Unifying Ranking and Generation in Query Auto-Completion via Retrieval-Augmented Generation and Multi-Objective Alignment

2026-02-01

AI Summary

This work addresses key limitations of traditional query auto-completion methods—particularly inadequate long-tail coverage, hallucination generation, and safety concerns—by proposing an end-to-end list generation framework that unifies the completion task under a single modeling paradigm. The approach integrates retrieval-augmented generation (RAG), multi-objective direct preference optimization (DPO), and an iterative critique-and-revision mechanism. High-quality training data are synthesized using a combination of rule-based filters, model-based evaluators, and large language model (LLM) judges. To meet stringent latency constraints, a hybrid serving architecture is designed for efficient deployment. Experimental results demonstrate significant improvements in offline metrics, with human evaluation showing a 0.40–0.69 increase in preference scores. Online A/B tests further reveal a 5.44% reduction in keystrokes and a 3.46% increase in suggestion adoption rate.

Technology Category

• Natural Language Processing: Generation
• Search and Optimization: Learning to Search
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Query Auto-Completion (QAC) suggests query completions as users type, helping them articulate intent and reach results more efficiently. Existing approaches face fundamental challenges: traditional retrieve-and-rank pipelines have limited long-tail coverage and require extensive feature engineering, while recent generative methods suffer from hallucination and safety risks. We present a unified framework that reformulates QAC as end-to-end list generation through Retrieval-Augmented Generation (RAG) and multi-objective Direct Preference Optimization (DPO). Our approach combines three key innovations: (1) reformulating QAC as end-to-end list generation with multi-objective optimization; (2) defining and deploying a suite of rule-based, model-based, and LLM-as-judge verifiers for QAC, and using them in a comprehensive methodology that combines RAG, multi-objective DPO, and iterative critique-revision for high-quality synthetic data; (3) a hybrid serving architecture enabling efficient production deployment under strict latency constraints. Evaluation on a large-scale commercial search platform demonstrates substantial improvements: offline metrics show gains across all dimensions, human evaluation yields +0.40 to +0.69 preference scores, and a controlled online experiment achieves 5.44\% reduction in keystrokes and 3.46\% increase in suggestion adoption, validating that unified generation with RAG and multi-objective alignment provides an effective solution for production QAC. This work represents a paradigm shift to end-to-end generation powered by large language models, RAG, and multi-objective alignment, establishing a production-validated framework that can benefit the broader search and recommendation industry.

Problem

• Query Auto-Completion
• hallucination
• long-tail coverage
• safety risks
• retrieve-and-rank

Innovation

• Retrieval-Augmented Generation
• Multi-Objective Alignment
• Direct Preference Optimization
• Query Auto-Completion
• End-to-End List Generation