LiteRAG：低成本图检索增强生成方法
原标题：LiteRAG: Cost-Efficient Graph-Based Retrieval-Augmented Generation
AI 摘要
研究者提出 LiteRAG，一种基于图的检索增强生成方法，用查询条件驱动的算法探索和推理链上下文构建替代检索时昂贵的 LLM 控制。在 DistComp 多跳检索基准上，LiteRAG 取得最高整体质量（0.798），相比 GraphRAG Global 和 DRIFT 将单查询延迟降低超过 100 倍、成本降低超过 99%；在 UltraDomain 上与 LinearRAG 质量相当但少用约 14 倍 token。消融实验表明查询自适应阈值和社区感知枢纽惩罚是 token 效率提升的主要来源。
正文节选
LiteRAG: Cost-Efficient Graph-Based Retrieval-Augmented Generation Abstract Graph-based retrieval can improve multi-hop question answering, but existing approaches often incur high query-time costs and produce diffuse, oversized contexts that reduce generation efficiency. We present LiteRAG, a graph-based retrieval method that replaces expensive retrieval-time LLM control with query-conditioned algorithmic exploration and reasoning-chain context construction. On DistComp, a benchmark for multi-h