Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework

Ben Chen, Siyuan Wang, Yufei Ma, Zihan Liang, Xuxin Zhang, Yue Lv, Ying Yang, Huangyu Dai, Lingtao Mao, Tong Zhao, Zhipeng Qian, Xinyu Sun, Zhixin Zhai, Yang Zhao, Bochao Liu, Jingshan Lv, Xiao Liang, Hui Kong, Jing Chen, Han Li, Chenyi Lei, Wenwu Ou, Kun Gai

Abstract

Generative Retrieval (GR) has emerged as a promising paradigm for modern search systems. Compared to multi-stage cascaded architecture, it offers advantages such as end-to-end joint optimization and high computational efficiency. OneSearch, as a representative industrial-scale deployed generative search framework, has brought significant commercial and operational benefits. However, its inadequate understanding of complex queries, inefficient exploitation of latent user intents, and overfitting to narrow historical preferences have limited its further performance improvement. To address these challenges, we propose \textbf{OneSearch-V2}, a latent reasoning enhanced self-distillation generative search framework. It contains three key innovations: (1) a thought-augmented complex query understanding module, which enables deep query understanding and overcomes the shallow semantic matching limitations of direct inference; (2) a reasoning-internalized self-distillation training pipeline, which uncovers users' potential yet precise e-commerce intentions beyond log-fitting through implicit in-context learning; (3) a behavior preference alignment optimization system, which mitigates reward hacking arising from the single conversion metric, and addresses personal preference via direct user feedback. Extensive offline evaluations demonstrate OneSearch-V2's strong query recognition and user profiling capabilities. Online A/B tests further validate its business effectiveness, yielding +3.98\% item CTR, +3.05\% buyer conversion rate, and +2.11\% order volume. Manual evaluation further confirms gains in search experience quality, with +1.65\% in page good rate and +1.37\% in query-item relevance. More importantly, OneSearch-V2 effectively mitigates common search system issues such as information bubbles and long-tail sparsity, without incurring additional inference costs or serving latency.

OneSearch-V2: The Latent Reasoning Enhanced Self-distillation Generative Search Framework

Abstract

Generative Retrieval (GR) has emerged as a promising paradigm for modern search systems. Compared to multi-stage cascaded architecture, it offers advantages such as end-to-end joint optimization and high computational efficiency. OneSearch, as a representative industrial-scale deployed generative search framework, has brought significant commercial and operational benefits. However, its inadequate understanding of complex queries, inefficient exploitation of latent user intents, and overfitting to narrow historical preferences have limited its further performance improvement. To address these challenges, we propose \textbf{OneSearch-V2}, a latent reasoning enhanced self-distillation generative search framework. It contains three key innovations: (1) a thought-augmented complex query understanding module, which enables deep query understanding and overcomes the shallow semantic matching limitations of direct inference; (2) a reasoning-internalized self-distillation training pipeline, which uncovers users' potential yet precise e-commerce intentions beyond log-fitting through implicit in-context learning; (3) a behavior preference alignment optimization system, which mitigates reward hacking arising from the single conversion metric, and addresses personal preference via direct user feedback. Extensive offline evaluations demonstrate OneSearch-V2's strong query recognition and user profiling capabilities. Online A/B tests further validate its business effectiveness, yielding +3.98\% item CTR, +3.05\% buyer conversion rate, and +2.11\% order volume. Manual evaluation further confirms gains in search experience quality, with +1.65\% in page good rate and +1.37\% in query-item relevance. More importantly, OneSearch-V2 effectively mitigates common search system issues such as information bubbles and long-tail sparsity, without incurring additional inference costs or serving latency.

Paper Structure

This paper contains 39 sections, 19 equations, 6 figures, 15 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Generative Retrieval and Recommendation
  4. Latent Reasoning and Self-Distillation
  5. Preference Alignment for GRs
  6. Methodology
  7. Multimodal or Unimodal SID Tokenization?
  8. Thought-Augmented Query Understanding
  9. Keyword-based CoT
  10. Training Paradigm Refinement
  11. Reasoning-Internalized Self-Distillation
  12. Self-Distillation Formulation
  13. Mitigating Representation Instability
  14. Prediction Consistency via R-Drop.
  15. Input Robustness via Adversarial Perturbation.
  16. ...and 24 more sections

Figures (6)

  • Figure 1: OneSearch-V2 vs. V1. OneSearch-V2 extends the generative search framework with thought-augmented query understanding, reasoning-internalized self-distillation, and behavior feedback preference alignment.
  • Figure 2: The Overall Framework of OneSearch V2. It contains (a) a thought-augmented complex query understanding module, (b) a reasoning-internalized self-distillation training pipeline, and (c) a behavior preference alignment optimization system. OneSearch-V2 effectively mitigates common search system issues such as information bubbles and long-tail sparsity, without incurring additional inference costs or serving latency.
  • Figure 3: Three-step keyword-based CoT extraction pipeline for diverse complex query types, along with the corresponding CoT tasks.
  • Figure 4: The sid rate of the proposed innovations with OneSearch on the industry dataset.
  • Figure 5: The online CTR relative gains for top/middle/tail 10 industries respectively.
  • ...and 1 more figures