发表机构
Capsens（Capsens）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
本研究针对法语PDF问答，提出五种RAG变体，发现页面选择比语义检索更关键，完整系统在MRR@10和Recall@10上显著优于BM25基线。
AI 中文摘要
我们研究了针对法语PDF文档问题的检索增强生成（RAG），其中答案及其支持文档页面均被评估。五个系统变体在BM25基线基础上增加了稠密检索、排名融合、重排序和查询分解。在595个挑战性问题上，完整系统的MRR@10得分为0.4450，Recall@10为0.4013，而BM25分别为0.3430和0.2994。仅使用稠密检索以及简单的词法-稠密融合均不如BM25。重排序改善了混合系统，而添加查询分解带来了最大的进一步增益，但延迟更高且检测到的输出伪影更多。完整系统在两个答案指标上略微超过报告中的匿名总体均值，但在大多数页面检索指标上低于该均值。这些结果指出，准确的页面选择（而非孤立的语义检索）是此设置中主要的改进机会。
英文摘要
We study retrieval-augmented generation (RAG) for questions about French PDF documents when both the answer and its supporting document pages are evaluated. Five system variants add dense retrieval, rank fusion, reranking, and query decomposition to a BM25 baseline. On 595 challenge questions, the complete system scores 0.4450 MRR@10 and 0.4013 Recall@10, compared with 0.3430 and 0.2994 for BM25. Dense retrieval alone and a simple lexical--dense fusion both underperform BM25. Reranking improves the hybrid system, whereas adding query decomposition produces the largest further gain, with higher latency and more detected output artifacts. The complete system slightly exceeds the reported anonymous overall mean on two answer metrics but falls below it on most page-retrieval metrics. These results identify accurate page selection, rather than semantic retrieval in isolation, as the main opportunity for improvement in this setting.
