# STORM: Internalized Modeling for Spatial-Temporal Reasoning in Video-Language Models

Authors: Yiming Liang, Yixiao Chen, Yiyang Zhou, Yixuan Wang, Shoubin Yu, Andong Deng, Fuxiao Liu, Qin Zhang, Chen Chen, Mohit Bansal, Huaxiu Yao.

Trains a video-language model to answer through a bounded sequence of continuous hidden states. First aligns these states with features from generated thought videos, then trains with answer-only supervision; inference uses the latent sequence without generating intermediate videos.

First public: 2026-05-25.
ArXiv initial submission: 2026-05-25.

[Original paper](https://arxiv.org/abs/2605.26014) · [PDF](https://arxiv.org/pdf/2605.26014) · [HTML](https://arxiv.org/html/2605.26014v1) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2605.26014/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2605.26014/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2605.26014/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2605.26014/record.json)

## Approaches

- [Internalized reasoning](https://latent-implicit-thinking.github.io/topics/internalized-reasoning/): Learn from worked solutions so that some or all of the reasoning can later happen without writing those steps. Training may use a separate teacher model or progressively remove written steps from the model being trained.
- [Continuous-state reasoning](https://latent-implicit-thinking.github.io/topics/continuous-states/): Use vectors—lists of numbers—to carry intermediate reasoning from one step to the next, instead of selecting a text token at every step. Some methods use the model's hidden states; others combine several candidate token representations.

Reading-list version 0.12.0; literature checked through 2026-11-02.