数据工程与质量 · Data Engineering and Quality

Ego-centric 数据在具身模型训练中的问题与困难

Egocentric Data for Embodied Model Training

更新于

30 秒结论

先读这个判断

本文回顾 2025 年 7 月 15 日至 2026 年 7 月 15 日公开的具身智能论文,聚焦人类第一视角视频、头戴式相机数据及其派生的手—物轨迹,如何进入 VLA、模仿学习、世界模型和灵巧操作训练。检索注册表共收录 1,409 个候选,恢复 162 篇全文,最终以 15 篇通过全文阅读和主张审计的论文作为证据。

精读论文
15
正式证据事件
26
知识单元
EA-DATA, EA-XEMBODIMENT, EA-MODEL, EA-SENSOR
审计状态
当前结算 run 已通过仓库校验

v0 · 完整正文

研究综述

摘要

本文回顾 2025 年 7 月 15 日至 2026 年 7 月 15 日公开的具身智能论文,聚焦人类第一视角视频、头戴式相机数据及其派生的手—物轨迹,如何进入 VLA、模仿学习、世界模型和灵巧操作训练。检索注册表共收录 1,409 个候选,恢复 162 篇全文,最终以 15 篇通过全文阅读和主张审计的论文作为证据。

核心结论是:Ego-centric 数据的主要瓶颈并非单一的视觉域差,而是从“人类行为记录”到“机器人可执行监督”的转换损失。 原始视频擅长提供任务语义、物体交互顺序、视点选择和长尾场景,却通常缺少机器人所需的夹爪状态、关节/末端动作、力、接触、奖励与失败原因。相机自运动、遮挡、尺度歧义和本体差异又会进一步污染动作标签。近一年表现较好的工作,并不是把更多视频直接送进模型,而是在中间加入三维手—物表示、坐标解耦、接触与物理约束、质量过滤、模态监督掩码,以及少量目标机器人数据。规模有效,但只在这些接口成立之后有效。

研究边界

本文所称 Ego-centric 数据,包括自然人类第一视角 RGB/双目视频、头戴设备记录,以及由这些观测派生的头部、手腕、手指和手—物轨迹;研究对象是它们在 VLA、模仿学习、世界模型与灵巧操作训练中的作用。时间范围为 2025-07-15 至 2026-07-15。证据来自 15 篇通过完整非 OCR 全文阅读和主张核验的论文。本文能归纳当前方法反复遇到的机制性困难,但不能据此估计整个领域的论文占比,也不能把有限桌面操作、本体和传感配置下的结果外推为通用规模律。

中心判断

Ego-centric 数据的价值取决于一个可证伪条件:只有当训练管线能够显式表示不可观测变量,并把视觉行为转换成坐标一致、本体匹配、接触可行且经目标机器人闭环验证的监督时,增加人类视频规模才会稳定提高具身模型性能;若缺少这些接口,扩规模主要会扩大伪标签、物理不可执行轨迹和域错配。目标机器人数据因而不是人类视频规模化之后可以删除的残余成本,而是定义动作语义与检验转换是否成立的必要锚点。

1. 为什么 Ego-centric 数据仍然值得投入

第一视角数据与机器人腕部或头部相机在观察位置上较接近,能够记录“人在操作时看哪里、何时移动视点、如何接近物体”,并覆盖机器人示教难以经济采集的大量环境和任务。规模实验也给出了积极证据:EgoScale 在同一训练框架下把人类动作预训练从 1K 小时扩展到 20K 小时,五项真实机器人任务的平均完成度从 0.30 提升到 0.71。这个结果说明,第一视角数据确实可以形成可扩展的行为先验,而不仅是视觉表征预训练材料。

但这一结论有严格边界。EgoScale 的最佳结果还依赖少量精确对齐的人—机器人中间训练;EgoVLA 在没有机器人微调时,12 个仿真人形机器人任务的成功率全部为 0%。因此,更准确的判断是:Ego-centric 数据可以扩大“先验”的规模,却不能自动完成“控制”的落地。

2. 六类关键问题

2.1 视频里没有完整的机器人动作

普通第一视角视频只记录可见运动,不记录产生运动的控制量。Developing VLA from Egocentric Videos 从视频恢复物体轨迹,但无法获得夹爪开合状态,最终只能把动作表示为 9 维物体位姿增量。腕部轨迹也会遗漏手指姿态、接触建立/释放时刻以及施力方向。EgoScale 的消融进一步表明,wrist-only 动作会丢失灵巧操作所需的信息,而很小的指尖估计误差又可能在逆运动学后放大成异常关节姿态和接触丢失。

力敏感任务的缺口更明显。UMI-FT 之所以能处理顺应操作,是因为采集端额外记录六轴力/力矩、深度、位姿和抓握力。这反向说明:纯 RGB 或纯轨迹数据并不足以监督插入、穿刺、擦拭、旋拧等接触丰富任务。对世界模型而言,也需要明确“哪些模态有标签”。τ0-WM 将机器人遥操作、UMI、人类第一视角视频和策略 rollout 混合训练时,使用按模态的监督掩码,而不是假定所有样本都具有同一种动作真值。

