arXiv:2510.17139v3 Announce Type: replace Abstract: Recent advances in large language models (LLMs) have led to a surge of interest in query augmentation for information retrieval (IR). Two main approaches have emerged. The first prompts LLMs to generate answers or pseudo-documents that serve as new queries, relying purely on the model's parametric knowledge or contextual information. The second applies reinforcement learning (RL) to fine-tune LLMs for query rewriting, directly optimizing retrieval metrics. While having respective advantages and limitations, the two approaches have not been co
The rapid advancement and integration of large language models into information retrieval systems necessitate continuous optimization research to maximize their effectiveness.
This research outlines improved methods for LLMs to generate more effective queries, directly impacting the quality and efficiency of search and information access.
The proposed 'on-policy optimization' offers a potentially more robust and adaptable framework for LLMs to refine their query augmentation strategies, moving beyond static prompts or purely RL-driven fine-tuning.
- · Information Retrieval developers
- · Search engine companies
- · AI research institutions
- · Companies relying on less sophisticated query augmentation methods
- · Users dealing with inefficient search results
Improved relevance and user satisfaction in search engines and information retrieval systems.
Reduced computational costs for achieving high-quality search results as query generation becomes more efficient.
Accelerated development of more fluid and intuitive human-computer interaction through advanced conversational AI search interfaces.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.CL