---
格式版本: 2
标题: "The Embodiment Gap in Robot Foundation Models"
原文链接: "https://arxiv.org/abs/2608.18433"
发布日期: "2026-08-19"
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, 19 Aug 2026 01:55:04 UTC (321 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: "2026-08-20T15:26:46+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: "2026-08-20T15:18:01+08:00"
入库时间: "2026-08-20T07:26:46.234Z"
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=deployment&searchtype=all"
匹配关键词:
  - "deployment"
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: 7
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "该资料为arXiv机器人基础模型论文，讨论机器人具身泛化问题，与超节点、AI Rack、机柜级AI基础设施及供电散热互连等主题完全无关。"
AI质检模型: "ali-deepseek-v4-flash"
AI质检时间: "2026-08-20T15:30:23+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: 2
AI摘要: "该文提出“具身差距”概念，指机器人基础模型（含VLA策略）可泛化但尚未适配到特定机器人本体上执行的差距；通过二维图谱梳理可跨本体复用的结构，并提出用报告框架评估跨本体学习中的剩余适配工作，而不只看成功率。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-09-07T03:19:02.883Z"
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/2608.18433"
---

## Computer Science > Robotics

## Title:The Embodiment Gap in Robot Foundation Models

[View PDF](https://arxiv.org/pdf/2608.18433) [HTML (experimental)](https://arxiv.org/html/2608.18433v1)

> Abstract:Robot foundation models (RFMs), including vision-language-action (VLA) policies, are often discussed through a scaling view: more data, larger models, and broader benchmarks should improve generalization. In robotics, however, a model can generalize while work still remains before it can run on a robot with a particular body. The work required differs across methods and target robots, and those differences affect practical deployment. We call the gap between reusable models, representations, or data and their use in execution on the target robot the embodiment gap. This survey examines what can be reused across robot embodiments and what must still be implemented on a new robot. We place existing methods on a two-axis map that shows the type of shared structure and the stage at which adaptation is needed for execution on the target robot. We then examine recent work through three overlapping research directions: sharing semantics and perception, sharing robot data and interfaces, and learning correspondence across embodiments. We also propose a reporting framework for adaptation work that success rate alone does not reveal. The framework identifies the work that should be checked when comparing cross-embodiment learning and highlights work that remains on a new robot and questions for future study.

| Comments: |  |
| --- | --- |
| Subjects: | Robotics (cs.RO); Machine Learning (cs.LG) |
| Cite as: | [arXiv:2608.18433](https://arxiv.org/abs/2608.18433) \[cs.RO\] |
|  | (or [arXiv:2608.18433v1](https://arxiv.org/abs/2608.18433v1) \[cs.RO\] for this version) |
|  | [https://doi.org/10.48550/arXiv.2608.18433](https://doi.org/10.48550/arXiv.2608.18433) |
| Journal reference: | Transactions on Machine Learning Research, August 2026 |

## Submission history

From: Yukiyasu Domae PhD. \[[view email](https://arxiv.org/show-email/ee8a8670/2608.18433)\]  
**\[v1\]** Wed, 19 Aug 2026 01:55:04 UTC (321 KB)

[Which authors of this paper are endorsers?](https://arxiv.org/auth/show-endorsers/2608.18433) | Disable MathJax ([What is MathJax?](https://info.arxiv.org/help/mathjax.html))