2.2 相机运动、遮挡和尺度歧义会被误当作动作

第一视角的优势——相机随人移动——也是动作恢复的主要噪声源。手腕位姿与相机自运动天然耦合,如果坐标系没有统一,转头或走动会被错误写进手部动作。ActiveMimic 通过显式分离头部轨迹和腕部轨迹,把相机运动转化为“视点动作”;但在 HOT3D 的严格阈值下,自动恢复率仍只有头部 78.82%、左腕 65.93%、右腕 61.72%。这意味着,RGB-only 自动标注越大,标签噪声也可能同步放大。

动态遮挡、弱纹理与绝对深度同样关键。UMI-3D 指出单目视觉 SLAM 在遮挡、动态场景和跟踪失败时会直接损害示教数据质量;HumanEgo 的高成功率依赖稳定的手和物体跟踪,作者仍把单目深度、动态遮挡、级联误差和约 1 厘米精度瓶颈列为限制。换言之,第一视角“看起来接近机器人视角”,不等于它已经处于稳定、度量一致的机器人坐标系中。

2.3 人体与机器人之间不是一个简单的外观域差

跨本体差距至少包括外观、几何、运动学、控制频率和动作空间五层。仅把画面中的人手替换成机器人手,解决不了这些问题。HumanEgo 的 Water Flowers 消融中,纯视觉适配的成功率最高约 32.5%;加入显式的手—物 6DoF 交互表征后,使用原始人类 RGB 的成功率达到 85%。这表明真正起作用的是空间关系接口,而不是“图像更像机器人”。

对灵巧手而言,转换本身仍很昂贵。UniDex 先用逆运动学自动重定向,再用图形界面人工调整 6DoF 偏移,接触丰富片段尤其需要人工复核。在 Make Coffee 实验中,只有转换后的人类数据而没有机器人数据时,成功率仍接近 0;人类数据只能在一定区间内部分替代目标机器人示范。这说明目标机器人数据不是最后可有可无的“微调集”,而是定义可执行动作接口的锚点。

2.4 几何正确不代表物理可执行

从单目视频恢复的轨迹可能在视觉上合理,却包含穿模、悬空抓取、错误摩擦或不可能的关节配置。VideoManip 指出,即使手—物运动重建正确,对象几何、手尺度和姿态误差仍会造成无效接触;其当前系统还依赖静态或近静态相机,并在物体被手遮挡时采用固定手—物相对位姿近似。

SUGAR 将遮挡、接触伪影和重定向误差视为人类视频 motion prior 不能直接用于策略学习的原因。消融显示,去掉物理修正器会显著退化;去掉交互奖励时,机器人能模仿弯腰动作,却不能真正抬起物体。ObjRetarget 也显示,仅做姿态/关节对齐不能维持接触;移除手部几何模块后,物体滑移从 0.012 米增至 0.045 米,成功率从 75.8% 降至 53.3%。因此,接触几何、动力学可行性和力学信息应被视为独立的数据质量维度。

2.5 扩规模会同时扩噪声,过滤不能只看单帧质量

Ego-centric 数据预处理是一条长链:人物/物体检测、深度、姿态、轨迹、重定向和动作生成。任一模块的错误都可能级联。更危险的是,放宽过滤阈值常能快速增加数据量,却未必增加有效监督。在 Developing VLA 的实验中,较严格的背景轨迹过滤保留 45,157 条 episode,真实机器人得分 55.0;放宽阈值后保留 86,427 条,得分反而降到 38.8。数据数量几乎翻倍,性能却显著下降。

过滤标准还必须与任务结构和机器人能力相关。Imitating What Works 用仿真排除不可达、位姿估计错误或抓取不兼容的人类轨迹;ATHENA 提醒,多任务数据若只做全局效用排序,会把某些任务几乎清空;SIEVE 则按可复用原语和转换接口分配预算,再在每种结构内部选择稳定轨迹。由此可见,Ego-centric 数据治理至少需要同时评价:标签可信度、机器人可执行性、任务覆盖与行为结构,而不是一个统一“质量分”。

2.6 目标机器人锚定与闭环验证仍不可替代

现有证据反复出现同一模式:人类数据负责扩大语义和行为覆盖,少量机器人数据负责校准动作定义、观察分布和闭环误差。ActiveMimic 把第一视角相机运动转为主动感知先验后,仍需目标机器人的监督微调;EgoVLA、EgoScale 和 UniDex 也都显示,缺少机器人锚定时性能会大幅下降甚至归零。

因此,离线重建指标不能替代闭环成功率。标签在像素或位姿空间“接近真值”,并不保证误差不会在控制循环中累积,也不保证接触时序正确。现阶段最可信的评估,应同时报告前端恢复质量、动作/物理可执行率、按任务覆盖和目标机器人闭环结果。

