---
格式版本: 2
标题: "Projecting BrowseComp-Plus onto ClimbMix: Toward More Realistic Corpora for Agentic Search"
原文链接: "https://arxiv.org/abs/2608.20317"
发布日期: "2026-08-20"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:local:strict_original_body"
发布时间证据: "**\\[v1\\]** Thu, 20 Aug 2026 17:54:48 UTC (284 KB)"
发布时间校准原因: "规则确认唯一严格发布时间，来源 local:strict_original_body"
发布时间校准置信度: "high"
发布时间候选数量: 18
发布时间严格候选数量: 6
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-21T16:14:53+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-21T16:14:13+08:00"
入库时间: "2026-08-21T08:14:53.212Z"
来源平台: "arXiv 学术论文搜索"
搜索渠道: "source_template"
搜索词: "https://arxiv.org/search/?query=NVIDIA&searchtype=all"
匹配关键词:
  []
相关厂家:
  - "NVIDIA"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 0
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文主题为信息检索与Agentic Search语料构建，仅提及NVIDIA发布的ClimbMix语料库，与超节点/AI Rack/机柜级AI基础设施等核心关注点完全无关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-21T16:15:03+08:00"
AI主题相关性: 0
AI来源权威性: 0
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
AI摘要: "研究者将BrowseComp-Plus评测问题迁移到NVIDIA发布的ClimbMix大规模语料上，提出BrowseComp-Plus_CM基准，通过逐跳分解和多重验证实现更真实的智能体搜索评测。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T02:12:24.709Z"
采集批次: "2026年8月21日13点55分53秒"
采集批次ID: "20260821-135553-743"
去重键: "https://arxiv.org/abs/2608.20317"
---

## Computer Science > Information Retrieval

## Title:Projecting BrowseComp-Plus onto ClimbMix: Toward More Realistic Corpora for Agentic Search

Authors:[Sahel Sharifymoghaddam](https://arxiv.org/search/cs?searchtype=author&query=Sharifymoghaddam,+S), [Lingwei Gu](https://arxiv.org/search/cs?searchtype=author&query=Gu,+L), [Yijun Ge](https://arxiv.org/search/cs?searchtype=author&query=Ge,+Y), [Jimmy Lin](https://arxiv.org/search/cs?searchtype=author&query=Lin,+J)

[View PDF](https://arxiv.org/pdf/2608.20317) [HTML (experimental)](https://arxiv.org/html/2608.20317v1)

> Abstract:The BrowseComp-Plus benchmark disentangled the evaluation of agentic search by replacing opaque web search with a fixed corpus, so that an agent's role can be separated from the retriever's. That corpus, however, holds only about 100K documents and was assembled from the supporting documents of the benchmark's own queries plus mined hard negatives, so the evidence and the distractors were both selected per query. We introduce $\text{BrowseComp-Plus}_{\text{CM}}$, which keeps the BrowseComp-Plus questions but relocates their evidence to ClimbMix, a 400B-token, 553M-document mixture of web text released by NVIDIA for pre-training language models and built without reference to any benchmark. Our main contribution is the projection pipeline that makes this possible: it decomposes each question into atomic reasoning hops and grounds every hop in the new corpus, retaining a question only when automatic verification, an independent agent, and human review all confirm that every hop is supported. The pipeline is dataset-agnostic and applies to any benchmark whose questions decompose into verifiable facts. Applied to the 830 BrowseComp-Plus test questions, our pipeline yields 57 fully grounded questions with question-level relevance judgments. Projection shifts the difficulty onto retrieval, as the strongest agent we evaluate loses five points of answer accuracy but sees its evidence recall fall from 84.3% to 21.4% while issuing 63% more search calls. As the first of a series of projections, we release the pipeline, the benchmark, and our analyses at [this https URL](https://github.com/castorini/cmass).

| Subjects: | Information Retrieval (cs.IR) |
| --- | --- |
| Cite as: | [arXiv:2608.20317](https://arxiv.org/abs/2608.20317) \[cs.IR\] |
|  | (or [arXiv:2608.20317v1](https://arxiv.org/abs/2608.20317v1) \[cs.IR\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.20317](https://doi.org/10.48550/arXiv.2608.20317) |

## Submission history

From: Sahel Sharifymoghaddam \[[view email](https://arxiv.org/show-email/1c59a7b5/2608.20317)\]  
**\[v1\]** Thu, 20 Aug 2026 17:54:48 UTC (284 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.20317) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))