← Back to the collection
Full-stack recurrence ·
SpiralFormer
SpiralFormer: Looped Transformers Can Learn Hierarchical Dependencies via Multi-Resolution Recursion
Explores recurrent computation at different sequence resolutions to change the cost of repeated processing.
Inside the method
[ Multi-resolution states ] × R
Simplified conceptual schematic. Consult the paper for the complete architecture.
- Recurrence family
- Full-stack recurrence
- Depth control
- Multi-resolution recurrence
- KV / state strategy
- See paper
Reading note
No official code or weights were confirmed in this review. This entry is not the similarly named speech model.
Sources checked 2026-09-15. This catalog does not imply independent reproduction.
Cite this work
@misc{spiralformer2026,
title = {SpiralFormer: Looped Transformers Can Learn Hierarchical Dependencies via Multi-Resolution Recursion},
author = {Chengting Yu and Xiaobo Shu and Yadao Wang and Yizhen Zhang and Haoyi Wu and You Wu and Rujiao Long and Ziheng Chen and Yuchi Xu and Wenbo Su and Bo Zheng},
year = {2026},
eprint = {2602.11698},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2602.11698}
}