---
格式版本: 2
标题: "Cubic-Root Gaussian Approximation under Unrestricted Covariance"
原文链接: "https://arxiv.org/abs/2608.30221"
发布日期: "2026-08-31"
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\\]** Mon, 31 Aug 2026 04:10:35 UTC (40 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: "2026-09-02T03:50:39+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: "2026-09-02T03:49:17+08:00"
入库时间: "2026-09-01T19:50:39.899Z"
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=Scale-up&searchtype=all"
匹配关键词:
  - "Scale-up"
相关厂家:
  - "OpenAI"
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: 15
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "该内容为arXiv数学统计理论论文，讨论高斯近似，与超节点/AI Rack/机柜级AI基础设施完全无关，仅因搜索词命中Scale-up。"
AI质检模型: "zj-deepseek-v4-flash"
AI质检时间: "2026-09-02T03:51:17+08:00"
AI主题相关性: 0
AI来源权威性: 10
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 5
AI摘要: "Gao与Zhang在无限制协方差下证明了高维矩形高斯近似的误差上界；在坐标次指数条件和边际方差下界下，当维数为多项式且尺度有界时，误差可达n^{-1/3}，从而否证了Chernozhukov等人关于n^{-1/4}近最优的猜想。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-01T23:06:24.670Z"
采集批次: "2026年9月1日23点56分52秒"
采集批次ID: "20260901-235652-761"
去重键: "https://arxiv.org/abs/2608.30221"
---

## Mathematics > Statistics Theory

## Title:Cubic-Root Gaussian Approximation under Unrestricted Covariance

Authors:[Zijun Gao](https://arxiv.org/search/math?searchtype=author&query=Gao,+Z), [Weihan Zhang](https://arxiv.org/search/math?searchtype=author&query=Zhang,+W)

[View PDF](https://arxiv.org/pdf/2608.30221) [HTML (experimental)](https://arxiv.org/html/2608.30221v1)

> Abstract:For Gaussian approximation over high-dimensional rectangles under unrestricted covariance, Chernozhukov et al. (2023b) conjectured that the $n^{-1/4}$ rate, up to logarithmic factors, is near-optimal. We show that, under the coordinatewise subexponential condition with scale $B_n$ and the marginal variance lower bound condition with constant $b$ in Chernozhukov et al. (2023b), the approximation error in dimension $d$ is bounded by \\begin{align\*} C\_b\\min\\left\\{ 1,\\, \\left(\\frac{B\_n^2}{n}\\right)^{1/3}\\{\\log(2dn)\\}^{7/3} + \\frac{B\_n}{\\sqrt n}\\{\\log(2dn)\\}^{5/2} \\right\\}. \\end{align\*} In particular, for bounded $B_n$ and polynomial dimension, the new bound is $n^{-1/3}$ and therefore falsifies the polynomial-dimensional $n^{-1/4}$ near-optimality conjecture. The proof uses a two-stage interpolation and a rank-free matrix-weighted Gaussian surface bound, which may be of independent interest.  
> The initial proof attempt was generated by ChatGPT 5.6 Pro (OpenAI) and subsequently corrected and rewritten by the authors. The machine-checked Lean formalization of the proof can be found at the GitHub repository ([this https URL](https://github.com/WeihanZhang2001/cubic-root-gaussian-approximation-under-unrestricted-covariance)).

| Comments: |  |
| --- | --- |
| Subjects: | Statistics Theory (math.ST) |
| Cite as: | [arXiv:2608.30221](https://arxiv.org/abs/2608.30221) \[math.ST\] |
|  | (or [arXiv:2608.30221v1](https://arxiv.org/abs/2608.30221v1) \[math.ST\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.30221](https://doi.org/10.48550/arXiv.2608.30221) |

## Submission history

From: Zijun Gao \[[view email](https://arxiv.org/show-email/1cb939c9/2608.30221)\]  
**\[v1\]** Mon, 31 Aug 2026 04:10:35 UTC (40 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.30221) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))