Feed Entities Podcasts Deep Dives Developers About

[KongchangAI]

Real-time AI knowledge base. Every claim is traceable to its source, with data freshness under 24 hours.

Powered by 240+ first-hand AI sources

Knowledge

Feed
Hot Entities
Podcasts

Developers

CLI
MCP Server
REST API
About

© 2026 KongchangAI. The freshest, most accurate AI knowledge

Unverified50% confidenceFactExact time

JEPA过程无需解码器重建像素，编码器将观测压缩为语义嵌入，预测网络在嵌入空间完成跨时间步状态预测

1

Sources

50%

Confidence

Long-term

Relevance

7/19/2026

First Seen

Sources

LeMario：用超级马里奥训练JEPA世界模型的开源实践

Related Claims

UnverifiedJEPA 借鉴了 BYOL 的动量编码器思想，同时吸收了 MAE 的掩码预测机制，但将重建目标替换为抽象嵌入68% similar UnverifiedJEPA 引入非对称结构，其中目标编码器的梯度通常通过指数移动平均（EMA）更新，而非直接反向传播66% similar UnverifiedJEPA 与自监督学习谱系的关键区别在于明确将'预测'而非'重建'或'对比'作为核心训练信号，且预测目标是经过编码的抽象表示65% similar UnverifiedLeCun在2022年提出的JEPA（Joint Embedding Predictive Architecture）在表征空间中直接预测目标嵌入，而非重构原始输入63% similar UnverifiedHa与Schmidhuber的世界模型采用三模块架构：视觉编码器V负责感知压缩、循环神经网络M负责时序记忆、控制器C负责决策60% similar

Cite This Claim

Stable URI

https://kongchang.com/claim/564469

API

curl https://kongchang.com/api/v1/knowledge/claims/564469

MCP

get_claim(id=564469)