Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SODA: Semantic-Oriented Distributional Alignment for Generative Recommendation

Ziqi Xue, Dingxian Wang, Yimeng Bai, Shuai Zhu, Jialei Li, Xiaoyan Zhao, Frank Yang, Andrew Rabinovich, Yang Zhang, Pablo N. Mendes

TL;DR

Semantic-Oriented Distributional Alignment (SODA) is introduced, a plug-and-play contrastive supervision framework based on Bayesian Personalized Ranking, which aligns semantically rich distributions via negative KL divergence while enabling end-to-end differentiable training.

Abstract

Generative recommendation has emerged as a scalable alternative to traditional retrieve-and-rank pipelines by operating in a compact token space. However, existing methods mainly rely on discrete code-level supervision, which leads to information loss and limits the joint optimization between the tokenizer and the generative recommender. In this work, we propose a distribution-level supervision paradigm that leverages probability distributions over multi-layer codebooks as soft and information-rich representations. Building on this idea, we introduce Semantic-Oriented Distributional Alignment (SODA), a plug-and-play contrastive supervision framework based on Bayesian Personalized Ranking, which aligns semantically rich distributions via negative KL divergence while enabling end-to-end differentiable training. Extensive experiments on multiple real-world datasets demonstrate that SODA consistently improves the performance of various generative recommender backbones, validating its effectiveness and generality. Codes will be available upon acceptance.

SODA: Semantic-Oriented Distributional Alignment for Generative Recommendation

TL;DR

Semantic-Oriented Distributional Alignment (SODA) is introduced, a plug-and-play contrastive supervision framework based on Bayesian Personalized Ranking, which aligns semantically rich distributions via negative KL divergence while enabling end-to-end differentiable training.

Abstract

Generative recommendation has emerged as a scalable alternative to traditional retrieve-and-rank pipelines by operating in a compact token space. However, existing methods mainly rely on discrete code-level supervision, which leads to information loss and limits the joint optimization between the tokenizer and the generative recommender. In this work, we propose a distribution-level supervision paradigm that leverages probability distributions over multi-layer codebooks as soft and information-rich representations. Building on this idea, we introduce Semantic-Oriented Distributional Alignment (SODA), a plug-and-play contrastive supervision framework based on Bayesian Personalized Ranking, which aligns semantically rich distributions via negative KL divergence while enabling end-to-end differentiable training. Extensive experiments on multiple real-world datasets demonstrate that SODA consistently improves the performance of various generative recommender backbones, validating its effectiveness and generality. Codes will be available upon acceptance.
Paper Structure (16 sections, 9 equations, 1 figure, 3 tables)

This paper contains 16 sections, 9 equations, 1 figure, 3 tables.

Table of Contents

  1. Introduction
  2. Methodology
  3. Task Formulation
  4. SODA Framework
  5. Supervision Strategy
  6. Distributional Representation
  7. Training Pipeline
  8. Experiment
  9. Experimental Setup
  10. Datasets
  11. Baselines
  12. Evaluation Metrics
  13. Implementation Details
  14. Main Result (RQ1)
  15. Ablation Study (RQ2)
  16. ...and 1 more sections

Figures (1)

  • Figure 1: Overview of the proposed SODA framework. SODA leverages probability distributions over the tokenizer’s multi-layer codebooks as soft semantic representations for auxiliary supervision, enabling fine-grained semantic alignment and joint optimization between the tokenizer and the recommender.