利用LLM未选候选词提升查询扩展，兼顾效果与效率
Upcycling Candidate Tokens of Large Language Models for Query Expansion
- 从一次LLM推理中提取未选候选词，实现高效查询扩展
- 在不增加计算成本下，检索性能超越或媲美更昂贵方法
- 适合追求低延迟、高性价比检索系统的开发者
查询扩展（QE）通过添加相关术语提升检索效果。近期，大语言模型（LLMs）被用于QE，但现有方法存在权衡：生成多样词汇可提升性能，却增加计算开销。为此，我们提出候选词查询扩展（CTQE），通过利用单次LLM解码中未选中的候选词，在不增加推理次数的前提下，提取多样且相关的术语。这些未被选中的词虽未进入最终输出，但受完整查询条件约束，仍蕴含有效信息。通过聚合这些词，CTQE在无需额外推理的情况下实现了相关性与多样性的平衡，显著降低开销与延迟。实验表明，CTQE在保持高性能的同时大幅减少成本，表现优于或媲美更昂贵的方法。代码已开源：https://github.com/bluejeans8/CTQE
原文摘要 · Abstract (English)
Query Expansion (QE) improves retrieval performance by enriching queries with related terms. Recently, Large Language Models (LLMs) have been used for QE, but existing methods face a trade-off: generating diverse terms boosts performance but increases computational cost. To address this challenge, we propose Candidate Token Query Expansion (CTQE), which extracts diverse and relevant terms from a single LLM decoding pass by leveraging unselected candidate tokens. These tokens, though not part of the final output, are conditioned on the full query and capture useful information. By aggregating them, CTQE achieves both relevance and diversity without extra inference, reducing overhead and latency. Experiments show that CTQE delivers strong retrieval performance with significantly lower cost, outperforming or comparable to more expensive methods. Code is available at: https://github.com/bluejeans8/CTQE
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。