GRASP用分层图结构高效完成多跳问答，省 token 又提准确率。
GRASP: Graph Agentic Search over Propositions for Multi-hop Question Answering
- 将多跳问题分解为依赖感知计划，动态调度子代理数量。
- 在三个数据集上准确率领先，比现有方法少用30%-50%的token。
- 提出‘成功经济’指标，推动效率导向的智能体评估新标准。
Agentic retrieval 通过赋予语言模型自主性，迭代获取证据来提升多跳问答性能。近期工作引入知识图谱以实现结构化遍历，但带来高昂成本：索引时的图构建开销和推理时的令牌消耗叠加。我们提出图命题代理搜索（GRASP），一种在多跳问答中同时优化高准确率与最小令牌使用量的代理系统。不同于执行单一固定查询，GRASP通过将多跳查询分解为依赖感知计划，主动协调检索策略，并根据问题复杂度动态调整子代理数量。每个子代理通过新型三层层次图（实体、命题、段落）探索单跳查询，利用实体层进行精准遍历，命题层通过倒数排名投票实现高召回段落检索。我们在 MuSiQue、2WikiMultihopQA、HotpotQA 上评估了 GRASP，采用两种设置：开放语料检索与扩展上下文推理（LongBench）。在开放检索设置下，GRASP 在 MuSiQue 与 2WikiMultihopQA 上取得最高准确率，且令牌使用量比 IRCoT+HippoRAG2 少 40%-50%。在 LongBench 设置下，其在所有三个数据集上的 EM 与 F1 均领先，且比最先进方法少用 30% 令牌。最后，我们提出‘成功经济’——每正确答案的加权平均令牌成本（按难度加权），并倡导效率意识评估作为智能体问答的标准实践。
原文摘要 · Abstract (English)
Agentic retrieval improves multi-hop question answering by giving language models autonomy to iteratively gather evidence. Recent work augments these systems with knowledge graphs for structured traversal, but this combination introduces significant cost: expensive graph construction at index time and compounding token usage at inference time. We introduce Graph Agentic Search over Propositions (GRASP), an agentic system that simultaneously optimizes for high accuracy and minimal token usage in multi-hop question answering. Rather than executing a rigid, singular query, GRASP actively coordinates its retrieval strategy by decomposing multi-hop queries into dependency-aware plans. This enables GRASP to dynamically scale the number of sub-agents according to the complexity of the problem. Each sub-agent resolves its single-hop query by exploring a novel three-layer hierarchical graph of entities, propositions, and passages, using the entity layer for targeted traversal and the proposition layer for high-recall passage retrieval via reciprocal-rank voting. We evaluate GRASP on MuSiQue, 2WikiMultihopQA, and HotpotQA under two settings: open-corpus retrieval and extended context reasoning (LongBench). GRASP achieves the highest QA accuracy in the open retrieval setting on MuSiQue and 2Wiki while using 40-50 percent fewer tokens than IRCoT+HippoRAG2. Furthermore, GRASP leads on EM and F1 across all three datasets in the LongBench setting while using 30 percent fewer tokens than the next most accurate method. Finally, we introduce success economy - the amortized token cost per correct answer, weighted by difficulty - and advocate for efficiency-aware evaluation as a standard practice for agentic QA.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。