AgentIR：面向长期对话记忆的工作负载自适应级联检索基座
AgentIR: A Workload-Adaptive Cascade Retrieval Substrate for Long-Term Conversational Memory
AI总结：
针对长期对话记忆中索引动态增长、查询类型变化和亚10毫秒延迟预算的挑战，提出一种基于置信度触发的级联路由策略，通过自适应选择融合方法和是否运行稠密通道，在保持准确率的同时显著降低延迟并提升并发能力。
摘要：
长期对话记忆是一种经典信息检索未针对其设计的检索工作负载：索引在查询流期间增长，查询类型在会话内发生变化，每次检索的延迟预算低于10毫秒。Lucene类引擎将索引视为静态、查询视为无状态，未利用工作负载的结构。AgentIR将融合视为每个查询沿两个轴的决定：应用哪种融合（BM25、稠密、RRF或智能体感知RRF），以及是否值得运行约52毫秒的稠密通道。第二个轴是一个置信度触发的级联路由器，仅根据BM25 top-k间隔做出决定，并在不同工作负载间重新调整而无需重新训练。在LongMemEval（n=500）上，稠密通道确实增加了信息，级联在LLM判断准确率相当的情况下跳过了63%的查询（在两种评判者下快2.67倍，配对bootstrap p>=0.88）；按查询类型阈值在5折交叉验证下将其扩展到5.76倍。在LoCoMo（n=1,982）上，BM25单独已经是最强的单一系统，同一触发器自动调整到100%跳过率（快132倍，+0.089 Hit@5）。在共享8核虚拟机上的容量从约154个并发智能体增加到约1,400个（9倍）。在级联之下，时间分区索引执行O(log 1/epsilon)的工作，与语料库大小无关：1234倍语料库增长仅导致3.6倍延迟，最终在500万条记录上以亚100微秒p50达到顺序搜索的1769倍。在与Lucene在9个BEIR数据集（最多880万文档）上质量相当的情况下，该基座在Pyserini 8T上运行速度几何平均快10倍，在PISA-1T BlockMax-WAND上快11倍；A100达到Pyserini 8T的1.8-39倍；分块索引构建在MS MARCO上维持56.8K文档/秒。记录并修复了三个使nDCG@10静默下降6-8倍的微妙BM25/GPU正确性陷阱；修复后，在适合单个A100的所有八个数据集上，CPU和GPU的nDCG@10差异在0.0002以内。
英文摘要：
Long-term conversational memory is a retrieval workload classical IR was not built for: the index grows during the query stream, query types shift intra-session, and the latency budget per retrieval is sub-10 ms. Lucene-class engines treat the index as static and the query as stateless, leaving the workload's structure unexploited. AgentIR treats fusion as a per-query decision along two axes: which fusion to apply (BM25, Dense, RRF, or agent-aware RRF), and whether the ~52 ms dense channel is worth running at all. The second axis is a confidence-triggered cascade router that decides from the BM25 top-k margin alone and re-tunes across workloads without retraining. On LongMemEval (n=500), where the dense channel does add information, the cascade skips 63% of queries at parity LLM-judged accuracy (2.67x faster under two judges, paired bootstrap p>=0.88); per-qtype thresholds extend this to 5.76x under 5-fold cross-validation. On LoCoMo (n=1,982), where BM25 alone is already the strongest single system, the same trigger auto-tunes to a 100% skip rate (132x faster, +0.089 Hit@5). Capacity on a shared 8-core VM rises from ~154 to ~1,400 concurrent agents (9x). Underneath the cascade, a time-partitioned index does O(log 1/epsilon) work independent of corpus size: 1234x corpus growth costs only 3.6x latency, ending in 1769x over sequential at sub-100 us p50 on 5M records. At parity quality with Lucene on 9 BEIR datasets up to 8.8M docs, the substrate runs 10x geo-mean over Pyserini 8T and 11x over PISA-1T BlockMax-WAND; an A100 reaches 1.8-39x over Pyserini 8T; chunked index build sustains 56.8K docs/sec on MS MARCO. Three subtle BM25/GPU correctness pitfalls that silently regress nDCG@10 by 6-8x are documented and fixed; post-fix CPU and GPU agree within 0.0002 nDCG@10 on all eight datasets that fit a single A100.
机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。