3. 一个更实用的训练数据栈

综合上述证据,Ego-centric 数据不宜被视为与机器人示教同构的样本池,而应经过六层转换:

  1. 原始观测层:保留 RGB、深度、相机标定、时间戳和可用的头/手传感信息,记录缺失模态。
  2. 可观测监督层:为每条样本建立 supervision mask,区分可见事实、模型估计和不可观测变量,避免把伪标签当真值。
  3. 几何动作层:统一相机、世界、手和物体坐标系;恢复手—物 6DoF 关系,而不只保存二维轨迹或腕部位置。
  4. 本体转换层:根据目标机器人的运动学、关节限位、夹爪/灵巧手结构和控制频率进行重定向,并保留转换误差与人工修改记录。
  5. 接触与可执行性层:检查碰撞、持续接触、抓取稳定性、力/力矩和动力学可行性;必要时使用仿真或物理修正器过滤。
  6. 目标域锚定层:用少量高质量机器人示范、中间训练和真实闭环测试校准模型;数据选择同时约束任务覆盖,避免只保留“容易样本”。

这一数据栈把问题从“收多少小时视频”改写为“多少视频能通过各层质量闸门,以及每一层损失了什么”。这是比总时长更可解释的规模指标。

4. 尚未解决的研究空白

第一,现有工作使用的动作表示不统一,物体位姿、腕部位姿、指尖、关节和接触图之间缺少可比较的基准。第二,力、触觉和失败原因在自然第一视角视频中几乎不可观测,纯视觉预训练对接触任务的上限尚不清楚。第三,许多方法依赖专用头显、双目相机、已知物体模型或人工重定向,成本核算往往只报视频小时数,没有报告转换工时与算力。第四,当前结果多集中于桌面操作和有限本体,跨相机、跨操作者、跨机器人及长时任务的联合泛化证据仍不足。第五,近一年论文大多证明“某种转换管线有效”,还没有统一回答:哪类 Ego-centric 样本在什么目标机器人、什么动作空间下具有最高边际价值。

结论

Ego-centric 数据不是廉价机器人示教的直接替代品,而是一种高覆盖、低控制完备度的行为语料。它最擅长提供任务语义、交互顺序和主动视点先验;最薄弱的是精确动作、接触、力和本体适配。近一年的共同进展,是把研究重点从“缩小视觉域差”推进到“构建可执行监督接口”。对训练工程而言,优先级应是:先建立坐标、动作、接触和监督缺失的显式表示,再扩大数据规模;先用目标机器人小数据校准转换链,再讨论人类视频替代比例。否则,规模化很可能只是把不可执行、不可观测和不可验证的误差一起放大。

References

Accepted evidence

去重论文引用

  1. $τ_0$-WM: A Unified Video-Action World Model for Robotic ManipulationPengfei Zhou, Shengcong Chen, Di Chen, Jiaxu Wang, Rongjun Jin, Bingwen Zhu, Yike Pan, Songen Gu et al. · 2026
  2. ActiveMimic: Egocentric Video Pretraining with Active PerceptionXingyao Lin, Guojin Zhong, Tianyi Lu, Ziyi Ye, Yichen Zhu, Zuxuan Wu, Yu-Gang Jiang · 2026
  3. ATHENA: Accelerated Multi-Task Heterogeneous Influence Functions for Robot Data CurationTao Xu, Jiaxin Wang, Runhao Zhang, Jiayi Guan, Xianchao Zeng, Weixi Song, Xinyu Zhou, Zhetao Chen et al. · 2026
  4. Developing Vision-Language-Action Model from Egocentric VideosTomoya Yoshida, Shuhei Kurita, Taichi Nishimura, Shinsuke Mori · 2025
  5. Dexterous Manipulation Policies from RGB Human Videos via 3D Hand-Object Trajectory ReconstructionHongyi Chen, Tony Dong, Tiancheng Wu, Liquan Wang, Yash Jangir, Yaru Niu, Yufei Ye, Homanga Bharadhwaj et al. · 2026
  6. EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human DataRuijie Zheng, Dantong Niu, Yuqi Xie, Jing Wang, Mengda Xu, Yunfan Jiang, Fernando Castañeda, Fengyuan Hu et al. · 2026
  7. EgoVLA: Learning Vision-Language-Action Models from Egocentric Human VideosRuihan Yang, Qinxi Yu, Yecheng Wu, Rui Yan, Borui Li, An-Chieh Cheng, Xueyan Zou, Yunhao Fang et al. · 2025
  8. HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric VideosZhi Wang, Botao He, Kelin Yu, Seungjae Lee, Ruohan Gao, Furong Huang, Yiannis Aloimonos · 2026
  9. Imitating What Works: Simulation-Filtered Modular Policy Learning from Human VideosAlbert J. Zhai, Kuo-Hao Zeng, Jiasen Lu, Ali Farhadi, Shenlong Wang, Wei-Chiu Ma · 2026
  10. In-the-Wild Compliant Manipulation with UMI-FTHojung Choi, Yifan Hou, Chuer Pan, Seongheon Hong, Austin Patel, Xiaomeng Xu, Mark R. Cutkosky, Shuran Song · 2026
  11. ObjRetarget: An Object-Aware Motion Retargeting Framework with Anthropomorphic Arm Constraints and Polyhedral Hand ModelingYuanchuan Lai, Qing Gao, Ziyan Liang, Junjie Hu, Zhaojie Ju · 2026
  12. SIEVE: Structure-Aware Data Selection for Imitation Learning with VLA ModelsChangti Wu, Bin Yu, Zhaolong Shen, Shijie Lian, Xiaopeng Lin, Cong Huang, Zhirui Zhang, Lei Zhang et al. · 2026
  13. SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning FrameworkTianshu Wu, Xiangqi Kong, Yue Chen, Qize Yu, Hang Ye, Jia Li, Yizhou Wang, Hao Dong · 2026
  14. UMI-3D: Extending Universal Manipulation Interface from Vision-Limited to 3D Spatial PerceptionZiming Wang · 2026
  15. UniDex: A Robot Foundation Suite for Universal Dexterous Hand Control from Egocentric Human VideosGu Zhang, Qicheng Xu, Haozhe Zhang, Jianhan Ma, Long He, Yiming Bao, Zeyu Ping, Zhecheng Yuan et al. · 2026
完整证据附录(26 条正式证据事件)
  • Time range: 2025-07-15..2026-07-15
  • Events: 26
  • 每个事件一节,标题即锚点;trace-map 中的 event 链接跳转到这里。

