导航启发式嵌入：基于智能体搜索轨迹的密集检索器适配
Navigation-Informed Embeddings: Dense-Retriever Adaptation from Agent Search Traces
浏览论文内容
中文总结 AI 辅助
该研究提出 NIE 方法，利用智能体检索轨迹无需额外标注即可适配密集检索器，在基准任务上提升了 Recall@20、长路径性能及 nDCG@10 指标，提供轻量适配通道。
中文摘要 AI 辅助
智能体检索工作流在回答问题时会生成查询、检索和停止轨迹作为副产品。我们研究如何利用这些轨迹在无需新相关性标签、合成查询或大语言模型（LLM）判断的情况下，将已部署的密集检索器适配到变化的工作流分布。我们提出导航启发式嵌入（Navigation-Informed Embeddings，NIE），这是一系列源自轨迹的目标函数。NIE-Stop 将停止检索的文档作为软正样本；NIE-Path 则额外将路径中前面的文档作为硬对比样本，并施加带几何衰减的序约束。从保留的源轨迹适配的 BGE 编码器，在独立目标基准上将支持度 Recall@20 从 72.2 提升至总体 78.0；NIE-Stop 达到总体 76.9，长路径场景为 52.3；NIE-Path 将长路径性能提升至 55.4，而未适配编码器的长路径性能为 46.7。在完整路径目标下的乱序对照实验中，性能损失 3.2 个百分点。在未使用公共基准训练的情况下，同一适配器在标准 BEIR HotpotQA 上也将 nDCG@10 提升了 1.9 个百分点。因此，NIE 为已保留轨迹且无需额外标注成本的场景提供了轻量适配通道。
英文摘要
Agentic retrieval workflows produce query, retrieval, and stopping traces as a byproduct of answering questions. We study how these traces can adapt a deployed dense retriever to changing workflow distributions without new relevance labels, synthetic queries, or LLM judgments. We introduce Navigation-Informed Embeddings (NIE), a family of trace-derived objectives. NIE-Stop turns the stopping document into a soft positive; NIE-Path additionally uses preceding path documents as hard comparisons and imposes ordinal constraints with geometric decay. A BGE encoder adapted from retained source trajectories improves support Recall@20 on an independent target benchmark from 72.2 to 78.0 overall. NIE-Stop reaches 76.9 overall and 52.3 on long paths; NIE-Path raises long-path performance to 55.4, compared with 46.7 for the unadapted encoder. A shuffled-order control under the full path objective loses 3.2 points. Without public-benchmark training, the same adapter also improves nDCG@10 by 1.9 points on standard BEIR HotpotQA. NIE therefore provides a lightweight adaptation channel for settings where trajectories are already retained, with zero incremental labeling cost.
