From Prompting to Alignment: A Generative Framework for Query Recommendation

2025-04-14

AI Summary

To address cold-start challenges, insufficient long-tail coverage, and task fragmentation in search query recommendation, this paper proposes the first end-to-end, generalizable generative unified framework. Methodologically: (1) instruction-tuned large language models unify suggestion, auto-completion, and query clarification under a single generative paradigm; (2) a novel CTR-aligned reward modeling scheme coupled with list-wise preference optimization—a PPO variant—enables multi-level alignment between generated queries and implicit user click feedback; (3) proactive modeling integrates co-occurrence-based query retrieval with historical log augmentation. Evaluated on multiple real-world datasets, the framework achieves a 23.6% improvement in NDCG@5 under cold-start conditions, 89.4% long-tail query coverage, and a 17.2% average click-through rate gain, significantly outperforming state-of-the-art methods.

Technology Category

• Search and Optimization: Learning to Search
• Natural Language Processing: Generation
• Data Mining & Knowledge Management: Intelligent Query Processing

Application Category

• User Modeling, Personalization and Recommendation: ML for personalized search and recommendations
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Querying, indexing, and retrieval in Web-related graphs

Abstract

In modern search systems, search engines often suggest relevant queries to users through various panels or components, helping refine their information needs. Traditionally, these recommendations heavily rely on historical search logs to build models, which suffer from cold-start or long-tail issues. Furthermore, tasks such as query suggestion, completion or clarification are studied separately by specific design, which lacks generalizability and hinders adaptation to novel applications. Despite recent attempts to explore the use of LLMs for query recommendation, these methods mainly rely on the inherent knowledge of LLMs or external sources like few-shot examples, retrieved documents, or knowledge bases, neglecting the importance of the calibration and alignment with user feedback, thus limiting their practical utility. To address these challenges, we first propose a general Generative Query Recommendation (GQR) framework that aligns LLM-based query generation with user preference. Specifically, we unify diverse query recommendation tasks by a universal prompt framework, leveraging the instruct-following capability of LLMs for effective generation. Secondly, we align LLMs with user feedback via presenting a CTR-alignment framework, which involves training a query-wise CTR predictor as a process reward model and employing list-wise preference alignment to maximize the click probability of the generated query list. Furthermore, recognizing the inconsistency between LLM knowledge and proactive search intents arising from the separation of user-initiated queries from models, we align LLMs with user initiative via retrieving co-occurrence queries as side information when historical logs are available.

Problem

• Addresses cold-start and long-tail issues in query recommendation models
• Unifies diverse query tasks via LLM-based universal prompt framework
• Aligns LLM-generated queries with user feedback and proactive intents

Innovation

• Unified prompt framework for diverse query tasks
• CTR-alignment framework with user feedback
• Retrieve co-occurrence queries for proactive alignment