---
格式版本: 2
标题: "Semantic-Aware Sub-Band Allocation for Terahertz Communications"
原文链接: "https://arxiv.org/abs/2608.30984"
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 15:42:48 UTC (543 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: "2026-09-02T03:56:19+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: "2026-09-02T03:54:32+08:00"
入库时间: "2026-09-01T19:56:19.885Z"
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=Oracle&searchtype=all"
匹配关键词:
  - "performance"
相关厂家:
  - "Oracle"
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: 10
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "内容为太赫兹通信中语义感知子带分配的信号处理论文，与超节点、AI Rack、机柜级AI基础设施、供电散热互连等主题完全无关，属无关学术文献。"
AI质检模型: "zj-deepseek-v4-flash"
AI质检时间: "2026-09-02T03:56:28+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: 5
AI摘要: "该论文提出一种面向太赫兹通信的语义感知子带分配方法，利用SBERT代理模型预测语义保真度，并通过模仿学习训练神经调度器。与DeepSC完整评估相比，运行时间降低200倍以上，在真实THz信道条件下逼近oracle性能并优于基准方案。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-01T23:06:13.242Z"
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/2608.30984"
---

## Electrical Engineering and Systems Science > Signal Processing

## Title:Semantic-Aware Sub-Band Allocation for Terahertz Communications

Authors:[Fatima Ismail](https://arxiv.org/search/eess?searchtype=author&query=Ismail,+F), [Hadi Sarieddeen](https://arxiv.org/search/eess?searchtype=author&query=Sarieddeen,+H), [Jihad Fahs](https://arxiv.org/search/eess?searchtype=author&query=Fahs,+J)

[View PDF](https://arxiv.org/pdf/2608.30984) [HTML (experimental)](https://arxiv.org/html/2608.30984v1)

> Abstract:This paper studies semantic-aware sub-band al- location for terahertz (THz) communication systems, where frequency-selective molecular absorption creates highly non- uniform sub-band qualities. Unlike conventional formulations, semantic fidelity depends nonlinearly on the signal-to-noise ratio (SNR) and is also sentence-specific, leading to a non-separable assignment problem that is generally not solvable using simple ordering-based policies. To address this, we use a sentence-BERT (SBERT)-based surrogate model that predicts semantic fidelity from the sentence embedding and sub-band SNR. We propose an importance-aware scheduler that assigns sentences to sub- bands based on their semantic contribution using an oracle utility function that captures importance-weighted semantic similarity across sentence-sub-band pairs. A neural scheduler is then trained through imitation learning to approximate the oracle policy at more than 200x lower runtime than full DeepSC- based oracle evaluation. Integrated with a deep-learning-enabled semantic communication (DeepSC) system, the proposed method consistently outperforms all benchmark schemes and approaches oracle-level performance under realistic THz channel conditions.

| Comments: |  |
| --- | --- |
| Subjects: | Signal Processing (eess.SP) |
| Cite as: | [arXiv:2608.30984](https://arxiv.org/abs/2608.30984) \[eess.SP\] |
|  | (or [arXiv:2608.30984v1](https://arxiv.org/abs/2608.30984v1) \[eess.SP\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.30984](https://doi.org/10.48550/arXiv.2608.30984) |

## Submission history

From: Fatima Ismail \[[view email](https://arxiv.org/show-email/65a406fe/2608.30984)\]  
**\[v1\]** Mon, 31 Aug 2026 15:42:48 UTC (543 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.30984) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))