---
格式版本: 2
标题: "Baseline-Relative Counterfactual Refinement for Bit-Aware Visual Token Communication"
原文链接: "https://arxiv.org/abs/2608.16192"
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 07:16:37 UTC (259 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: "2026-08-20T14:59:47+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: "2026-08-20T14:56:53+08:00"
入库时间: "2026-08-20T06:59:47.844Z"
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=C-Link&searchtype=all"
匹配关键词:
  - "C-Link"
  - "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: 10
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "论文主题为视觉令牌通信，与超节点/AI Rack/机柜级AI基础设施完全无关，仅搜索词命中C-Link但正文未涉及。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-20T15:00:26+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: "针对生成式视觉token通信中token选择难以直接优化最终重建质量的问题，作者提出门控反事实修正层GCR-C：构建候选集并以完整预算Local-MDL续算，仅在相对基线获得正增益时替换动作。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:24:56.962Z"
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/2608.16192"
---

## Computer Science > Artificial Intelligence

## Title:Baseline-Relative Counterfactual Refinement for Bit-Aware Visual Token Communication

Authors:[Jia Guo](https://arxiv.org/search/cs?searchtype=author&query=Guo,+J), [Xiaohan Zhao](https://arxiv.org/search/cs?searchtype=author&query=Zhao,+X), [Changwang Liu](https://arxiv.org/search/cs?searchtype=author&query=Liu,+C), [Shuqing He](https://arxiv.org/search/cs?searchtype=author&query=He,+S), [Chenyang Zhang](https://arxiv.org/search/cs?searchtype=author&query=Zhang,+C), [Bingchuan Zhao](https://arxiv.org/search/cs?searchtype=author&query=Zhao,+B), [Jinqi Zhu](https://arxiv.org/search/cs?searchtype=author&query=Zhu,+J)

[View PDF](https://arxiv.org/pdf/2608.16192) [HTML (experimental)](https://arxiv.org/html/2608.16192v1)

> Abstract:Generative visual-token communication reduces transmission load by sending only selected discrete tokens and reconstructing missing content at the receiver. However, existing token-selection criteria based on local uncertainty, importance, or diversity do not directly determine whether changing the current selection improves the final reconstruction under the same packet budget. To address this problem, we propose Gated Counterfactual Refinement for Communication (GCR-C), a rollout-style correction layer over Local-MDL. GCR-C constructs a compact diversified candidate set, evaluates each candidate through matched full-budget Local-MDL continuation, and replaces the baseline action only when a positive baseline-relative reconstruction gain is obtained. Experiments on CIFAR-10, STL-10, a coded 5G-LDPC link, and a limited high-resolution Kodak transfer show that GCR-C consistently improves reconstruction quality at active low- and medium-rate operating points without increasing the realized packet rate, while remaining effective across changes in dataset, channel condition, resolution, token grid, and tokenizer. The results also reveal a clear quality--computation tradeoff due to the additional encoder-side counterfactual evaluation.

| Subjects: | Artificial Intelligence (cs.AI) |
| --- | --- |
| Cite as: | [arXiv:2608.16192](https://arxiv.org/abs/2608.16192) \[cs.AI\] |
|  | (or [arXiv:2608.16192v1](https://arxiv.org/abs/2608.16192v1) \[cs.AI\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.16192](https://doi.org/10.48550/arXiv.2608.16192) |

## Submission history

From: Jia Guo \[[view email](https://arxiv.org/show-email/359f53ce/2608.16192)\]  
**\[v1\]** Mon, 17 Aug 2026 07:16:37 UTC (259 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.16192) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))