---
格式版本: 2
标题: "Auditable AI-Assisted Research Writing: An Engineering Discipline with Pre-Registered Process Observation"
原文链接: "https://arxiv.org/abs/2608.10858"
发布日期: "2026-08-11"
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\\]** Tue, 11 Aug 2026 12:30:42 UTC (142 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: "2026-08-12T11:29:18+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: "2026-08-12T11:26:18+08:00"
入库时间: "2026-08-12T03:29:18.496Z"
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=AI&searchtype=all"
匹配关键词:
  - "AI"
相关厂家:
  []
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: 5
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "资料为arXiv上的AI辅助研究写作可审计性论文，与超节点/AI Rack/机柜级AI基础设施完全无关，命中‘AI’关键词但正文无关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-12T11:29:29+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
采集批次: "2026年8月12日2点04分20秒"
采集批次ID: "20260812-020420-289"
去重键: "https://arxiv.org/abs/2608.10858"
---

## Computer Science > Digital Libraries

## Title:Auditable AI-Assisted Research Writing: An Engineering Discipline with Pre-Registered Process Observation

Authors:[Yang Zhou](https://arxiv.org/search/cs?searchtype=author&query=Zhou,+Y), [Chengqun Yu](https://arxiv.org/search/cs?searchtype=author&query=Yu,+C)

[View PDF](https://arxiv.org/pdf/2608.10858) [HTML (experimental)](https://arxiv.org/html/2608.10858v1)

> Abstract:Language models now draft, classify and criticise inside research production, yet the artifacts they help produce carry little accountable history. Rather than detecting machine involvement afterwards, we specify an auditability discipline built at production time: git sealing with an anchor lineage, hash-bound provenance, red-line gates that refuse non-compliant artifacts and log every refusal, cross-model role separation, and programmatic assembly from registered sources. Adherence is instrumented by metric cards, each carrying a pre-registered blind spot and evidential standing, frozen before the prospective case it observes. In that case the observed project's pre-registered confirmatory test was executed under seal and returned No-Go, and that project's frozen stopping rule halted the work, against its own operators. A lower-graded retrospective case covers families whose machinery predates the protocol. Current observations are provisional; we release a package from which a third party can recompute every primary metric.

| Comments: |  |
| --- | --- |
| Subjects: | Digital Libraries (cs.DL); Computers and Society (cs.CY); Human-Computer Interaction (cs.HC) |
| Cite as: | [arXiv:2608.10858](https://arxiv.org/abs/2608.10858) \[cs.DL\] |
|  | (or [arXiv:2608.10858v1](https://arxiv.org/abs/2608.10858v1) \[cs.DL\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.10858](https://doi.org/10.48550/arXiv.2608.10858) |

## Submission history

From: Yang Zhou \[[view email](https://arxiv.org/show-email/50dc1936/2608.10858)\]  
**\[v1\]** Tue, 11 Aug 2026 12:30:42 UTC (142 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.10858) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))