A Review of Modern Recommender Systems Using Generative Models (Gen-RecSys)

TL;DR

This survey surveys the expansion of recommender systems beyond tradition by integrating generative models across interaction data, text, and multimodal content. It systematically organizes methods into auto-encoding, autoregressive, GANs, diffusion, and other generative paradigms, and then extends to LLM-based, retrieval-augmented, and multimodal approaches, including conversational interfaces. It also articulates comprehensive evaluation strategies for offline and online performance, societal impact, and holistic welfare, outlining open challenges and future directions. The work aims to provide a broad, taxonomy-driven framework to guide Gen-RecSys research, development, and responsible deployment in real-world systems.

Abstract

Traditional recommender systems (RS) typically use user-item rating histories as their main data source. However, deep generative models now have the capability to model and sample from complex data distributions, including user-item interactions, text, images, and videos, enabling novel recommendation tasks. This comprehensive, multidisciplinary survey connects key advancements in RS using Generative Models (Gen-RecSys), covering: interaction-driven generative models; the use of large language models (LLM) and textual data for natural language recommendation; and the integration of multimodal models for generating and processing images/videos in RS. Our work highlights necessary paradigms for evaluating the impact and harm of Gen-RecSys and identifies open challenges. This survey accompanies a tutorial presented at ACM KDD'24, with supporting materials provided at: https://encr.pw/vDhLq.

Figures (1)

• Figure 1: Overview of the areas of interest in generative models in recommendation.