English

Beyond Pass@k: Breadth-Depth Metrics for Reasoning Boundaries

Artificial Intelligence 2025-10-22 v2 Computation and Language Machine Learning

Abstract

Reinforcement Learning with Verifiable Rewards (RLVR) has emerged as a powerful paradigm to improve Large Language Models on reasoning tasks such as coding, math or logic. To assess the reasoning boundary (the fraction of problems a model can solve) researchers often report Pass@k at large sampling budgets. Recent results reveal a crossover phenomenon: while RLVR models outperform the base model at small k values, the base model usually outperforms them when sampling a very large number of completions. This has been interpreted as evidence that base models have a larger reasoning boundary. We argue that on tasks with discrete answer spaces, such as math with numeric outputs, Pass@k at large k reflects the increasingly higher chance of success in the limit of the number of trials rather than genuine reasoning, and can therefore be misleading. We propose Cover@tau, which measures the fraction of problems that a model can solve for which at least a tau proportion of completions are correct. Unlike Pass@k, Cover@tau captures reasoning under an explicit reliability threshold: models that rely on random guessing degrade rapidly as tau increases. We evaluate several RLVR models using Cover@tau-based metrics and illustrate how the relative rankings of popular algorithms change compared to Pass@1, offering a different perspective on reasoning boundaries.

Keywords

Cite

@article{arxiv.2510.08325,
  title  = {Beyond Pass@k: Breadth-Depth Metrics for Reasoning Boundaries},
  author = {Marius Dragoi and Ioana Pintilie and Florin Gogianu and Florin Brad},
  journal= {arXiv preprint arXiv:2510.08325},
  year   = {2025}
}

Comments

10 pages, 3 figures. v2 adds discussion of related work (G-Pass@k)

R2 v1 2026-07-01T06:27:01.800Z