EUVIMEDEUVIMED entdecken
Crossref

OptScale: Probabilistic Optimality for Inference-time Scaling

Youkang Wang, Jian Wang, Rubing Chen, Xiao-Yong Wei

Proceedings of the AAAI Conference on Artificial Intelligence · 14.03.2026 · 40 (40) · 33710-33718

DOI: 10.1609/aaai.v40i40.40661 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Proceedings of the AAAI Conference on Artificial Intelligence, online erschienen am 14.03.2026. Verfasst von Youkang Wang und 3 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Inference-time scaling has emerged as a powerful technique for enhancing the reasoning performance of Large Language Models (LLMs). However, existing approaches often rely on heuristic strategies for parallel sampling, lacking a principled foundation. To address this gap, we propose a probabilistic framework that formalizes the optimality of inference-time scaling under the assumption that parallel samples are independently and identically distributed (i.i.d.), and where the Best-of-N selection strategy follows a probability distribution that can be estimated. Within this framework, we derive a theoretical lower bound on the required number of samples to achieve a target performance level, providing the first principled guidance for compute-efficient scaling. Leveraging this insight, we develop OptScale, a practical algorithm that dynamically determines the optimal number of sampled responses. OptScale employs a language model-based predictor to estimate probabilistic prior parameters, enabling the decision of the minimal number of samples needed that satisfy predefined performance thresholds and confidence levels. Extensive experiments on representative reasoning benchmarks (including MATH-500, GSM8K, AIME, and AMC) demonstrate that OptScale significantly reduces sampling overhead while remaining better or on par with state-of-the-art reasoning performance. Our work offers both a theoretical foundation and a practical solution for principled inference-time scaling, addressing a critical gap in the efficient deployment of LLMs for complex reasoning.

Zitieren

APA

Wang, Y., Wang, J., Chen, R., & Wei, X. (2026). OptScale: Probabilistic Optimality for Inference-time Scaling. Proceedings of the AAAI Conference on Artificial Intelligence, 40(40), 33710-33718. https://doi.org/10.1609/aaai.v40i40.40661

Vancouver

Wang Y, Wang J, Chen R, Wei X. OptScale: Probabilistic Optimality for Inference-time Scaling. Proceedings of the AAAI Conference on Artificial Intelligence. 2026;40(40):33710-33718. doi:10.1609/aaai.v40i40.40661

BibTeX
@article{Wang2026,
  author = {Wang, Youkang and Wang, Jian and Chen, Rubing and Wei, Xiao-Yong},
  title = {OptScale: Probabilistic Optimality for Inference-time Scaling},
  journal = {Proceedings of the AAAI Conference on Artificial Intelligence},
  year = {2026},
  volume = {40},
  number = {40},
  pages = {33710--33718},
  doi = {10.1609/aaai.v40i40.40661}
}
Kostenlos starten →