---
格式版本: 2
标题: "Conditional Total Correlation and the Serial Depth of Adaptive Parallel Sampling"
原文链接: "https://arxiv.org/abs/2608.25505"
发布日期: "2026-08-26"
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\\]** Wed, 26 Aug 2026 08:16:24 UTC (78 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: "2026-08-27T23:24:56+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: "2026-08-27T23:21:22+08:00"
入库时间: "2026-08-27T15:24:56.884Z"
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:
  []
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: 33
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "正文主线是掩码扩散模型并行解码的自适应采样理论，新增条件总相关恒等式及多类分布的串行深度结论；来源为可追溯的arXiv预印本摘要页。固定知识库未见相同研究，但该成果不涉及超节点、AI机架、互连、供电、液冷或机架级部署，也无客户、量产和生产平台指标。命中应用与模型效率及缺乏生产级基础设施机制的否决项，不具备业务准入价值。"
AI质检模型: "gpt-5.6-sol"
AI质检时间: "2026-08-27T23:25:05+08:00"
AI主题相关性: 1
AI来源权威性: 10
AI新颖性: 12
AI技术细节: 3
AI商业部署信号: 0
AI完整性: 7
AI评分提示词版本: "v17-精简生产版"
AI评分提示词SHA256: "48fb9777f386026761b4873eaff30807694fb11e9b352d7c69bf2dfde750cc7d"
AI评分知识库版本: "knowledge_base_v1-20260819+runtime.12"
AI评分知识库SHA256: "13aade27529fda213430b6def2f11efe60b599f8676ae127cad35798cf7169c8"
AI评分知识库检索词: "[\"Scale-up\",\"PDF\",\"arxiv.org/pdf/2608.25505\",\"HTML\",\"arxiv.org/html/2608.25505v1\",\"v1\",\"arxiv.org/abs/2608.25505\",\"arxiv.org/abs/2608.25505v1\",\"doi.org/10.48550/arXiv.2608.25505\",\"arxiv.org/show-email/9a1cccc2/2608.25505\",\"UTC\",\"KB\"]"
AI评分知识库命中: "[{\"id\":\"july-correct-0111\",\"title\":\"StrataCL: Fabric-Native Communication Library for Production Supernodes\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"PDF\",\"HTML\",\"v1\",\"UTC\",\"KB\"],\"rank\":-22.625644289950294},{\"id\":\"july-correct-0010\",\"title\":\"Schneider Electric and AMD release first Helios platform reference design to accelerate AI Factory deployment\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"PDF\",\"v1\",\"KB\"],\"rank\":-9.545382929171423},{\"id\":\"historical-jun-010\",\"title\":\"爱建证券-电子行业专题报告：Vera Rubin量产提速，RTX Spark打开终端AI新空间-260608.pdf\",\"sourceType\":\"curated_item\",\"time\":\"2026-06\",\"matchedTerms\":[\"PDF\"],\"rank\":-6.876785924495043},{\"id\":\"july-correct-0115\",\"title\":\"锚定 300kW 整机柜演进方向 OAII 社区三项规范联合发布，树立AIDC基础设施建设标准答案\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"HTML\",\"v1\"],\"rank\":-6.545627953632225},{\"id\":\"historical-may-043\",\"title\":\"2026安全可靠测评结果(第2号)出炉，壁仞科技壁砺166获I级认证\",\"sourceType\":\"curated_item\",\"time\":\"2026-05\",\"matchedTerms\":[\"v1\"],\"rank\":-6.534802061811721}]"
AI摘要: "条件总相关被证明是自适应并行采样中每一轮并行性的精确信息代价：任意策略的前向KL散度等于其揭示轮次中累积的条件总相关。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-08-27T18:59:27.721Z"
采集批次: "2026年8月27日21点49分30秒"
采集批次ID: "20260827-214930-609"
去重键: "https://arxiv.org/abs/2608.25505"
---

## Computer Science > Information Theory

## Title:Conditional Total Correlation and the Serial Depth of Adaptive Parallel Sampling

Authors:[Chuling Wen](https://arxiv.org/search/cs?searchtype=author&query=Wen,+C), [Weijie Liang](https://arxiv.org/search/cs?searchtype=author&query=Liang,+W), [Jian Lu](https://arxiv.org/search/cs?searchtype=author&query=Lu,+J)

[View PDF](https://arxiv.org/pdf/2608.25505) [HTML (experimental)](https://arxiv.org/html/2608.25505v1)

> Abstract:Motivated by parallel decoding in masked diffusion models, we study adaptive parallel sampling of discrete vectors: in each round, a deterministic policy selects unrevealed coordinates on the basis of the values observed so far, and the selected coordinates are sampled independently from their exact conditional marginals. Approximation error is measured by forward Kullback-Leibler divergence, and serial depth is the minimum target-averaged number of rounds meeting a prescribed error budget. Our central result is an exact identity: the divergence of every policy equals the expected conditional total correlation accumulated over its reveal rounds, so conditional total correlation is the exact information cost of within-round parallelism. The identity yields zero-error schedules for finite-order Markov chains with round complexity proportional to the Markov order and logarithmic in sequence length, a matching logarithmic characterization of the Bernoulli walk at every fixed error budget, and a linear-versus-logarithmic separation between left-to-right and hierarchical reveal orders. Uniform random permutations require linearly many expected rounds at every fixed budget; their hard-cap round-error tradeoff is an exact integer-composition problem whose fixed-round asymptotics and joint-scaling frontier we determine. Uniform balanced binary strings have depth of order squared logarithm, and binary one-hot blocks have square-root depth, with rectangular versions realizing every polynomial exponent up to one half. These results separate serial depth from entropy and negative log-likelihood, and establish conditional-dependence structure as a fundamental determinant of parallelizability. Experiments with a masked diffusion language model show that the pseudo-cost distinguishes deployed decoding rules and that its policy rankings agree closely with the quality of self-sampled outputs.

| Subjects: | Information Theory (cs.IT); Computation and Language (cs.CL) |
| --- | --- |
| Cite as: | [arXiv:2608.25505](https://arxiv.org/abs/2608.25505) \[cs.IT\] |
|  | (or [arXiv:2608.25505v1](https://arxiv.org/abs/2608.25505v1) \[cs.IT\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.25505](https://doi.org/10.48550/arXiv.2608.25505) |

## Submission history

From: Jian Lu \[[view email](https://arxiv.org/show-email/9a1cccc2/2608.25505)\]  
**\[v1\]** Wed, 26 Aug 2026 08:16:24 UTC (78 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.25505) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))