# Looped Language Models: A Survey and Meta-Analysis of Recurrent-Depth Transformers

Authors: Seyedarmin Azizi, Erfan Baghaei Potraghloo, Massoud Pedram.

Surveys models that reuse transformer layers across repeated hidden-state updates. Compares their architecture, training and inference choices, separating parameter-matched results from compute-matched results.

First public: 2026-10-10.
Preprints.org preprint, version 1: 2026-10-15.
DOI: 10.20944/preprints202610.1404.v1.

[Original paper](https://www.preprints.org/manuscript/202610.1404) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/azizi2026loopedsurvey/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/azizi2026loopedsurvey/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/azizi2026loopedsurvey/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/azizi2026loopedsurvey/record.json)

Date note: Author repository PDF release. [Source](https://github.com/ArminAzizi98/looped-lm-survey/commit/f55bb84a885d6800a3776ba73308750396527b86).

## 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.
- [Analysis and evaluation](https://latent-implicit-thinking.github.io/topics/implicit-reasoning-evaluation/): Measure how well these methods work, investigate what their internal computation does, and identify where they fail. This includes experiments, studies of model internals, and mathematical results.

Reading-list version 0.11.0; literature checked through 2026-10-26.