让检索模型直接识别细粒度相关片段，效率远超大模型。
FGR-ColBERT: Identifying Fine-Grained Relevance Tokens During Retrieval
- 在ColBERT基础上融合大模型提炼的细粒度相关信号
- 110M小模型达64.5分token级F1，超27B大模型
- 保持高效检索能力，延迟仅增加1.12倍
文档检索能找出相关文档，但无法提供如具体相关片段等细粒度证据。一种解决方案是在检索后使用大语言模型（LLM），但这带来显著计算开销，限制实际部署。本文提出FGR-ColBERT，对ColBERT检索模型进行改进，将从LLM中蒸馏出的细粒度相关信号直接融入检索函数。在MS MARCO数据集上的实验表明，FGR-ColBERT（110M）达到64.5的token级F1，超过Gemma 2（27B）的62.8，且模型规模仅为后者的约1/245。同时，其检索有效性保持良好（相对Recall@50为99%），效率高，相比原始ColBERT仅增加约1.12倍延迟。
原文摘要 · Abstract (English)
Document retrieval identifies relevant documents but does not provide fine-grained evidence cues, such as specific relevant spans. A possible solution is to apply an LLM after retrieval; however, this introduces significant computational overhead and limits practical deployment. We propose FGR-ColBERT, a modification of ColBERT retrieval model that integrates fine-grained relevance signals distilled from an LLM directly into the retrieval function. Experiments on MS MARCO show that FGR-ColBERT (110M) achieves a token-level F1 of 64.5, exceeding the 62.8 of Gemma 2 (27B), despite being approximately 245 times smaller. At the same time, it preserves retrieval effectiveness (99% relative Recall@50) and remains efficient, incurring only a ~1.12x latency overhead compared to the original ColBERT.
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。