EA-EGO-2026-0007

  • Claim: 在 EgoScale 的测量区间内,egocentric human action pretraining 确有规模收益:1K 到 20K 小时使真实机器人平均任务完成度从 0.30 升到 0.71。
  • Stance: support | Confidence: direct
  • Paper: 2602.16710 EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human Data
  • Locator: 3.3 Policy Performance Scales with Pretraining Data Size
  • Evidence: 五个数据规模的同架构实验报告单调提升,并限制结论不外推到测量区间之外。
  • Quote: “Average task completion rises monotonically from 0.30 at 1k hours to 0.71 at 20k hours”
  • Authors: ruijie-zheng; dantong-niu; yuqi-xie; et al.

EA-WMDATA-READ-0001

  • Claim: τ0-WM 使用真实机器人遥操作、UMI 式交互、人类第一视角视频与 rollout/失败轨迹的异构语料,并用按模态的监督掩码训练统一视频—动作世界模型。
  • Stance: support | Confidence: direct
  • Paper: 2606.01027 $τ_0$-WM: A Unified Video-Action World Model for Robotic Manipulation
  • Locator: Abstract (full-text section)
  • Evidence: 摘要直接报告了异构数据组成与 modality-specific supervision masks。
  • Quote: “Abstract Robotic manipulation requires models that generate executable actions while anticipating and evaluating their future consequences before physical execution. We present -World Model ( -WM), a unified video-action world model that integrates policy learning, video prediction, and action evaluation within a single future-predictive framework. Built on a shared video diffusion backbone, -WM provides two complementary interfaces. First, a video action model jointly predicts future visual lat”
  • Authors: pengfei-zhou; shengcong-chen; di-chen; et al.

EA-DQ-YEAR-READ-0010

  • Claim: 在多任务 VLA 微调中,数据质量治理要同时考虑样本效用和任务覆盖;单一全局排序会让某些任务被几乎淘汰,导致任务级覆盖坍缩。
  • Stance: support | Confidence: direct
  • Paper: 2606.16208 ATHENA: Accelerated Multi-Task Heterogeneous Influence Functions for Robot Data Curation
  • Locator: C.4 Retention Balance, Single-Task Curation, and Real-Robot Failure Modes
  • Evidence: ATHENA 指出 VLA 性能不只取决于规模,也取决于 demonstration quality,大规模冗余数据甚至可能伤害性能;在六任务真实机器人设置中,naive global influence ranking 让 Stack Bowls 只保留 13 条示教,而 MII 结合 task-local 和 cross-task influence utilities 后保留分布更均衡。
  • Quote: “To further ablate the role of Multitask Influence Interaction (MII), we visualize the retained task distributions after data curation in Fig. 8 . We consider the six-task real-robot setting with 120 demonstrations per task and an overall retention ratio of 66.7%. Without MII, naively ranking demonstrations with a single global influence score results in a highly skewed retained set: Pick Fruits, Shelf Retrieval, and Wipe Board retain 115, 113, and 104 demonstrations, respectively, whereas Stack”
  • Authors: tao-xu; jiaxin-wang; runhao-zhang; et al.

