---
格式版本: 2
标题: "Learning constructive models of emergent systems"
原文链接: "https://arxiv.org/abs/2608.18248"
发布日期: "2026-08-18"
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, 18 Aug 2026 18:55:31 UTC (3,557 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: "2026-08-20T14:22:53+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: "2026-08-20T14:19:37+08:00"
入库时间: "2026-08-20T06:22:53.581Z"
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: 15
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "该论文为统计物理与机器学习研究，讨论层受限玻尔兹曼机，与超节点/AI Rack/机柜级AI基础设施完全无关，未涉及任何核心部件、供电、散热或量产部署。"
AI质检模型: "tx-deepseek-v4-flash"
AI质检时间: "2026-08-20T14:24:39+08:00"
AI主题相关性: 0
AI来源权威性: 5
AI新颖性: 5
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 5
AI摘要: "研究人员提出层限制玻尔兹曼机（LRBM），通过约束信息从低熵尺度流向高熵尺度，以数据驱动方式逐步构建涌现系统，并在图像、自然语言和酶序列上验证了层级逻辑。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:21:53.578Z"
采集批次: "2026年8月20日14点19分32秒"
采集批次ID: "20260820-141932-079"
去重键: "https://arxiv.org/abs/2608.18248"
---

## Condensed Matter > Disordered Systems and Neural Networks

## Title:Learning constructive models of emergent systems

Authors:[Bipul Pandey](https://arxiv.org/search/cond-mat?searchtype=author&query=Pandey,+B), [Caden Proctor](https://arxiv.org/search/cond-mat?searchtype=author&query=Proctor,+C), [Vinay Ramanathan](https://arxiv.org/search/cond-mat?searchtype=author&query=Ramanathan,+V), [Nico Roth](https://arxiv.org/search/cond-mat?searchtype=author&query=Roth,+N), [Arjun S. Raman](https://arxiv.org/search/cond-mat?searchtype=author&query=Raman,+A+S)

[View PDF](https://arxiv.org/pdf/2608.18248)

> Abstract:Emergent systems - systems containing multi-scale interactions - often arise through iteration rather than explicit forward design. As such, design principles for building emergent systems have been under-explored. Current artificial intelligence architectures, while useful for generation, do not provide a logic for how to construct emergent systems. Inspired by statistical phylogenetics, we show that inferring scales of system entropy then explicitly constraining information flow from low to high entropic scales yields data-driven, stepwise 'constructive models' of emergent systems. We term the resulting architecture the Layer-Restricted Boltzmann Machine (LRBM). Applied identically to digit images, natural language, and enzyme sequences, LRBM models follow hierarchical logic: lower layers encode global structure from which higher layers encode fine-grained features. As a stringent experimental challenge for LRBM-based construction, we built and evaluated 160 constructive trajectories of synthetic and 1,130 natural chorismate mutase (CM) enzymes in an in vivo functional assay. Designed CMs functioned up to 53% sequence divergent from the nearest natural homolog. Notably, enzyme fold emerged at intermediate LRBM layers while function arose only upon satisfying all layers. Fold was therefore necessary but insufficient for function in this enzyme family. Our results suggest a shared hierarchical, learnable statistical logic for constructing emergent systems.

| Comments: |  |
| --- | --- |
| Subjects: | Disordered Systems and Neural Networks (cond-mat.dis-nn); Populations and Evolution (q-bio.PE) |
| Cite as: | [arXiv:2608.18248](https://arxiv.org/abs/2608.18248) \[cond-mat.dis-nn\] |
|  | (or [arXiv:2608.18248v1](https://arxiv.org/abs/2608.18248v1) \[cond-mat.dis-nn\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.18248](https://doi.org/10.48550/arXiv.2608.18248) |

## Submission history

From: Bipul Pandey \[[view email](https://arxiv.org/show-email/9626abec/2608.18248)\]  
**\[v1\]** Tue, 18 Aug 2026 18:55:31 UTC (3,557 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.18248) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))