EUVIMEDEUVIMED entdecken
Crossref

Promoting Efficient Reasoning with Verifiable Stepwise Reward

Chuhuai Yue, Chengqi Dong, Yinan Gao, Hang He, Jiajun Chai, Wei Lin, Guojun Yin

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (41) · 34530-34538

DOI: 10.1609/aaai.v40i41.40752 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Chuhuai Yue und 6 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Large reasoning models (LRMs) have recently achieved significant progress in complex reasoning tasks, aided by reinforcement learning with verifiable rewards. However, LRMs often suffer from overthinking, expending excessive computation on simple problems and reducing efficiency. Existing efficient reasoning methods typically require accurate task assessment to preset token budgets or select reasoning modes, which limits their flexibility and reliability. In this work, we revisit the essence of overthinking and identify that encouraging effective steps while penalizing ineffective ones is key to its solution. To this end, we propose a novel rule-based verifiable stepwise reward mechanism (VSRM), which assigns rewards based on the performance of intermediate states in the reasoning trajectory. This approach is intuitive and naturally fits the step-by-step nature of reasoning tasks. We conduct extensive experiments on standard mathematical reasoning benchmarks, including AIME24 and AIME25, by integrating VSRM with PPO and Reinforce++. Results show that our method achieves substantial output length reduction while maintaining original reasoning performance, striking an optimal balance between efficiency and accuracy. Further analysis of overthinking frequency and pass@k score before and after training demonstrates that our approach indeed effectively suppresses ineffective steps and encourages effective reasoning, fundamentally alleviating the overthinking problem.

Zitieren

APA

Yue, C., Dong, C., Gao, Y., He, H., Chai, J., Lin, W., & Yin, G. (2026). Promoting Efficient Reasoning with Verifiable Stepwise Reward. Proceedings of the AAAI Conference on Artificial Intelligence, 40(41), 34530-34538. https://doi.org/10.1609/aaai.v40i41.40752

Vancouver

Yue C, Dong C, Gao Y, He H, Chai J, Lin W, et al. Promoting Efficient Reasoning with Verifiable Stepwise Reward. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(41):34530-34538. doi:10.1609/aaai.v40i41.40752

BibTeX
@article{Yue2026,
  author = {Yue, Chuhuai and Dong, Chengqi and Gao, Yinan and He, Hang and Chai, Jiajun and Lin, Wei and Yin, Guojun},
  title = {Promoting Efficient Reasoning with Verifiable Stepwise Reward},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {41},
  pages = {34530--34538},
  doi = {10.1609/aaai.v40i41.40752}
}
Kostenlos starten →