---
id: 20260926-T0-07
title: "新训练方法让世界模型学会客体永久性"
title_en: "New Training Method Teaches World Models Object Permanence"
url: https://ai.daily.yangsir.net/daily/20260926-T0-07
issue_date: 2026-09-26
publish_date: 2026-09-25T04:00:00.000Z
category: research
source_name: "arXiv cs.AI"
source_url: https://arxiv.org/abs/2609.28654
---

# 新训练方法让世界模型学会客体永久性

客体永久性和固体性是人类认知的基本先验，但当前视频生成模型作为世界模型的代表，在这方面仍有欠缺。研究者提出一种训练方法，通过设计特定任务让模型学会追踪被遮挡后重新出现的物体，以及理解物体不可穿透的性质。实验显示，经过训练的模型在相关场景中的生成质量明显提升。这为构建更符合物理直觉的世界模型提供了可行路径。

## English Version

**New Training Method Teaches World Models Object Permanence**

Object permanence and solidity are hallmarks of human cognitive priors, but video generation models — a leading class of world models — still lack them. Researchers propose a training method that teaches models to track objects that disappear and reappear, and to understand that objects cannot pass through each other. Experiments show trained models generate noticeably better results in these scenarios. The work offers a practical path toward world models with more physical intuition.

---

**来源**：[arXiv cs.AI](https://arxiv.org/abs/2609.28654)

**详情页**：https://ai.daily.yangsir.net/daily/20260926-T0-07

---

*智语观潮 · Daily — https://ai.daily.yangsir.net/llms.txt*