMs For Sequential Recommendation
- LLM提取表征,LLM使用进行推荐:HLLM: Enhancing Sequential Recommendations via Hierarchical Large
Language Models for Item and User Modeling,Breaking the Length Barrier: LLM-Enhanced CTR Prediction in Long Textual User Behaviors,Knowledge Adaptation from LLM to Recommendation for Practical Industrial Application