# LTA-thinker: Latent Thought-Augmented Training Framework for Large Language Models on Complex Reasoning

Authors: Jiaqi Wang, Binquan Ji, Haibo Luo, Yiyang Qi, Ruiting Li, Huiyan Wang, Yuantao Han, Cangyi Yang, Jiaxu Zhang, Feiliang Ren.

Trains a small transformer module to produce continuous thought vectors for a frozen language model. Aligns those vectors with the question and important solution steps; the language model then generates a written explanation and answer.

First public: 2025-09-16.
ArXiv initial submission: 2025-09-16.

[Original paper](https://arxiv.org/abs/2509.12875) · [PDF](https://arxiv.org/pdf/2509.12875) · [HTML](https://arxiv.org/html/2509.12875v3) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2509.12875/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2509.12875/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2509.12875/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2509.12875/record.json)

## Approaches

- [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 0.15.2; literature checked through 2026-11-03.