Research paper
Latent Thinking Optimization: Your Latent Reasoning Language Model Secretly Encodes Reward Signals in Its Latent Thoughts
ICLR 2026
Abstract
The complete author-written abstract has not yet been supplied to this site. The verified publication metadata and primary source are available below; experimental claims should be checked against that source.
Citation
@article{2026-latent-thinking-optimization,
title = {Latent Thinking Optimization: Your Latent Reasoning Language Model Secretly Encodes Reward Signals in Its Latent Thoughts},
author = {H Du and Y Dong and X Ning},
year = {2026},
journal = {ICLR 2026},
eprint = {2509.26314},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2509.26314},
}