---
title: "Where Search Agents Go Off the Rails: Graph-Grounded Step-Level Verification"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.06662v1"
permalink: "https://feed.astralor.com/items/itm_05eeb040fdf74c99ba90c41f"
published: "2026-11-04T15:37:36.000Z"
score: 54
channel: "ai"
tags: ["AI Agent","Where Search Agents","Off","Rails","Graph-Grounded Step-Level Verification","Search-augmented","Reliable"]
---

# Where Search Agents Go Off the Rails: Graph-Grounded Step-Level Verification

## 速览

Search-augmented reasoning systems can reach correct answers while making errors in intermediate steps. Reliable auditing must check whether evidence $D$ covers the question $Q$, reasoning $T$ addresses $Q$, and $T$ is …

## 关键点

- Search-augmented reasoning systems can reach correct answers while making errors in intermediate steps.
- Reliable auditing must check whether evidence $D$ covers the question $Q$, reasoning $T$ addresses $Q$, and $T$ is grounded in $D$.
- We introduce GraphRail, which represents requirements, evidence, and reasoning as structured claims and compares them without a generative judge or gold answer.

## 全文

Search-augmented reasoning systems can reach correct answers while making errors in intermediate steps. Reliable auditing must check whether evidence $D$ covers the question $Q$, reasoning $T$ addresses $Q$, and $T$ is grounded in $D$. We introduce GraphRail, which represents requirements, evidence, and reasoning as structured claims and compares them without a generative judge or gold answer. Across three multi-hop QA benchmarks and three reasoning frameworks, GraphRail improves error recall over generative verifiers while remaining stable under stochastic extraction. Its diagnoses can also guide online reasoning, providing a traceable way to detect and correct intermediate errors.

---

引用前请核对原文;本文由星际流动整理:https://feed.astralor.com/items/itm_05eeb040fdf74c99ba90c41f