Su and colleagues: rotary position embeddings in RoFormer

What changed

Encoded token position through rotations of query and key vectors, making their attention inner products depend on relative offsets.

Scientific scope

Encoded token position through rotations of query and key vectors, making their attention inner products depend on relative offsets. This record binds the five-author April 2021 first preprint and preliminary non-biological experiments. Later revised-paper authorship, extended evaluations and long-context variants are separate.

Evidence

The linked original sources describe the stated contribution and its evidence limits. WRS has not independently rerun the experiments or certified a proof. This record binds the five-author April 2021 first preprint and preliminary non-biological experiments. Later revised-paper authorship, extended evaluations and long-context variants are separate.

Contributors

• Source-listed authors and technical contributors · SCIENCE: Source-listed authors or explicitly credited technical contributors for the cited versions. Family entries retain source-specific attribution; a combined list is not one original-paper byline or equal individual work. Names alone do not establish resolved identities.

• RoFormer: Enhanced Transformer with Rotary Position Embedding
• Zhuiyi Technology: original RoFormer project

Provisional WRS assessment, edition 2026-10-16.1. Independent expert calibration is pending. This grade does not certify correctness or assign ownership.

Full scientific record · Browse contribution grades · Request a correction