DyCon: Dynamic Reasoning Control via Evolving Difficulty Modeling

Tengyao Tu, Yulin Li, Hui-Ling Zhen, Libo Qin, Zhoujun Wei, Jinghua Piao, Zhuotao Tian, Yong Li, Min Zhang
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:122448-122494, 2026.

Abstract

Recent advances in Large Reasoning Models (LRMs) demonstrate remarkable performance improvements by iteratively reflecting, exploring, and executing complex tasks, yet suffer from inefficiencies due to redundant reasoning, known as “overthinking”. Existing methods to mitigate this issue either rely on static difficulty estimates or require task-specific training, and thus fail to adapt to the dynamic complexity during reasoning. In this work, we empirically show that the problem difficulty evolves dynamically throughout the reasoning process and is linearly encoded in the LRM’s step-level embeddings. Building on this insight, we propose DyCon, a training-free framework that leverages latent step-level representations to explicitly model the evolving task difficulty, enabling the dynamic control of reasoning depth to mitigate the overthinking issue. Extensive experiments conducted on four models ranging from 4B to 32B, and across twelve benchmarks in math reasoning, general question answering, and coding tasks demonstrate that DyCon significantly enhances reasoning efficiency by reducing redundant steps without sacrificing accuracy or generalization. Project page and code are available at https://github.com/yu-lin-li/DyCon.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-tu26c, title = {{D}y{C}on: Dynamic Reasoning Control via Evolving Difficulty Modeling}, author = {Tu, Tengyao and Li, Yulin and Zhen, Hui-Ling and Qin, Libo and Wei, Zhoujun and Piao, Jinghua and Tian, Zhuotao and Li, Yong and Zhang, Min}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {122448--122494}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/tu26c/tu26c.pdf}, url = {https://proceedings.mlr.press/v306/tu26c.html}, abstract = {Recent advances in Large Reasoning Models (LRMs) demonstrate remarkable performance improvements by iteratively reflecting, exploring, and executing complex tasks, yet suffer from inefficiencies due to redundant reasoning, known as “overthinking”. Existing methods to mitigate this issue either rely on static difficulty estimates or require task-specific training, and thus fail to adapt to the dynamic complexity during reasoning. In this work, we empirically show that the problem difficulty evolves dynamically throughout the reasoning process and is linearly encoded in the LRM’s step-level embeddings. Building on this insight, we propose DyCon, a training-free framework that leverages latent step-level representations to explicitly model the evolving task difficulty, enabling the dynamic control of reasoning depth to mitigate the overthinking issue. Extensive experiments conducted on four models ranging from 4B to 32B, and across twelve benchmarks in math reasoning, general question answering, and coding tasks demonstrate that DyCon significantly enhances reasoning efficiency by reducing redundant steps without sacrificing accuracy or generalization. Project page and code are available at https://github.com/yu-lin-li/DyCon.} }
Endnote
%0 Conference Paper %T DyCon: Dynamic Reasoning Control via Evolving Difficulty Modeling %A Tengyao Tu %A Yulin Li %A Hui-Ling Zhen %A Libo Qin %A Zhoujun Wei %A Jinghua Piao %A Zhuotao Tian %A Yong Li %A Min Zhang %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-tu26c %I PMLR %P 122448--122494 %U https://proceedings.mlr.press/v306/tu26c.html %V 306 %X Recent advances in Large Reasoning Models (LRMs) demonstrate remarkable performance improvements by iteratively reflecting, exploring, and executing complex tasks, yet suffer from inefficiencies due to redundant reasoning, known as “overthinking”. Existing methods to mitigate this issue either rely on static difficulty estimates or require task-specific training, and thus fail to adapt to the dynamic complexity during reasoning. In this work, we empirically show that the problem difficulty evolves dynamically throughout the reasoning process and is linearly encoded in the LRM’s step-level embeddings. Building on this insight, we propose DyCon, a training-free framework that leverages latent step-level representations to explicitly model the evolving task difficulty, enabling the dynamic control of reasoning depth to mitigate the overthinking issue. Extensive experiments conducted on four models ranging from 4B to 32B, and across twelve benchmarks in math reasoning, general question answering, and coding tasks demonstrate that DyCon significantly enhances reasoning efficiency by reducing redundant steps without sacrificing accuracy or generalization. Project page and code are available at https://github.com/yu-lin-li/DyCon.
APA
Tu, T., Li, Y., Zhen, H., Qin, L., Wei, Z., Piao, J., Tian, Z., Li, Y. & Zhang, M.. (2026). DyCon: Dynamic Reasoning Control via Evolving Difficulty Modeling. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:122448-122494 Available from https://proceedings.mlr.press/v306/tu26c.html.

Related Material