Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training
Abstract
Mixture-of-Experts (MoE) architectures within Large Reasoning Models (LRMs) have achieved impressive reasoning capabilities by selectively activating experts to facilitate structured cognitive processes. Despite notable advances, existing reasoning models often suffer from cognitive inefficiencies like overthinking and underthinking. To address these limitations, we introduce a novel inference-time steering methodology called Reinforcing Cognitive Experts (RICE), designed to improve reasoning depth and efficiency without additional training or complex heuristics. Leveraging normalized Pointwise Mutual Information (nPMI), we systematically identify specialized experts, termed cognitive experts that orchestrate meta-level reasoning operations characterized by tokens like <think>. Empirical evaluations with leading MoE-based LRMs (DeepSeek-R1 and Qwen3-235B) on rigorous quantitative and scientific reasoning benchmarks (AIME and GPQA Diamond) demonstrate noticeable and consistent improvements in reasoning accuracy, cognitive efficiency, and cross-domain generalization. Crucially, our lightweight approach substantially outperforms prevalent reasoning-steering techniques, such as prompt design and decoding constraints, while preserving the model's general instruction-following skills. These results highlight reinforcing cognitive experts as a promising, practical, and interpretable direction to enhance cognitive efficiency within advanced reasoning models.
Cite
Text
Wang et al. "Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training." Advances in Neural Information Processing Systems, 2025.Markdown
[Wang et al. "Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training." Advances in Neural Information Processing Systems, 2025.](https://mlanthology.org/neurips/2025/wang2025neurips-two/)BibTeX
@inproceedings{wang2025neurips-two,
title = {{Two Experts Are All You Need for Steering Thinking: Reinforcing Cognitive Effort in MoE Reasoning Models Without Additional Training}},
author = {Wang, Mengru and Chen, Xingyu and Wang, Yue and He, Zhiwei and Xu, Jiahao and Liang, Tian and Liu, Qiuzhi and Yao, Yunzhi and Wang, Wenxuan and Ma, Ruotian and Mi, Haitao and Zhang, Ningyu and Tu, Zhaopeng and Li, Xiaolong and Yu, Dong},
booktitle = {Advances in Neural Information Processing Systems},
year = {2025},
url = {https://mlanthology.org/neurips/2025/wang2025neurips-two/}
}