# LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation

Authors: Ahmadreza Jeddi, Marco Ciccone, Babak Taati.

Trains a shared transformer stack on long and short computation paths, conditioning each pass on its location and step size. A consistency loss aligns shorter paths with the longest path so one model can run under different depth budgets.

First public: 2026-02-11.
ArXiv initial submission: 2026-02-11.

[Original paper](https://arxiv.org/abs/2602.11451) · [PDF](https://arxiv.org/pdf/2602.11451) · [HTML](https://arxiv.org/html/2602.11451v1) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2602.11451/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2602.11451/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2602.11451/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2602.11451/record.json)

## Published version

LoopFormer: Elastic-Depth Looped Transformers for Latent Reasoning via Shortcut Modulation. Ahmadreza Jeddi, Marco Ciccone, Babak Taati. International Conference on Learning Representations, 2026.
[Publication](https://arxiv.org/abs/2602.11451) · [BibTeX](https://latent-implicit-thinking.github.io/papers/2602.11451/publication.bib) · [CSL-JSON](https://latent-implicit-thinking.github.io/papers/2602.11451/publication.csl.json)
[Publication record](https://iclr.cc/virtual/2026/poster/10009450).

## Approaches

- [Recurrent reasoning](https://latent-implicit-thinking.github.io/topics/recurrent-reasoning/): Repeat parts of the model or keep intermediate information in memory while working through a problem. Some methods also learn how many passes to run.

Reading-list version 0.15.2; literature checked through 2026-11-03.