CRoPE: Efficient Parametrization of Rotary Positional Embedding
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866915904869105664 |
|---|---|
| author | Lou, Beicheng Xu, Zifei Wong, Vivian W. H. |
| author_facet | Lou, Beicheng Xu, Zifei Wong, Vivian W. H. |
| contents | Rotary positional embedding has become the state-of-the-art approach to encode position information in transformer-based models. While it is often succinctly expressed in complex linear algebra, we note that the actual implementation of $Q/K/V$-projections is not equivalent to a complex linear transformation. We argue that complex linear transformation is a more natural parametrization and saves near 50\% parameters within the attention block. We show empirically that removing such redundancy has negligible impact on the model performance. Our modification achieves more efficient parameter usage, as well as a cleaner interpretation of the representation space. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2601_02728 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | CRoPE: Efficient Parametrization of Rotary Positional Embedding Lou, Beicheng Xu, Zifei Wong, Vivian W. H. Machine Learning Rotary positional embedding has become the state-of-the-art approach to encode position information in transformer-based models. While it is often succinctly expressed in complex linear algebra, we note that the actual implementation of $Q/K/V$-projections is not equivalent to a complex linear transformation. We argue that complex linear transformation is a more natural parametrization and saves near 50\% parameters within the attention block. We show empirically that removing such redundancy has negligible impact on the model performance. Our modification achieves more efficient parameter usage, as well as a cleaner interpretation of the representation space. |
| title | CRoPE: Efficient Parametrization of Rotary Positional Embedding |
| topic | Machine Learning |
| url | https://arxiv.org/abs/2601.02728 |