# Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens

Authors: Wei-Lin Chen, Liqian Peng, Tian Tan, Chao Zhao, Blake JianHang Chen, Ziqian Lin, Alec Go, Yu Meng.

Measures how late in a transformer’s layers each token prediction settles, then uses this signal to score a reasoning trace. Tests early rejection of low-scoring samples on math and science questions, measuring savings in generated tokens.

First public: 2026-02-13.
ArXiv initial submission: 2026-02-13.

[Original paper](https://arxiv.org/abs/2602.13517) · [PDF](https://arxiv.org/pdf/2602.13517) · [HTML](https://arxiv.org/html/2602.13517v2) · [Reading-list entry](https://latent-implicit-thinking.github.io/papers/2602.13517/)
[Preprint BibTeX](https://latent-implicit-thinking.github.io/papers/2602.13517/citation.bib) · [Preprint CSL-JSON](https://latent-implicit-thinking.github.io/papers/2602.13517/citation.csl.json) · [JSON record](https://latent-implicit-thinking.github.io/papers/2602.13517/record.json)

## Published version

Think Deep, Not Just Long: Measuring LLM Reasoning Effort via Deep-Thinking Tokens. Wei-Lin Chen, Liqian Peng, Tian Tan, Chao Zhao, Jianhang Chen, Ziqian Lin, Alec Go, Yu Meng. Proceedings of the 43rd International Conference on Machine Learning, 2026.
[Publication](https://proceedings.mlr.press/v306/chen26di.html) · [BibTeX](https://latent-implicit-thinking.github.io/papers/2602.13517/publication.bib) · [CSL-JSON](https://latent-implicit-thinking.github.io/papers/2602.13517/publication.csl.json)
[Publication record](https://proceedings.mlr.press/v306/chen26di.html).

## Approaches

- [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.