# Retrofitting Recurrent Vision-Language Models from Pretrained Backbones

Authors: Madhav Iyengar, Andreas Geiger, Hilde Kühne.

Prunes layers from a pretrained vision-language model, then loops the remaining middle block with a learned adapter. Uses text training, projector realignment, and visual instruction tuning to recover performance; on TinyLLaVA, the resulting model improves reasoning benchmarks with fewer language-model parameters.

First public: 2026-10-15.
ArXiv initial submission: 2026-10-15.

[Original paper](https://arxiv.org/abs/2610.22075) · [PDF](https://arxiv.org/pdf/2610.22075) · [HTML](https://arxiv.org/html/2610.22075v1) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2610.22075/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2610.22075/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2610.22075/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2610.22075/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.

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