QP-OneModel: A Unified Generative LLM for Multi-Task Query Understanding in Xiaohongshu Search

2026-02-10

AI Summary

This work addresses the challenge of reconciling informal language usage with business-specific constraints in social search on Xiaohongshu by proposing the first unified generative large language model framework for multi-task query understanding. The approach reformulates intent recognition, named entity recognition (NER), and term weighting as a unified sequence generation problem. It employs a three-stage alignment strategy and multi-reward reinforcement learning to jointly optimize semantic understanding and business objectives, enhanced by structured intent descriptions as high-fidelity semantic signals. Experimental results demonstrate a 7.35% overall improvement in offline evaluation, with NER and term weighting F1 scores increasing by 9.01% and 9.31%, respectively. The model also outperforms a 32B-parameter baseline by 7.60% on unseen tasks. Online A/B tests show significant gains in DCG (+0.21%) and user retention (+0.044%).

Technology Category

• Natural Language Processing: Generation
• Search and Optimization: Learning to Search
• Machine Learning: Large Multimodal Models (LMMs)

Application Category

• Search and Retrieval-Augmented AI: Large language models for search
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Social Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Query Processing (QP) bridges user intent and content supply in large-scale Social Network Service (SNS) search engines. Traditional QP systems rely on pipelines of isolated discriminative models (e.g., BERT), suffering from limited semantic understanding and high maintenance overhead. While Large Language Models (LLMs) offer a potential solution, existing approaches often optimize sub-tasks in isolation, neglecting intrinsic semantic synergy and necessitating independent iterations. Moreover, standard generative methods often lack grounding in SNS scenarios, failing to bridge the gap between open-domain corpora and informal SNS linguistic patterns, while struggling to adhere to rigorous business definitions. We present QP-OneModel, a Unified Generative LLM for Multi-Task Query Understanding in the SNS domain. We reformulate heterogeneous sub-tasks into a unified sequence generation paradigm, adopting a progressive three-stage alignment strategy culminating in multi-reward Reinforcement Learning. Furthermore, QP-OneModel generates intent descriptions as a novel high-fidelity semantic signal, effectively augmenting downstream tasks such as query rewriting and ranking. Offline evaluations show QP-OneModel achieves a 7.35% overall gain over discriminative baselines, with significant F1 boosts in NER (+9.01%) and Term Weighting (+9.31%). It also exhibits superior generalization, surpassing a 32B model by 7.60% accuracy on unseen tasks. Fully deployed at Xiaohongshu, online A/B tests confirm its industrial value, optimizing retrieval relevance (DCG) by 0.21% and lifting user retention by 0.044%.

Problem

• Query Processing
• Social Network Service
• Large Language Models
• Multi-Task Understanding
• Semantic Gap

Innovation

• Unified Generative LLM
• Multi-Task Query Understanding
• Sequence Generation Paradigm
• Intent Description
• Reinforcement Learning Alignment