发表机构
Pennsylvania State University（宾夕法尼亚州立大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
SAGE是面向生产级RAG系统的感知SLO自适应检索策略，通过模仿学习训练，能动态选段落数，在多数据集与LLM上提升SLO合规率、降本减时延且质量损失小。
AI 中文摘要
生产环境中的检索增强生成（RAG）系统需严格满足尾延迟和基础设施成本方面的服务水平目标（SLO）。然而，标准检索流水线依赖固定的检索预算，未考虑查询难度，对简单查询过度检索、对困难查询检索不足，迫使运维人员在答案质量与SLO合规性间做权衡。本文提出SAGE，一种感知SLO的学习型自适应检索策略，可动态为每个查询选择段落数量k。SAGE利用初始检索得到的轻量特征（如分数分布、排名间隙、词汇信号），通过模仿学习从近似最优延迟-质量权衡的专家模型进行离线训练；推理阶段不增加大语言模型（LLM）调用，仅产生极小开销。在Natural Questions数据集上，5秒P95延迟SLO约束下，SAGE的SLO合规率达95%，而最优静态基准（k=20）仅为30%；其P95延迟降低36%，检索成本降低51%，仅损失2个百分点的精确匹配（EM）指标。在Natural Questions上训练的单一策略可泛化至HotpotQA、UnSeenTimeQA数据集及四类LLM（Llama、Qwen、Mistral、Gemma），持续实现45至52点的SLO提升且无质量下降。
英文摘要
Retrieval-Augmented Generation (RAG) systems in production operate under strict service level objectives (SLOs) on tail latency and infrastructure cost. However, standard retrieval pipelines rely on fixed retrieval budgets that ignore query difficulty, over-retrieving for easy queries and under-serving hard ones, forcing operators to trade answer quality against SLO compliance. This paper proposes SAGE, a learned SLO-aware adaptive retrieval policy that dynamically selects the number of passages k per query. SAGE uses lightweight features derived from initial retrieval (e.g., score distributions, rank gaps, lexical signals) and is trained offline via imitation learning from an oracle that approximates optimal latency-quality trade-offs. At inference, it adds no LLM calls and minimal overhead. On Natural Questions, under a 5s P95 latency SLO, SAGE achieves 95% SLO compliance versus 30% for the best static baseline (k=20), reduces P95 latency by 36% and retrieval cost by 51% with only 2 percentage points Exact Match (EM) loss. A single policy trained on Natural Questions generalizes across HotpotQA, UnSeenTimeQA, and four LLM families (Llama, Qwen, Mistral, Gemma), consistently yielding +45-52 point SLO improvements without quality degradation.
Comments7 pages, 5 figures, 2 tables. Authors' accepted version of a paper published in Proc. IEEE CoDIT 2026. The version of record is available at the DOI below
Journal ref2026 12th International Conference on Control, Decision and Information Technologies (CoDIT), Bari, Italy, 2026, pp. 169-175
DOI：10.1109/CoDIT70676.2026.11631166