EA-EGO-2026-0002

  • Claim: 该预训练路线依赖 hand/wrist pose 注释,并仍需要中等规模目标机器人数据,因此 raw ego video 的可扩展性与可部署性之间存在结构性张力。
  • Stance: conditional | Confidence: direct
  • Paper: 2507.12440 EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videos
  • Locator: 7 Limitation
  • Evidence: 作者在限制段同时指出 pose annotations 限制数据可得性与机器人微调依赖。
  • Quote: “requires human data with hand and wrist pose annotations, which may limit data availability”
  • Authors: ruihan-yang; qinxi-yu; yecheng-wu; et al.

EA-UMI-READ-0002

  • Claim: UMI-style data is useful for contact-rich manipulation when the collection interface records force/torque, depth, pose, and grasp-force signals; the paper also implies that vision/trajectory-only data is insufficient for force-sensitive tasks.
  • Stance: conditional | Confidence: direct
  • Paper: 2601.09988 In-the-Wild Compliant Manipulation with UMI-FT
  • Locator: Abstract (full-text section)
  • Evidence: The HTML full text reports that UMI-FT mounts compact six-axis force/torque sensors on each finger, uses multimodal demonstrations to train adaptive compliance policies, and shows diverse in-the-wild data outperforming limited scene-diversity data in a skewer task.
  • Quote: “Abstract Many manipulation tasks require careful force modulation. With insufficient force the task may fail, while excessive force could cause damage. The high cost, bulky size and fragility of commercial force/torque (F/T) sensors have limited large-scale, force-aware policy learning. We introduce UMI-FT, a handheld data-collection platform that mounts compact, six-axis force/torque sensors on each finger, enabling finger-level wrench measurements alongside RGB, depth, and pose. Using the mult”
  • Authors: hojung-choi; yifan-hou; chuer-pan; et al.

EA-DQ-YEAR-READ-0003

  • Claim: 人类视频能扩大机器人学习数据来源,但其质量取决于机器人可执行性和任务兼容性;仿真过滤可把不可达、估计错误或 grasp 不兼容的轨迹排除或标注出来。
  • Stance: conditional | Confidence: direct
  • Paper: 2602.13197 Imitating What Works: Simulation-Filtered Modular Policy Learning from Human Videos
  • Locator: 3.3 Trajectory and Grasp Filtering via Simulation
  • Evidence: PSI 将人类演示转换为 6DoF object pose trajectories 后在仿真中执行,用于过滤不适合机器人学习的数据;不适合原因包括 pose estimation errors 和机器人 physically unachievable trajectories,并生成 grasp suitability labels 以学习 task-oriented grasping。
  • Quote: “Now that we have converted the human demonstrations into 6 DoF object pose trajectories, the next step is to execute them on a robot in simulation. This serves two purposes. One is to filter out those that may not be suitable for robot learning. There are two main reasons a trajectory may be unsuitable. First, pose estimation errors can lead to inaccurate trajectories. Second, the extracted trajectory may not be physically achievable by the robot. In either case, it would be harmful to train the”
  • Authors: albert-j-zhai; kuo-hao-zeng; jiasen-lu; et al.

EA-EGO-2026-0008

  • Claim: 大规模 human pretraining 仍需少量精确 aligned human-robot mid-training 才能最好地落到可执行控制;规模和本体对齐是互补条件。
  • Stance: conditional | Confidence: direct
  • Paper: 2602.16710 EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human Data
  • Locator: 3.2 Large-Scale Human Pretraining Is Key to Strong Dexterous Manipulation Policy Performance
  • Evidence: 四类 checkpoint 的消融中,pretrain+midtrain 最好;human pretraining 提供结构,mid-training 负责控制锚定。
  • Quote: “combining human pretraining with a small amount of aligned mid-training yields the best overall performance”
  • Authors: ruijie-zheng; dantong-niu; yuqi-xie; et al.

EA-EGO-2026-0011

  • Claim: Retargeted ego-human 数据只能部分替代目标机器人示范:在论文的 Make Coffee co-training 实验中,没有 robot data 时成功始终接近 0。
  • Stance: conditional | Confidence: direct
  • Paper: 2603.22264 UniDex: A Robot Foundation Suite for Universal Dexterous Hand Control from Egocentric Human Videos
  • Locator: 5.4 UniDex-Cap for Human-Robot Data Co-train
  • Evidence: 作者明确总结 human data helps but robot data is indispensable,并给出约 2:1 的局部替代斜率。
  • Quote: “Retargeted human data helps, but robot data is indispensable”
  • Authors: gu-zhang; qicheng-xu; haozhe-zhang; et al.

