Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation

Dengzhao Fang, Jingtong Gao, Yu Li, Xiangyu Zhao, Yi Chang

TL;DR

A novel generative recommendation framework, PRISM, with Purified Representation and Integrated Semantic Modeling is proposed, which consistently outperforms state-of-the-art baselines across four real-world datasets, demonstrating substantial performance gains, particularly in high-sparsity scenarios.

Abstract

Generative Sequential Recommendation (GSR) has emerged as a promising paradigm, reframing recommendation as an autoregressive sequence generation task over discrete Semantic IDs (SIDs), typically derived via codebook-based quantization. Despite its great potential in unifying retrieval and ranking, existing GSR frameworks still face two critical limitations: (1) impure and unstable semantic tokenization, where quantization methods struggle with interaction noise and codebook collapse, resulting in SIDs with ambiguous discrimination; and (2) lossy and weakly structured generation, where reliance solely on coarse-grained discrete tokens inevitably introduces information loss and neglects items' hierarchical logic. To address these issues, we propose a novel generative recommendation framework, PRISM, with Purified Representation and Integrated Semantic Modeling. Specifically, to ensure high-quality tokenization, we design a Purified Semantic Quantizer that constructs a robust codebook via adaptive collaborative denoising and hierarchical semantic anchoring mechanisms. To compensate for information loss during quantization, we further propose an Integrated Semantic Recommender, which incorporates a dynamic semantic integration mechanism to integrate fine-grained semantics and enforces logical validity through a semantic structure alignment objective. PRISM consistently outperforms state-of-the-art baselines across four real-world datasets, demonstrating substantial performance gains, particularly in high-sparsity scenarios.

PRISM: Purified Representation and Integrated Semantic Modeling for Generative Sequential Recommendation

TL;DR

A novel generative recommendation framework, PRISM, with Purified Representation and Integrated Semantic Modeling is proposed, which consistently outperforms state-of-the-art baselines across four real-world datasets, demonstrating substantial performance gains, particularly in high-sparsity scenarios.

Abstract

Generative Sequential Recommendation (GSR) has emerged as a promising paradigm, reframing recommendation as an autoregressive sequence generation task over discrete Semantic IDs (SIDs), typically derived via codebook-based quantization. Despite its great potential in unifying retrieval and ranking, existing GSR frameworks still face two critical limitations: (1) impure and unstable semantic tokenization, where quantization methods struggle with interaction noise and codebook collapse, resulting in SIDs with ambiguous discrimination; and (2) lossy and weakly structured generation, where reliance solely on coarse-grained discrete tokens inevitably introduces information loss and neglects items' hierarchical logic. To address these issues, we propose a novel generative recommendation framework, PRISM, with Purified Representation and Integrated Semantic Modeling. Specifically, to ensure high-quality tokenization, we design a Purified Semantic Quantizer that constructs a robust codebook via adaptive collaborative denoising and hierarchical semantic anchoring mechanisms. To compensate for information loss during quantization, we further propose an Integrated Semantic Recommender, which incorporates a dynamic semantic integration mechanism to integrate fine-grained semantics and enforces logical validity through a semantic structure alignment objective. PRISM consistently outperforms state-of-the-art baselines across four real-world datasets, demonstrating substantial performance gains, particularly in high-sparsity scenarios.
Paper Structure (24 sections, 18 equations, 7 figures, 4 tables)

This paper contains 24 sections, 18 equations, 7 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Problem Formulation
  5. Overall Pipeline
  6. Purified Semantic Quantizer
  7. Adaptive Collaborative Denoising for Signal Purification
  8. Hierarchical Semantic Anchoring for Latent Stability
  9. Dual-Head Reconstruction and Optimization
  10. Integrated Semantic Recommender
  11. Dynamic Semantic Integration via Mixture-of-Experts
  12. Semantic Structure Alignment for Generative Consistency
  13. Adaptive Temperature Scaling Generation
  14. Experiments
  15. Experimental Setup
  16. ...and 9 more sections

Figures (7)

  • Figure 1: Illustration of critical limitations in existing GSR frameworks. (a) Codebook Collapse: The unstable quantizer tokenizes diverse items into a narrow range of codes, making items indistinguishable. (b) Information Loss: Discrete SIDs fail to capture fine-grained continuous semantics, providing insufficient item features for recommendation.
  • Figure 2: The PRISM framework. PRISM first learns a robust vocabulary via the Purified Semantic Quantizer, and then the Integrated Semantic Recommender utilizes the vocabulary to tokenize items into semantic IDs for generative recommendation.
  • Figure 3: Performance comparison across item popularity groups, where 'n' denotes the number of test interactions.
  • Figure 4: t-SNE visualization of codebook embeddings.
  • Figure 5: t-SNE visualization of item embeddings, where different colors indicate different categories.
  • ...and 2 more figures