OneSearch: A Preliminary Exploration of the Unified End-to-End Generative Framework for E-commerce Search

2025-09-03

AI Summary

To address computational fragmentation, misaligned stage-wise objectives, and performance bottlenecks in conventional multi-stage cascaded e-commerce search architectures, this paper proposes OneSearch—the first industrial-scale end-to-end generative search framework. Methodologically, it introduces: (1) keyword-augmented hierarchical quantized encoding for semantic fidelity and efficient retrieval; (2) multi-view user behavioral sequence injection coupled with preference-aware reward modeling for fine-grained intent understanding; and (3) behavior-driven user ID construction, multi-stage supervised fine-tuning, and adaptive weighted ranking. Offline evaluations demonstrate substantial gains over strong baselines. Online A/B tests show statistically significant improvements of +1.67% in item click-through rate, +2.40% in unique buyers, and +3.22% in order volume, alongside an 8.4× increase in compute utilization and a 75.4% reduction in operational costs.

Technology Category

• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Search
• Machine Learning: Learning Preferences or Rankings

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions
• User Modeling, Personalization and Recommendation: ML for personalized search and recommendations

Abstract

Traditional e-commerce search systems employ multi-stage cascading architectures (MCA) that progressively filter items through recall, pre-ranking, and ranking stages. While effective at balancing computational efficiency with business conversion, these systems suffer from fragmented computation and optimization objective collisions across stages, which ultimately limit their performance ceiling. To address these, we propose extbf{OneSearch}, the first industrial-deployed end-to-end generative framework for e-commerce search. This framework introduces three key innovations: (1) a Keyword-enhanced Hierarchical Quantization Encoding (KHQE) module, to preserve both hierarchical semantics and distinctive item attributes while maintaining strong query-item relevance constraints; (2) a multi-view user behavior sequence injection strategy that constructs behavior-driven user IDs and incorporates both explicit short-term and implicit long-term sequences to model user preferences comprehensively; and (3) a Preference-Aware Reward System (PARS) featuring multi-stage supervised fine-tuning and adaptive reward-weighted ranking to capture fine-grained user preferences. Extensive offline evaluations on large-scale industry datasets demonstrate OneSearch's superior performance for high-quality recall and ranking. The rigorous online A/B tests confirm its ability to enhance relevance in the same exposure position, achieving statistically significant improvements: +1.67% item CTR, +2.40% buyer, and +3.22% order volume. Furthermore, OneSearch reduces operational expenditure by 75.40% and improves Model FLOPs Utilization from 3.26% to 27.32%. The system has been successfully deployed across multiple search scenarios in Kuaishou, serving millions of users, generating tens of millions of PVs daily.

Problem

• Replacing fragmented multi-stage e-commerce search systems with unified end-to-end framework
• Addressing computational inefficiency and objective conflicts in traditional search architectures
• Enhancing relevance modeling and user preference capture in e-commerce search

Innovation

• Keyword-enhanced Hierarchical Quantization Encoding module
• Multi-view user behavior sequence injection strategy
• Preference-Aware Reward System with fine-tuning