# Towards Looped Models Done Right, Part I: Topology, Input Injection, Recurrent-State Design

Authors: Benhao Huang, Chufan Shi, Junlin Chen, Shicheng Wen, Zhengzhong Liu, Eric Xing, Xuezhe Ma.

Compares where to place shared layers, how to reintroduce the input, and how to organize recurrent states in looped language models. Uses matched dense-model comparisons and a mixture-of-experts study to separate these design choices.

First public: 2026-10-16.
ArXiv initial submission: 2026-10-16.

[Original paper](https://arxiv.org/abs/2610.23036) · [PDF](https://arxiv.org/pdf/2610.23036) · [HTML](https://arxiv.org/html/2610.23036v1) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2610.23036/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2610.23036/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2610.23036/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2610.23036/record.json)

## 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.