EA-EGO-2026-0013

  • Claim: Physics refiner 和 interaction reward 是把 Ego-centric 视频数据变成可执行技能的必要中间层;只跟踪运动会在接触任务中失败。
  • Stance: conditional | Confidence: direct
  • Paper: 2605.20373 SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework
  • Locator: 4.4 Component Analysis
  • Evidence: 组件消融中去除 Refiner 显著退化,去除 interaction reward 时机器人只模仿弯腰而无法抬起物体。
  • Quote: “Removing the Refiner leads to substantial performance degradation”
  • Authors: tianshu-wu; xiangqi-kong; yue-chen; et al.

EA-EGO-2026-0014

  • Claim: 缩小 human/robot 图像外观差距并不足以让 ego 数据可训练;Water Flowers 消融中 visual-only 最高约 32.5%,显式 hand-object 6DoF ICT 才带来大幅闭环提升。
  • Stance: conditional | Confidence: direct
  • Paper: 2605.24934 HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos
  • Locator: 4.4 What Drives Performance of HumanEgo?
  • Evidence: raw RGB、inpainting、robot RGB 和 ICT 的阶梯消融把视觉外观与空间关系作用分离。
  • Quote: “Monocular RGB encodes appearance , not the 3D spatial relationships that manipulation demands.”
  • Authors: zhi-wang; botao-he; kelin-yu; et al.

EA-EGO-2026-0017

  • Claim: 自动 RGB-only ego 标签存在明显 fidelity ceiling:严格阈值下左右 wrist pose recovery 仅约 66% 和 62%,规模化以噪声为代价。
  • Stance: conditional | Confidence: direct
  • Paper: 2606.06194 ActiveMimic: Egocentric Video Pretraining with Active Perception
  • Locator: 4.3 Egocentric Video Yields Effective Pretraining Labels
  • Evidence: HOT3D ground truth 上的 10% sample 验证给出 head/wrist 三类严格阈值 recovery rate。
  • Quote: “Under the strict tier ( , rot6d L2 ), head recovery reaches 78.82%, with left and right wrist recovery at 65.93% and 61.72%, respectively;”
  • Authors: xingyao-lin; guojin-zhong; tianyi-lu; et al.

EA-EGO-2026-0018

  • Claim: 把 camera motion 当作 viewpoint action 可提供真实的 active-perception prior,但能力必须在有 head-camera/robot fine-tuning 的系统中承接。
  • Stance: conditional | Confidence: direct
  • Paper: 2606.06194 ActiveMimic: Egocentric Video Pretraining with Active Perception
  • Locator: 4.4 The Head Camera Enables Pretrained Active Perception
  • Evidence: Restocking 中 egocentric-pretrained model 的 placement 为 24/27,SFT-only 为 6/27;移除 head camera 降到 1/27。
  • Quote: “ActiveMimic scores 24 out of 27 on placement, whereas ActiveMimic sft-only achieves only 6 out of 27”
  • Authors: xingyao-lin; guojin-zhong; tianyi-lu; et al.

EA-EGO-2026-0020

  • Claim: 显式 contact geometry 在该系统中显著减少滑移并提高成功率,说明接触结构是 Ego-centric 数据转成可执行监督的独立质量维度。
  • Stance: conditional | Confidence: direct
  • Paper: 2607.03828 ObjRetarget: An Object-Aware Motion Retargeting Framework with Anthropomorphic Arm Constraints and Polyhedral Hand Modeling
  • Locator: IV-C 1 Hand–object geometric consistency module
  • Evidence: 去除 hand geometry 后 object slip 变大且 success 下降,full ObjRetarget 最好。
  • Quote: “also removing geometric consistency causes significant slippage, contact failures, and posture collapse”
  • Authors: yuanchuan-lai; qing-gao; ziyan-liang; et al.

EA-EGO-2026-0001

  • Claim: EgoVLA 的人类第一视角预训练不能直接消除本体差距:没有机器人微调时,仿真 humanoid 的全部任务成功率为 0%。
  • Stance: limit | Confidence: direct
  • Paper: 2507.12440 EgoVLA: Learning Vision-Language-Action Models from Egocentric Human Videos
  • Locator: 5.2 Humanoid Robot Evaluation
  • Evidence: 论文把失败归因于外观、感知和运动学错配,并用同一 benchmark 的零样本结果明确展示。
  • Quote: “zero-shot deployment on humanoid robots without fine-tuning on robot data results in 0% success across all tasks”
  • Authors: ruihan-yang; qinxi-yu; yecheng-wu; et al.

EA-EGO-2026-0003

  • Claim: 从 egocentric 视频恢复的 object trajectory 不是完整机器人动作:该路线无法获得 gripper state,只能把动作表示为 9D 物体位姿增量。
  • Stance: limit | Confidence: direct
  • Paper: 2509.21986 Developing Vision-Language-Action Model from Egocentric Videos
  • Locator: III-C Policy Training
  • Evidence: 策略训练段明确说明 gripper state 缺失,并以 object pose displacement 作为替代动作。
  • Quote: “Because gripper states cannot be obtained from Section III-B , each action is represented by a 9-dimensional vector”
  • Authors: tomoya-yoshida; shuhei-kurita; taichi-nishimura; et al.

