RankGR: Rank-Enhanced Generative Retrieval with Listwise Direct Preference Optimization in Recommendation

TL;DR

RankGR introduces a two-phase, listwise preference-optimized generative retrieval method that enhances recommendation accuracy by modeling user preferences more comprehensively and refining candidate evaluation, achieving high scalability and real-time performance.

Contribution

The paper proposes RankGR, a novel generative retrieval framework with listwise preference optimization and a two-stage process, improving recommendation quality and scalability.

Findings

Outperforms existing methods on research datasets.

Achieves near ten-thousand requests per second in real-time deployment.

Demonstrates significant online gains on Taobao's recommendation system.

Abstract

Generative retrieval (GR) has emerged as a promising paradigm in recommendation systems by autoregressively decoding identifiers of target items. Despite its potential, current approaches typically rely on the next-token prediction schema, which treats each token of the next interacted items as the sole target. This narrow focus 1) limits their ability to capture the nuanced structure of user preferences, and 2) overlooks the deep interaction between decoded identifiers and user behavior sequences. In response to these challenges, we propose RankGR, a Rank-enhanced Generative Retrieval method that incorporates listwise direct preference optimization for recommendation. RankGR decomposes the retrieval process into two complementary stages: the Initial Assessment Phase (IAP) and the Refined Scoring Phase (RSP). In IAP, we incorporate a novel listwise direct preference optimization strategy into GR, thus facilitating a more comprehensive understanding of the hierarchical user preferences and more effective partial-order modeling. The RSP then refines the top-{\lambda} candidates generated by IAP with interactions towards input sequences using a lightweight scoring module, leading to more precise candidate evaluation. Both phases are jointly optimized under a unified GR model, ensuring consistency and efficiency. Additionally, we implement several practical improvements in training and deployment, ultimately achieving a real-time system capable of handling nearly ten thousand requests per second. Extensive offline performance on both research and industrial datasets, as well as the online gains on the "Guess You Like" section of Taobao, validate the effectiveness and scalability of RankGR.