UniSearch: Rethinking Search System with a Unified Generative Architecture

2025-09-08

AI Summary

Traditional search cascade architectures suffer from modular fragmentation and misaligned optimization objectives, hindering end-to-end optimization and generalization. This paper introduces UniSearch—a truly end-to-end generative search framework tailored for Kuaishou’s short-video and live-streaming scenarios—where a unified generator directly models the “query → content” mapping. Our key contributions are: (1) a jointly optimized generator-video encoder architecture; (2) Search Preference Optimization (SPO), which integrates user behavioral feedback and reward modeling to align generated outputs with user preferences; and (3) a unified training paradigm combining latent item embeddings, tokenized representations, and behavior-driven reinforcement learning. Online A/B testing demonstrates that UniSearch achieves the largest single-experiment gain in live-stream search to date, validating its effectiveness and scalability for large-scale deployment.

Technology Category

• Search and Optimization: Learning to Search
• Constraint Satisfaction and Optimization: Search
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: ML for personalized search and recommendations
• Responsible Web: Human-perceived consequences of algorithmic deployment on the web

Abstract

Modern search systems play a crucial role in facilitating information acquisition. Traditional search engines typically rely on a cascaded architecture, where results are retrieved through recall, pre-ranking, and ranking stages. The complexity of designing and maintaining multiple modules makes it difficult to achieve holistic performance gains. Recent advances in generative recommendation have motivated the exploration of unified generative search as an alternative. However, existing approaches are not genuinely end-to-end: they typically train an item encoder to tokenize candidates first and then optimize a generator separately, leading to objective inconsistency and limited generalization. To address these limitations, we propose UniSearch, a unified generative search framework for Kuaishou Search. UniSearch replaces the cascaded pipeline with an end-to-end architecture that integrates a Search Generator and a Video Encoder. The Generator produces semantic identifiers of relevant items given a user query, while the Video Encoder learns latent item embeddings and provides their tokenized representations. A unified training framework jointly optimizes both components, enabling mutual enhancement and improving representation quality and generation accuracy. Furthermore, we introduce Search Preference Optimization (SPO), which leverages a reward model and real user feedback to better align generation with user preferences. Extensive experiments on industrial-scale datasets, together with online A/B testing in both short-video and live search scenarios, demonstrate the strong effectiveness and deployment potential of UniSearch. Notably, its deployment in live search yields the largest single-experiment improvement in recent years of our product's history, highlighting its practical value for real-world applications.

Problem

• Unified generative search to replace cascaded architecture
• Address objective inconsistency in existing generative approaches
• Improve generalization and representation quality in search systems

Innovation

• Unified end-to-end generative search architecture
• Joint optimization of generator and video encoder
• Search Preference Optimization with user feedback alignment