---
格式版本: 2
标题: "A Quantum Roadmap for Softmax Attention: Exact Born-Rule Analogs for Softmax Attention on the Probability Simplex"
原文链接: "https://arxiv.org/abs/2608.11173"
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 17:32:30 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: "2026-08-12T11:13:07+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: "2026-08-12T11:12:25+08:00"
入库时间: "2026-08-12T03:13:07.328Z"
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=roadmap&searchtype=all"
匹配关键词:
  - "roadmap"
  - "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: "论文为量子注意力机制理论，与超节点/AI Rack/机柜级AI基础设施等硬件架构、供电、散热、互连、量产落地完全无关，主题相关性0分。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-12T11:13:33+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/2608.11173"
---

## Quantum Physics

## Title:A Quantum Roadmap for Softmax Attention: Exact Born-Rule Analogs for Softmax Attention on the Probability Simplex

Authors:[Eric A. F. Reinhardt](https://arxiv.org/search/quant-ph?searchtype=author&query=Reinhardt,+E+A+F), [Adam J. Hauser](https://arxiv.org/search/quant-ph?searchtype=author&query=Hauser,+A+J)

[View PDF](https://arxiv.org/pdf/2608.11173) [HTML (experimental)](https://arxiv.org/html/2608.11173v1)

> Abstract:The attention mechanism forms the foundation of many modern AI models such as the Transformer. In one subclass of problems where attention is used, inputs and outputs are bound to the probability simplex so that all outputs sum to one. In this setting, softmax attention admits an exact, component-by-component quantum realization. Attention scores are Hadamard-test statistics on block-encoded projections of amplitude-encoded inputs. The exponential softmax is the interior of a cosine-squared family generated by Born-rule measurement under an exact bijection, whose boundary expresses sparse attention with exact zeros at finite parameter values. The softmax temperature is a repetition count where post-selected measurement rounds realize discretized inverse temperature exactly. Value aggregation is a deterministic column-loading channel that dilates the column-stochastic value matrix. The gated residual is the preparation angle of a single ancilla, with the additive identity at a mixing angle of {\\pi}/2. Every learnable parameter is a rotation-gate angle. The composed layer is exact in the infinite-shot limit with one measure-and-reload step per attention score; a fully-coherent variant is {\\epsilon}-approximate via quantum singular value transformation in the infinite depth limit. The algebraic core is machine-checked in Lean 4.

| Comments: |  |
| --- | --- |
| Subjects: | Quantum Physics (quant-ph); Machine Learning (cs.LG) |
| Cite as: | [arXiv:2608.11173](https://arxiv.org/abs/2608.11173) \[quant-ph\] |
|  | (or [arXiv:2608.11173v1](https://arxiv.org/abs/2608.11173v1) \[quant-ph\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.11173](https://doi.org/10.48550/arXiv.2608.11173) |

## Submission history

From: Eric Reinhardt \[[view email](https://arxiv.org/show-email/54b8250f/2608.11173)\]  
**\[v1\]** Tue, 11 Aug 2026 17:32:30 UTC (40 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.11173) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))