@inproceedings{hariri-etal-2026-ranking,
    title = "Ranking Reasoning {LLM}s under Test-Time Scaling",
    author = "Hariri, Mohsen  and
      Hinczewski, Michael  and
      Ma, Jing  and
      Chaudhary, Vipin",
    editor = "Liakata, Maria  and
      Moreira, Viviane P.  and
      Zhang, Jiajun  and
      Jurgens, David",
    booktitle = "Proceedings of the 64th Annual Meeting of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2026",
    address = "San Diego, California, United States",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/ingest-acl/2026.acl-long.1544/",
    pages = "33437--33478",
    ISBN = "979-8-89176-390-6",
    abstract = "Test-time scaling evaluates reasoning LLMs by sampling multiple outputs per prompt, but ranking models in this regime remains underexplored. We formalize dense benchmark ranking under test-time scaling and introduce Scorio, a library that implements statistical ranking methods such as paired-comparison models, item response theory (IRT) models, voting rules, and graph- and spectral-based methods. Across 20 reasoning models on four Olympiad-style math benchmarks (AIME{'}24, AIME{'}25, HMMT{'}25, and BrUMO{'}25; up to N = 80 trials), most full-trial rankings agree closely with the Bayesian gold standard Bayes{\_}{\ensuremath{\mathscr{U}}}@80 (mean Kendall{'}s {\ensuremath{\tau}}{\_}b = 0.93{--}0.95), and 19{--}34 methods recover exactly the same ordering. In the single-trial regime, the best methods reach {\ensuremath{\tau}}{\_}b {\ensuremath{\approx}} 0.86.Using greedy decoding as an empirical prior (Bayes{\_}R₀@N) reduces variance at N = 1 by 16{--}52{\%}, but can bias rankings when greedy and stochastic sampling disagree. These results identify reliable ranking methods for both high- and low-budget test-time scaling. We release Scorio as an open-source library at https://github.com/mohsenhariri/scorio."
}