@inproceedings{zhang-2025-confidence,
    title = "Confidence-Aware Reasoning: Optimizing Self-Guided Thinking Trajectories in Large Reasoning Models",
    author = "Zhang, Jiaxin",
    editor = "Potdar, Saloni  and
      Rojas-Barahona, Lina  and
      Montella, Sebastien",
    booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track",
    month = nov,
    year = "2025",
    address = "Suzhou (China)",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/author-page-yu-wang-polytechnic/2025.emnlp-industry.146/",
    doi = "10.18653/v1/2025.emnlp-industry.146",
    pages = "2081--2095",
    ISBN = "979-8-89176-333-3",
    abstract = "Chain-of-thought enables large reasoning models (LRMs) to reason through multi-step problems but often leads to unnecessarily long or redundant reasoning traces, a phenomenon known as \textit{overthinking}. This results in inflated inference costs and potential degradation in answer quality. To address these challenges, we propose Confidence-Aware Reasoning (), an inference-time framework that optimizes reasoning trajectories by selectively pruning low-utility reasoning blocks and halting early when sufficient confidence has been achieved. is theoretically grounded in Bayesian optimal experimental design, treating each reasoning block as a sequential decision whose utility is approximated by its marginal contribution to reducing final answer uncertainty. We introduce a lightweight implementation that leverages token-level confidence to dynamically modulate reasoning depth without additional supervision. Evaluations on multiple benchmarks, including AMC, AIME, GPQA-Diamond, and MATH-500 show that improves answer accuracy by up to +13.3{\%}, while reducing average reasoning length by 40{\%}{--}50{\%}. Our findings demonstrate that information-theoretic insights can effectively control self-guided reasoning and enable LRMs to ``think just enough'' at test time."
}