The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2025
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866909828871356416 |
|---|---|
| author | Fichtl, Alexander M. Bohn, Jeremias Kelber, Josefin Mosca, Edoardo Groh, Georg |
| author_facet | Fichtl, Alexander M. Bohn, Jeremias Kelber, Josefin Mosca, Edoardo Groh, Georg |
| contents | Transformers have dominated sequence processing tasks for the past seven years -- most notably language modeling. However, the inherent quadratic complexity of their attention mechanism remains a significant bottleneck as context length increases. This paper surveys recent efforts to overcome this bottleneck, including advances in (sub-quadratic) attention variants, recurrent neural networks, state space models, and hybrid architectures. We critically analyze these approaches in terms of compute and memory complexity, benchmark results, and fundamental limitations to assess whether the dominance of pure-attention transformers may soon be challenged. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_05364 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures Fichtl, Alexander M. Bohn, Jeremias Kelber, Josefin Mosca, Edoardo Groh, Georg Computation and Language Transformers have dominated sequence processing tasks for the past seven years -- most notably language modeling. However, the inherent quadratic complexity of their attention mechanism remains a significant bottleneck as context length increases. This paper surveys recent efforts to overcome this bottleneck, including advances in (sub-quadratic) attention variants, recurrent neural networks, state space models, and hybrid architectures. We critically analyze these approaches in terms of compute and memory complexity, benchmark results, and fundamental limitations to assess whether the dominance of pure-attention transformers may soon be challenged. |
| title | The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2510.05364 |