Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Large Language Models for Generative Recommendation: A Survey and Visionary Discussions

Lei Li, Yongfeng Zhang, Dugang Liu, Li Chen

TL;DR

This paper surveys the emergence of generative recommendation powered by large language models (LLMs), arguing for a shift from multi-stage discriminative pipelines to end-to-end generation of recommendations via token-based IDs. It defines ID representations for users and items, and reviews three ID creation methods (SVD-based, collaborative indexing, and residual-quantized VAE) that capture collaborative information in compact token sequences. The survey details seven generative tasks—rating prediction, top-N and sequential recommendations, explainable recommendations, review generation and summarization, conversational recommendation—and outlines evaluation protocols, challenges, and opportunities, including hallucination, bias, transparency, controllability, efficiency, and multimodality. It also discusses the role of LLM-based agents and retrieval-augmented approaches as practical pathways for real-world deployment, and provides a pragmatic framework and future directions for research in LLM-driven generative RS.

Abstract

Large language models (LLM) not only have revolutionized the field of natural language processing (NLP) but also have the potential to reshape many other fields, e.g., recommender systems (RS). However, most of the related work treats an LLM as a component of the conventional recommendation pipeline (e.g., as a feature extractor), which may not be able to fully leverage the generative power of LLM. Instead of separating the recommendation process into multiple stages, such as score computation and re-ranking, this process can be simplified to one stage with LLM: directly generating recommendations from the complete pool of items. This survey reviews the progress, methods, and future directions of LLM-based generative recommendation by examining three questions: 1) What generative recommendation is, 2) Why RS should advance to generative recommendation, and 3) How to implement LLM-based generative recommendation for various RS tasks. We hope that this survey can provide the context and guidance needed to explore this interesting and emerging topic.

Large Language Models for Generative Recommendation: A Survey and Visionary Discussions

TL;DR

This paper surveys the emergence of generative recommendation powered by large language models (LLMs), arguing for a shift from multi-stage discriminative pipelines to end-to-end generation of recommendations via token-based IDs. It defines ID representations for users and items, and reviews three ID creation methods (SVD-based, collaborative indexing, and residual-quantized VAE) that capture collaborative information in compact token sequences. The survey details seven generative tasks—rating prediction, top-N and sequential recommendations, explainable recommendations, review generation and summarization, conversational recommendation—and outlines evaluation protocols, challenges, and opportunities, including hallucination, bias, transparency, controllability, efficiency, and multimodality. It also discusses the role of LLM-based agents and retrieval-augmented approaches as practical pathways for real-world deployment, and provides a pragmatic framework and future directions for research in LLM-driven generative RS.

Abstract

Large language models (LLM) not only have revolutionized the field of natural language processing (NLP) but also have the potential to reshape many other fields, e.g., recommender systems (RS). However, most of the related work treats an LLM as a component of the conventional recommendation pipeline (e.g., as a feature extractor), which may not be able to fully leverage the generative power of LLM. Instead of separating the recommendation process into multiple stages, such as score computation and re-ranking, this process can be simplified to one stage with LLM: directly generating recommendations from the complete pool of items. This survey reviews the progress, methods, and future directions of LLM-based generative recommendation by examining three questions: 1) What generative recommendation is, 2) Why RS should advance to generative recommendation, and 3) How to implement LLM-based generative recommendation for various RS tasks. We hope that this survey can provide the context and guidance needed to explore this interesting and emerging topic.
Paper Structure (27 sections, 1 figure, 2 tables)

This paper contains 27 sections, 1 figure, 2 tables.

Table of Contents

  1. Introduction
  2. Why Generative Recommendation
  3. ID Creation Methods
  4. Singular Value Decomposition
  5. Collaborative Indexing
  6. Residual-Quantized Variational AutoEncoder
  7. How to Do Generative Recommendation
  8. Rating Prediction
  9. Top-$N$ Recommendation
  10. Sequential Recommendation
  11. Explainable Recommendation
  12. Review Generation
  13. Review Summarization
  14. Conversational Recommendation
  15. Evaluation Protocols
  16. ...and 12 more sections

Figures (1)

  • Figure 1: Pipeline comparison between traditional recommender systems and LLM-based generative recommendation.

Theorems & Definitions (2)

  • Definition 1: ID in Recommender Systems
  • Definition 2: Generative Recommendation