Joint-Embedding Predictive Architecture

Definition

A framework for learning world models by encoding both a full input and a corrupted version and predicting the abstract representation (not the raw pixels) of the full input from the corrupted one, allowing the encoder to discard unpredictable noise.

Cast within

No topic-region cast yet — this appears once Joint-Embedding Predictive Architecture's compiled claims are aligned into a topic region's argument tree.

Claims mentioning Joint-Embedding Predictive Architecture (3)

Your notes

Loading notes...