---
格式版本: 2
标题: "L-COIN: LLM-Assisted Counterfactual Inference for Game-Theoretic Distributed Computation Offloading in Sub-THz LEO Satellite Networks"
原文链接: "https://arxiv.org/abs/2608.16174"
发布日期: "2026-08-17"
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, 17 Aug 2026 06:45:19 UTC (461 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: "2026-08-20T15:43:54+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: "2026-08-20T15:39:49+08:00"
入库时间: "2026-08-20T07:43:54.444Z"
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=bandwidth&searchtype=all"
匹配关键词:
  - "bandwidth"
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: 0
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文主题为LEO卫星网络的子THz通信与计算卸载，与超节点、AI Rack、机柜级AI基础设施等完全无关，仅命中bandwidth关键词但属卫星通信带宽，明显不相关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-20T15:49:27+08:00"
AI主题相关性: 0
AI来源权威性: 0
AI新颖性: 0
AI技术细节: 0
AI商业部署信号: 0
AI完整性: 0
AI摘要: "本文提出 L-COIN 框架，利用大语言模型辅助的反事实推断，实现亚太赫兹低轨卫星网络中的博弈论分布式计算卸载，避免全局信息依赖并适应时变拓扑。实验显示，相比现有方案，该框架可将卸载成本降低 10.9% 至 27.7%。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:25:41.134Z"
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/2608.16174"
---

## Electrical Engineering and Systems Science > Systems and Control

## Title:L-COIN: LLM-Assisted Counterfactual Inference for Game-Theoretic Distributed Computation Offloading in Sub-THz LEO Satellite Networks

Authors:[Jinhao Yi](https://arxiv.org/search/eess?searchtype=author&query=Yi,+J), [Weijun Gao](https://arxiv.org/search/eess?searchtype=author&query=Gao,+W), [Chong Han](https://arxiv.org/search/eess?searchtype=author&query=Han,+C)

[View PDF](https://arxiv.org/pdf/2608.16174) [HTML (experimental)](https://arxiv.org/html/2608.16174v1)

> Abstract:As Space-Based Information Networks (SBINs) evolve toward high-capacity, intelligence-centric paradigms, integrating sub-Terahertz (sub-THz) communication into Low Earth Orbit (LEO) satellite constellations has emerged as a critical enabler for ultra-broadband and resilient global connectivity. By exploiting the ultra-wide bandwidth of sub-THz links to reduce transmission delays, resource-constrained ground devices can seamlessly offload compute-intensive tasks to LEO edge servers. However, satellite motion, short visibility windows, and limited onboard resources make offloading decisions highly time-varying. Existing distributed offloading schemes typically require repeated inter-device state exchange and poorly adapt to time-varying LEO topology or traffic conditions. To address these limitations, a decentralized game-theoretic offloading framework empowered by large language models (LLMs) and counterfactual inference is proposed in this paper. First, a realistic offloading system is established by integrating time-varying 3D-Walker topology. Second, a game-theoretic scheme using counterfactual inference is introduced to deduce unobserved states from local histories, eliminating global information reliance. Finally, an LLM-empowered semantic fusion algorithm is integrated into the counterfactual inference to enhance adaptability through zero-shot reasoning and self-reflection. Numerical results show that L-COIN reduces offloading cost by 10.9% to 27.7% relative to state-of-the-art baselines.

| Subjects: | Systems and Control (eess.SY) |
| --- | --- |
| Cite as: | [arXiv:2608.16174](https://arxiv.org/abs/2608.16174) \[eess.SY\] |
|  | (or [arXiv:2608.16174v1](https://arxiv.org/abs/2608.16174v1) \[eess.SY\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.16174](https://doi.org/10.48550/arXiv.2608.16174) |

## Submission history

From: Jinhao Yi \[[view email](https://arxiv.org/show-email/51a7108f/2608.16174)\]  
**\[v1\]** Mon, 17 Aug 2026 06:45:19 UTC (461 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.16174) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))