# 30000小时自我中心视频没有教会什么

> 原文链接：https://robotics.hellobike.com/papers/2026-10-08-2610-12464
> 频道：papers | 更新日期：2026-10-08 | 标签：arxiv, cs.RO
> 核心摘要：通过扩展3万小时自我中心视频训练并改进视觉条件与监督，在离分布基准验证世界模型交互保真度，发现物体动态仍滞后。

# 30000小时自我中心视频没有教会什么

::: v-pre
> 原文链接：[What 30,000 Hours of Ego-centric Video Does Not Teach](https://arxiv.org/abs/2610.12464)（arXiv 2610.12464 · 发布日期：2026-10-08）

**What 30,000 Hours of Ego-centric Video Does Not Teach**

通过扩展3万小时自我中心视频训练并改进视觉条件与监督，在离分布基准验证世界模型交互保真度，发现物体动态仍滞后。

## 核心要点

- 动机是评估仅靠扩展自我中心人类视频，能把世界模型推向多远。
- 方法直接评估主体与物体交互保真度，并设计视觉条件和新的监督方案。
- 训练数据增加100倍提升不均：主体建模较好，物体保真度低且提升慢。
- 新的监督将容量从场景外观转向物体动态，改善物体保真度但差距仍大。

**验证范围与局限**：验证范围为离分布基准上的交互保真度评估，并迁移到下游人形建模。

## 英文摘要 (Abstract)

World models offer a promising alternative to physics-based simulators, yet remain far from practical deployment. We ask how far scaling ego-centric human video takes them, using a dataset of 30,000 hours spanning over 1,000 scene types and 14,000 contributors. Rather than relying on opaque downstream metrics, we directly evaluate agent and object-interaction fidelity on a challenging out-of-distribution benchmark. Increasing training data by 100x improves both, but unevenly: the agent is modeled well, while object fidelity remains far lower and improves slowly. We show that the agent gains need not come from data, and a careful visual conditioning design saturates fidelity with a fraction of it, which lets us measure object fidelity on its own and discover its saturation point. We then introduce a supervision scheme that shifts capacity from scene appearance toward object dynamics, improving object fidelity though a substantial gap remains. Finally, our conclusions transfer to downstream humanoid modeling. Overall, our results suggest that scaling ego-centric data brings agent modeling close to its limit while leaving its effects on the world far behind, and that closing this gap will depend on how models are trained, not only on how much data they see.

作者：Jiahua Dong, Anurag Bagchi, Yash Jangir, Muhammad Zubair Irshad, Sergey Zakharov, Martial Hebert, Homanga Bharadhwaj, Yu-Xiong Wang, Vitor Guizilini, Pavel Tokmakov
:::

## 原始参考来源 (Sources)

- [arXiv 2610.12464：What 30,000 Hours of Ego-centric Video Does Not Teach](https://arxiv.org/abs/2610.12464)