智能体式RAG中部分答案质量与效用的预测
原标题：Predicting Partial Answer Quality and Utility in Agentic Retrieval-Augmented Generation
AI 摘要
该论文提出一种轨迹内探测框架，用于研究智能体式RAG（Agentic RAG）多跳问答中中间答案状态的变化。作者在每次检索-推理迭代后强制模型生成中间答案，定义“部分答案质量”和“部分效用”两个指标，并发现多数轨迹中仅一次迭代显著提升答案质量，后续迭代收益甚微。基于此，作者构建了部分答案质量预测与部分效用预测任务，实验显示质量预测更可预测（Pearson相关系数高于0.43），并利用预测结果进行早停，可将平均迭代次数减少约11%，同时保留约98%的最终答案质量。
正文节选
Predicting Partial Answer Quality and Utility in Agentic Retrieval-Augmented Generation Abstract. Agentic Retrieval-Augmented Generation (RAG) has become a promising paradigm for multi-hop question answering, where a reasoning model iteratively issues queries to a retriever and incorporates newly retrieved context into subsequent reasoning steps. While this iterative process can improve final answer quality, current evaluations of agentic RAG largely focus on end-to-end outcomes and provide limi