EA-EGO-2026-0004

  • Claim: Ego-centric 轨迹构建存在规模—质量冲突:保留更多疑似背景/错误轨迹会显著降低真实机器人表现。
  • Stance: limit | Confidence: direct
  • Paper: 2509.21986 Developing Vision-Language-Action Model from Egocentric Videos
  • Locator: IV-C Ablation Study
  • Evidence: BGTS=1.0 保留 86,427 episodes 但真实机器人分数低于 BGTS=0.7 的 45,157 episodes。
  • Quote: “Setting an appropriate curation threshold is crucial to balancing the scale and quality of our dataset”
  • Authors: tomoya-yoshida; shuhei-kurita; taichi-nishimura; et al.

EA-EGO-2026-0005

  • Claim: 单目 RGB 人类视频恢复出的 hand-object 轨迹常不具物理可执行性;对象几何、手尺度/姿态误差会形成穿模、无效接触和抓取失败。
  • Stance: limit | Confidence: direct
  • Paper: 2602.09013 Dexterous Manipulation Policies from RGB Human Videos via 3D Hand-Object Trajectory Reconstruction
  • Locator: III-B Dexterous Grasp and Manipulation Learning
  • Evidence: 方法段明确说明重建运动正确时,机器人—对象交互仍可能因几何误差而无效。
  • Quote: “the resulting robot–object interactions are not always physically feasible due to reconstruction errors”
  • Authors: hongyi-chen; tony-dong; tiancheng-wu; et al.

EA-EGO-2026-0006

  • Claim: 当前 VideoManip 依赖静态或近静态相机,并在真实闭环中用固定 hand-object 相对位姿绕过手部遮挡,限制了动态第一视角数据的可用范围。
  • Stance: limit | Confidence: direct
  • Paper: 2602.09013 Dexterous Manipulation Policies from RGB Human Videos via 3D Hand-Object Trajectory Reconstruction
  • Locator: V Conclusion, Limitations, and Future Work
  • Evidence: 作者在限制段明确列出 dynamic camera 未覆盖;实验段说明对象点云被 LEAP Hand 遮挡时采用固定相对位姿近似。
  • Quote: “The current framework assumes static or approximately static camera setups”
  • Authors: hongyi-chen; tony-dong; tiancheng-wu; et al.

EA-EGO-2026-0009

  • Claim: Ego-centric 数据的动作接口会决定训练成败:wrist-only 缺失手指/接触时序,小 fingertip 误差又会映射成不合理关节和接触丢失。
  • Stance: limit | Confidence: direct
  • Paper: 2602.16710 EgoScale: Scaling Dexterous Manipulation with Diverse Egocentric Human Data
  • Locator: 3.6 Hand Action Space Design for Human Pretraining
  • Evidence: 动作空间消融中 wrist-only 普遍较差,fingertip mapping 在 Cards/Bottle 等接触敏感任务不稳定。
  • Quote: “Small errors in fingertip pose often lead to implausible joint configurations after mapping”
  • Authors: ruijie-zheng; dantong-niu; yuqi-xie; et al.

EA-EGO-2026-0010

  • Claim: 将 egocentric hand trajectories 转为机器人可执行数据仍需 human-in-the-loop retargeting:基础坐标/形态偏差和 contact-rich 片段要人工校准。
  • Stance: limit | Confidence: direct
  • Paper: 2603.22264 UniDex: A Robot Foundation Suite for Universal Dexterous Hand Control from Egocentric Human Videos
  • Locator: 3.2.1 Kinematic Retargeting
  • Evidence: 论文的两阶段 retargeting 先自动 IK,再用 GUI 调整 6DoF offset;接触片段需人工复核。
  • Quote: “The whole pipeline is a two-stage, human-in-the-loop retargeting procedure”
  • Authors: gu-zhang; qicheng-xu; haozhe-zhang; et al.

EA-UMI-READ-0004

  • Claim: Original vision-only UMI data can fail to be useful in scenes with occlusion, dynamic changes, feature-poor regions, or tracking failure; adding LiDAR-centric 3D sensing improves data quality and expands the feasible task distribution.
  • Stance: limit | Confidence: direct
  • Paper: 2604.14089 UMI-3D: Extending Universal Manipulation Interface from Vision-Limited to 3D Spatial Perception
  • Locator: Abstract (full-text section)
  • Evidence: The HTML full text states that monocular visual SLAM makes UMI vulnerable to occlusions, dynamic scenes, and tracking failures, and reports that LiDAR-centric SLAM improves pose-estimation robustness and demonstration data quality under challenging real-world conditions.
  • Quote: “Abstract We present UMI-3D, a multimodal extension of the Universal Manipulation Interface (UMI) for robust and scalable data collection in embodied manipulation. While UMI enables portable, wrist-mounted data acquisition, its reliance on monocular visual SLAM makes it vulnerable to occlusions, dynamic scenes, and tracking failures, limiting its applicability in real-world environments. UMI-3D addresses these limitations by introducing a lightweight and low-cost LiDAR sensor tightly integrated i”
  • Authors: ziming-wang

