Neither Black nor White: Balancing Semantic and Collaborative Signals with Graph-Informed Semantic IDs (GrIS)

2026-10-01

AI Summary

This study addresses the oversight of collaborative signals in semantic ID construction for generative recommendation by proposing GrIS, a unified framework that reformulates semantic ID generation as recursive graph clustering. By integrating both semantic and collaborative signals, GrIS achieves hierarchical graph partitioning. Theoretically, we demonstrate that existing methods are incomplete special cases within this graph-based framework; decoupling graph construction from partitioning algorithms enables independent optimization along multiple axes and flexible combinations. Technically, the framework introduces differentiable graph pooling (RecDMoN) and a graph-aware RQ-VAE (RQ-GAE). Extensive experiments show that GrIS significantly outperforms state-of-the-art baselines across multiple datasets, achieving up to a 52% improvement in Hit@10.

Abstract

Existing work on Semantic IDs (SIDs) for generative recommendation treats SID construction as a representation learning problem: encode items into a quantised latent space and read off codes. We argue this view is incidental. SID construction is, at heart, a recursive clustering problem, and once stated this way the natural object to cluster is a graph whose nodes carry semantic content and whose edges carry collaborative signal; SID assignment becomes a hierarchical graph partition. This reframing yields a unified framework, Graph-Informed Semantic IDs (GrIS), that subsumes prior approaches rather than displacing them. RQ-VAE and RQ-KMeans are recovered as the special case where the graph is empty, exposing content-only quantisation as one corner of a larger design space along two so-far-collapsed axes: graph construction and recursive partition algorithm. We explore two contrasting instantiations: RecDMoN, which performs hierarchical assignment via differentiable graph pooling, and RQ-GAE, which extends RQ-VAE with graph-aware item representations and a graph reconstruction objective. On multiple real-world datasets, GrIS consistently improves over CF-aware SOTA, with gains of up to +52\% Hit@10. Because graph construction and partition are explicit, separately configurable components, improvements on either axis can be combined and evaluated systematically.

Problem

• Semantic IDs
• Generative Recommendation
• Recursive Clustering
• Graph Partition
• Collaborative Signal

Innovation

• Semantic IDs
• Generative Recommendation
• Hierarchical Graph Partitioning
• Differentiable Graph Pooling
• Graph Autoencoder