RAG 检索策略受控对比：重排与后期交互提升科学问答检索质量
原标题：Re-ranking and Late Interaction Drive Retrieval Quality: A Controlled Comparison of RAG Strategies for Scientific Question Answering
AI 摘要
该论文对科学问答场景下的六种 RAG 检索策略进行了受控对比，包括经典稠密检索、LLM 查询改写、查询改写加重排、基于 RRF 的多查询融合、智能体工具调用式检索，以及 ColBERTv2 的后期交互检索。所有流水线共享同一生成器（Llama-3.1-8B-Instruct）、提示词和评测协议，并从 463,971 篇 2024–2025 年 arXiv 论文中检索。作者还发布了由 Llama-3.1-8B-Instruct 生成的 19,484 条合成科学问题数据集及代码，为研究 RAG 设计的成本与质量权衡提供开放测试平台。
正文节选
Re-ranking and Late Interaction Drive Retrieval Quality: A Controlled Comparison of RAG Strategies for Scientific Question Answering Abstract— Retrieval-Augmented Generation (RAG) is now the standard way to ground Large Language Models (LLMs) in external knowledge, yet the design space of retrieval pipelines is large and the trade-offs between variants are not well understood, especially on domain-specific corpora at realistic scale. In this work, we present a controlled comparison of six retrie