Fig.1

Concept

3D RoPE

3D RoPE (three-dimensional Rotary Position Embedding) extends the rotary position encoding from language models to data with three positional axes, typically time plus height plus width in video. Standard RoPE encodes a token's 1D position by rotating pairs of dimensions in…

The rest of “3D RoPE” is a premium feature: every concept in the library gets a precise, practitioner-focused write-up like this one, cross-linked straight from the paper summaries that use it.

Log in to unlock

← Back to the library