Open Models Almanac
Methodsmodel architecturepositional encoding

specific method · filed under model architecture

Proportional Rotary Position Embedding

A named proportional variant of rotary position embedding used in the cited global layers.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See positional encoding for the methods filed beside it.

Also called Proportional RoPE (p-RoPE), proportional rotary position embeddings.

sources
2
model
1
lab adopt it
1
strongest
used

Documented in

Google DeepMind

Evidence

2 quotes from 2 sources, one card per model, each opening with its strongest.

How sources treat it
used 2

Gemma 4 31B

Google DeepMindadopted2 sources

To optimize memory for long contexts, global layers feature unified Keys and Values, and apply Proportional RoPE (p-RoPE).

usedmodel architecturein Gemma 4Google DeepMind
1 more from this model

global layers feature unified Keys and Values, and apply Proportional RoPE (p-RoPE).

usedmodel architecturein Gemma 4Google DeepMind
↑