Conference paper · 2025 · Open access

Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning

Xiaoxue Cheng, Junyi Li, Zhenduo Zhang, Xinyu Tang, Xin Zhao, Xinyu Kong, Zhiqiang Zhang

neural information processing systems, 169570–169591 · Published 1 January 2025 · DOI 10.52202/085713-5108

Free PDFView at the publisherCite this paper

In brief

“Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning” is a 2025 conference paper by Xiaoxue Cheng and 6 co-authors, published in neural information processing systems. OpenAlex counts 19 citations as of 16 November 2026, which puts it in the top 2% of works published in 2025 by citations. It is open access: the full text is free to read at arXiv (Cornell University).

Citations
19
Most cited in
2026 (10)
Field-weighted impact
6.69
References
18
Authors
7

Abstract

Large reasoning models (LRMs) have demonstrated strong performance on complex reasoning tasks, but often suffer from overthinking, generating redundant content regardless of task difficulty. Inspired by the dual process theory in cognitive science, we propose Adaptive Cognition Policy Optimization (ACPO), a reinforcement learning framework that enables LRMs to achieve efficient reasoning through adaptive cognitive allocation and dynamic system switch. ACPO incorporates two key components: (1) introducing system-aware reasoning tokens to explicitly represent the thinking modes thereby making the model's cognitive process transparent, and (2) integrating online difficulty estimation and token length budget to guide adaptive system switch and reasoning during reinforcement learning. To this end, we propose a two-stage training strategy. The first stage begins with supervised fine-tuning to cold start the model, enabling it to generate reasoning paths with explicit thinking modes. In the second stage, we apply ACPO to further enhance adaptive system switch for difficulty-aware reasoning. Experimental results demonstrate that ACPO effectively reduces redundant reasoning while adaptively adjusting cognitive allocation based on task complexity, achieving efficient hybrid reasoning.

Cite this paper

Formatted from the catalogue record. Check names, capitals and page numbers against the paper before you submit.

APA 7

Cheng, X., Li, J., Zhang, Z., Tang, X., Zhao, X., Kong, X., & Zhang, Z. (2025). Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning. neural information processing systems, 169570–169591. https://doi.org/10.52202/085713-5108

MLA 9

Cheng, Xiaoxue, et al. “Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning.” neural information processing systems, 2025, pp. 169570–169591. https://doi.org/10.52202/085713-5108.

Chicago

Cheng, Xiaoxue, Junyi Li, Zhenduo Zhang, Xinyu Tang, Xin Zhao, Xinyu Kong, and Zhiqiang Zhang. 2025. “Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning.” neural information processing systems: 169570–169591. https://doi.org/10.52202/085713-5108.

Harvard

Cheng, X. et al. (2025) ‘Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning’, neural information processing systems, pp. 169570–169591. doi:10.52202/085713-5108.

IEEE

X. Cheng et al., “Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning,” neural information processing systems, pp. 169570–169591, 2025, doi: 10.52202/085713-5108.

BibTeX

@inproceedings{cheng2025incentivizing,
  title     = {{Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning}},
  author    = {Cheng, Xiaoxue and Li, Junyi and Zhang, Zhenduo and Tang, Xinyu and Zhao, Xin and Kong, Xinyu and Zhang, Zhiqiang},
  booktitle = {neural information processing systems},
  year      = {2025},
  pages     = {169570--169591},
  doi       = {10.52202/085713-5108},
  url       = {https://doi.org/10.52202/085713-5108}
}

Download for a reference manager: BibTeX (.bib) · RIS for Zotero, Mendeley and EndNote (.ris)

Topic and keywords

Physical Sciences › Computer Science › Artificial Intelligence › Topic Modeling

  • large language model reasoning
  • dual-process theory
  • overthinking
  • task difficulty estimation
  • hybrid reasoning

Work with this paper

Most-cited papers in Computer Science

All highly cited Computer Science papers

Questions about this paper

How do I cite this paper in APA style?

Cheng, X., Li, J., Zhang, Z., Tang, X., Zhao, X., Kong, X., & Zhang, Z. (2025). Incentivizing Dual Process Thinking for Efficient Large Language Model Reasoning. neural information processing systems, 169570–169591. https://doi.org/10.52202/085713-5108

How many times has this paper been cited?

19 times, according to OpenAlex on 16 November 2026. Google Scholar and Scopus count different sets of sources, so their numbers differ.

Is this paper free to read?

Yes. It is open access: read the free full text at arXiv (Cornell University).

Who wrote this paper?

Xiaoxue Cheng, Junyi Li, Zhenduo Zhang, Xinyu Tang, Xin Zhao, Xinyu Kong, Zhiqiang Zhang.

What is the DOI of this paper?

10.52202/085713-5108. The permanent link is https://doi.org/10.52202/085713-5108.

Bibliographic data from OpenAlex (CC0), last checked 16 November 2026. ResearchGuru does not host this paper; the links go to the publisher or the repository that does.