# TwT: Thinking without Tokens by Habitual Reasoning Distillation with Multi-Teachers' Guidance

Authors: Jingxian Xu, Mengyu Zhou, Weichang Liu, Hanbing Liu, Shi Han, Dongmei Zhang.

Selects diverse worked solutions from multiple teachers, then trains a student on full solutions, compressed solutions, and finally answers alone. The intermediate compression stage uses the student’s responses to guide how a teacher shortens its explanations.

First public: 2025-03-31.
ArXiv initial submission: 2025-03-31.

[Original paper](https://arxiv.org/abs/2503.24198) · [PDF](https://arxiv.org/pdf/2503.24198) · [HTML](https://arxiv.org/html/2503.24198v2) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2503.24198/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2503.24198/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2503.24198/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2503.24198/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.

Reading-list version 0.15.2; literature checked through 2026-11-03.