---
title: "结构化边界智能体图探索：面向证据支撑的学术深度搜索"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2608.24809v1"
permalink: "https://feed.astralor.com/items/itm_4c93fcbfe455e7e2a8f9f854"
published: "2026-08-25T16:51:26.000Z"
score: 76
channel: "ai"
tags: ["arXiv","Crase","学术搜索","引用图","深度研究智能体"]
---

# 结构化边界智能体图探索：面向证据支撑的学术深度搜索

## AI 点评

> 这篇论文最值得关注的是把‘可检查性’作为核心设计目标，而不是事后给结果加解释。对于需要严格追溯证据的科研场景，这种有界方法可能比追求全能型智能体更实用。如果你在构建文献综述工具或评估搜索Agent的可靠性，建议仔细看它的剪枝与排序细节。

## 速览

arXiv最新论文提出Crase系统，一种有界且可检查的深度学术搜索替代方案。与传统开放搜索循环不同，Crase仅对搜索引擎查询一次以获取种子论文，沿1.5跳引用邻域扩展，剪除缺乏证据支持的引用边，并用考虑时效性的随机游走排序剩余论文。候选集、保留理由及停止条件在推理前即明确固定。在LitSearch及另一基准上，面对50万篇arXiv论文语料，Crase以约三分之一成本在recall@50上超越专有模型深度研究智能体最多3倍。

## 关键点

- Crase采用一次性查询+局部图扩展，避免开放搜索循环，实现有界推理。
- 通过1.5跳引用邻域与意图剪枝，确保候选论文均有证据支撑。
- 在LitSearch等基准上，recall@50超越专有深度研究智能体最高3倍，成本仅约三分之一。

## 全文

We present Crase, a bounded and inspectable alternative to deep research agents for scholarly search. Instead of an open-ended search loop, Crase queries a search engine once for seed papers, expands them along their 1.5-hop citation neighborhood, prunes citation edges whose claims lack entailment support, and ranks the remaining papers with a recency-aware random walk. This makes the candidate set, the reason each paper is kept, and the stopping condition explicit and fixed before inference. On LitSearch and one further benchmarks over a 500K-paper arXiv corpus, Crase outperforms deep research agents built on proprietary models by up to 3$\times$ recall@50 at roughly a third of the cost.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_4c93fcbfe455e7e2a8f9f854