EA-EGO-2026-0012

  • Claim: 从人类视频恢复的 motion prior 会因遮挡、接触伪影和 retargeting 误差而物理不合理,不能直接当作 humanoid policy 的示范。
  • Stance: limit | Confidence: direct
  • Paper: 2605.20373 SUGAR: A Scalable Human-Video-Driven Generalizable Humanoid Loco-Manipulation Learning Framework
  • Locator: 1 Introduction
  • Evidence: 引言直接列出三类误差并说明它们使数据 unsuitable for direct policy learning。
  • Quote: “Severe occlusion, contact artifacts, and retargeting errors render this data physically implausible for direct imitation”
  • Authors: tianshu-wu; xiangqi-kong; yue-chen; et al.

EA-EGO-2026-0015

  • Claim: HumanEgo 的高成功率依赖强 hand/object tracking 前端;单目绝对深度、动态遮挡、模块级联误差和亚厘米接触精度仍是未解决困难。
  • Stance: limit | Confidence: direct
  • Paper: 2605.24934 HumanEgo: Zero-Shot Robot Learning from Minutes of Human Egocentric Videos
  • Locator: 5 Conclusion
  • Evidence: 作者在 limitation 段逐项列出 stereo hand tracking、occlusion-robust tracking、cascading failures 和 1 cm plateau。
  • Quote: “monocular substitutes drop real-world success sharply”
  • Authors: zhi-wang; botao-he; kelin-yu; et al.

EA-EGO-2026-0016

  • Claim: Ego-centric wrist trajectory 与相机自运动天然耦合;若不统一参考系,动作监督会把 camera rotation/translation 错算为 hand motion。
  • Stance: limit | Confidence: direct
  • Paper: 2606.06194 ActiveMimic: Egocentric Video Pretraining with Active Perception
  • Locator: 3 Method
  • Evidence: 方法段明确说明 current-frame wrist pose 与 first-frame camera path 的坐标差异会混合两类位移。
  • Quote: “using these wrist poses directly as action supervision would therefore conflate wrist movement with camera motion”
  • Authors: xingyao-lin; guojin-zhong; tianyi-lu; et al.

EA-EGO-2026-0019

  • Claim: Ego-human motion 的 pose/joint 对齐只能保证自由空间几何相似;不显式建模 hand-object contact,就难以保持持续接触、物体交换和多阶段操作。
  • Stance: limit | Confidence: direct
  • Paper: 2607.03828 ObjRetarget: An Object-Aware Motion Retargeting Framework with Anthropomorphic Arm Constraints and Polyhedral Hand Modeling
  • Locator: II-B Human-to-Robot Motion Retargeting
  • Evidence: 相关工作和引言都指出现有方法多假设 object-free/weak-contact,忽略手臂与手的不同功能。
  • Quote: “most methods assume object-free or weak-contact settings and focus on geometric consistency or joint error minimization”
  • Authors: yuanchuan-lai; qing-gao; ziyan-liang; et al.

EA-DQ-YEAR-READ-0015

  • Claim: SIEVE 按可复用原语组合和转换接口分配选择预算,再优先保留各组合模式中稳定、中心的轨迹;论文报告其用 50% 示教和 50% 训练步数可优于全量训练。
  • Stance: limit | Confidence: direct
  • Paper: 2607.06442 SIEVE: Structure-Aware Data Selection for Imitation Learning with VLA Models
  • Locator: Introduction
  • Evidence: 引言的贡献列表同时说明了结构暴露、学习友好轨迹选择和半量数据超过全量训练的结果。
  • Quote: “Our contributions are as follows: • We propose a primitive-compositional view of trajectory utility, realized by Primitive Discovery and Structural Exposure Allocation, which allocate selection budgets according to reuse-aware primitive and transition exposure under diminishing returns. • We introduce Learning-Friendly Trajectory Selection, which selects medoid trajectories within each composition-pattern bucket to favor central, stable, and predictable realizations for behavior cloning. • We pr”
  • Authors: changti-wu; bin-yu; zhaolong-shen; et al.

References

Interactive Wiki

进入交互阅读

静态页采用默认版本;v0、v0、v0可在同一交互 Wiki 中阅读,不建立相互竞争的索引页。

Method and limits

研究方法、证据边界与引用说明

本页由当前结算的证据层自动派生,不改写已结算 run。论文只有在完整非 OCR 全文、结构化论文笔记和 claim-support audit 通过后,才进入接受证据。

页面结论是对所列论文与研究窗口的综合,不等同于无限外推。引用本专题时,请同时保留本页 canonical 地址、访问日期,并在需要时回溯原论文。