Deep Think with Confidence
Abstract
Large Language Models (LLMs) have shown great potential in reasoning tasks through test-time scaling methods like self-consistency with majority voting. However, this approach often leads to diminishing returns in accuracy and high computational overhead. To address these challenges, we introduce \textbf{Deep Think with Confidence (DeepConf)}, a simple yet powerful method that enhances both reasoning efficiency and performance at test time. DeepConf leverages model-internal confidence signals to dynamically filter out low-quality reasoning traces during or after generation. It requires no additional model training or hyperparameter tuning and can be seamlessly integrated into existing serving frameworks. We evaluate DeepConf across a variety of tasks and the latest open-source models, including Qwen3 and GPT-OSS series. Notably, on challenging benchmarks such as AIME 2025, DeepConf@512 achieves up to 99.9\% accuracy and reduces generated tokens by up to 84.7\% compared to full parallel thinking. Our code is available at https://github.com/facebookresearch/deepconf
Cite
Text
Fu et al. "Deep Think with Confidence." International Conference on Learning Representations, 2026.Markdown
[Fu et al. "Deep Think with Confidence." International Conference on Learning Representations, 2026.](https://mlanthology.org/iclr/2026/fu2026iclr-deep/)BibTeX
@inproceedings{fu2026iclr-deep,
title = {{Deep Think with Confidence}},
author = {Fu, Yichao and Wang, Xuewei and Zhang, Hao and Tian, Yuandong and Zhao, Jiawei},
booktitle = {International Conference on Learning Representations},
year = {2026},
url = {https://mlanthology.org/iclr/2026/fu2026iclr-deep/}
}