# Open-HumanScale开源5万小时第一视角数据与具身预训练配方

> 原文链接：https://robotics.hellobike.com/papers/2026-11-04-2611-06817-3da22286
> 频道：papers | 更新日期：2026-11-04 | 标签：auto, arxiv, hf-daily-papers
> 核心摘要：发布Open-HumanScale-50K数据集及全开源预训练流程，验证人类视频预训练对机器人策略的log-linear缩放律。HumanWAM模型在RoboTwin 2.0达92.7%成功率，并展现零样本视觉目标泛化与物理实机上下文学习能力。

# Open-HumanScale开源5万小时第一视角数据与具身预训练配方

::: v-pre
> 本文由 AI 根据公开内容整理，可能有误，请以原文为准：[Open-HumanScale: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models](https://arxiv.org/abs/2611.06817)（HF Daily Papers · arXiv 2611.06817）

## 核心内容

研究团队推出Open-HumanScale，这是首个面向具身基础模型的全开源人类第一视角视频预训练配方，核心资产为包含5万小时密集标注手部位姿与自然语言的Open-HumanScale-50K语料库。基于该 corpus 的闭环仿真评估揭示了纯人类视频预训练时长与下游机器人任务成功率之间存在经验性 log-linear 缩放定律，为利用海量人类视频数据指导机器人学习提供了量化依据与可复现基准。

基于此发现构建的 HumanWAM-Vanilla 世界-动作模型仅用人类视频预训练，后训练阶段在 RoboTwin 2.0 Clean2Clean/Clean2Random 分别达 92.08%/68.88% 成功率。进一步引入人机联合预训练的 HumanWAM-Goal 模型将 Clean2Random 性能提升至 84.60%，并在物理机器人上涌现出无需参数更新即可适应未见视觉目标的具身上下文学习能力。项目已开源数据集、Pipeline、模型权重及评测套件。

## 关键看点

- **缩放定律验证**：通过闭环仿真确立人类视频预训练时长与机器人成功率的log-linear关系，量化数据规模收益
- **SOTA性能突破**：HumanWAM-Goal在RoboTwin 2.0 Clean2Random任务达84.60%成功率，刷新该基准纪录
- **涌现能力验证**：物理实机测试证实模型具备零样本视觉目标泛化与动态适应未见目标的具身上下文学习能力
- **全栈开源资产**：开放5万小时EgoVideo语料、数据清洗Pipeline、预训练/后训练权重及仿真评测代码

## 工程与产业观察

该工作确立了人类第一视角视频作为具身智能通用预训练数据的有效性边界，其揭示的log-linear缩放律为算法工程师规划数据预算提供关键参考。全栈开源的HumanWAM模型与50K数据集降低了世界-动作模型研发门槛，验证了纯人类数据预训练结合少量机器人微调即可实现高泛化操作策略的技术路径。

## 英文摘要 (Abstract)

Egocentric human video offers a scalable source of manipulation priors for robot learning, yet translating raw data scale into generalizable robot policies demands rigorous data curation and training recipes. We introduce Open-HumanScale, the first fully open recipe for scaling egocentric pretraining to 50,000 hours, anchored by Open-HumanScale-50K---a curated egocentric video corpus densely annotated with hand motion and natural language. Leveraging this corpus, we systematically investigate pure human pretraining via closed-loop simulation evaluations and uncover an empirical log-linear scaling law linking pretraining data hours to downstream robot success rates. Grounded in these findings, we develop HumanWAM-Vanilla, a world--action model pretrained exclusively on 50k hours of egocentric human video. Upon robot post-training, HumanWAM-Vanilla reaches 92.08% success on RoboTwin 2.0 Clean2Clean and 68.88% on Clean2Random, alongside a 14.17% mean success rate and a 19.00 mean progress score on RoboDojo. To push the capability frontier further, we explore joint human--robot pretraining with HumanWAM-Goal, a visual-goal-conditioned world--action model. Hybrid data scaling unlocks broad zero-shot generalization across both simulation and physical benchmarks, propelling post-trained RoboTwin 2.0 performance to new state-of-the-art levels: 92.70% on Clean2Clean and a substantial leap to 84.60% on Clean2Random. On physical robots, HumanWAM-Goal further exhibits emergent embodied in-context learning, dynamically adapting its manipulation behavior to unseen visual goals without parameter updates. To support reproducible scaling research in embodied AI, we open-source the Open-HumanScale-50K corpus, data curation and pretraining pipelines, alongside the pretrained and post-trained HumanWAM checkpoints, inference code, and simulation evaluation suites.

Hugging Face Daily Papers 热度：2 upvotes
作者：Juncheng Ma, Jiaxin Li, Jiankai Tu, Bo Liang, Guodong Lv, Yufan Deng, Junzhe Tang, Liangzhu Yang, Chen Bian, Zicong Cao, Baohao Zhu, Jianxin Bi, Tengfei Yan, Yinyuan Zhang, Kewei Zhang, Xinglong Chen,…
:::

## 原始参考来源 (Sources)

- [HF Daily Papers · arXiv 2611.06817：Open-HumanScale: An Open Recipe for Scaling Human and Robot Experience for Embodied Foundation Models](https://arxiv.org/abs/2611.06817)