Evalci, a pure-Python library that turns a per-item results table into a publication-ready claim, and re-analyzes a public comparison of nine language models'MMLU accuracy to find that 3 of the 8 adjacent leaderboard-rank gaps are not statistically significant after correcting for the 36 pairwise comparisons the ranking implies.
Abstract
The dominant practice in language model evaluation is to report a single accuracy number per model and declare the higher one better, without testing whether the gap could plausibly be sampling noise. On benchmarks of a few thousand items, and under temperature sampling where a model can differ from itself run to run by more than the reported gap between models, this practice routinely overstates confidence in headline claims. The statistical machinery to fix this -- confidence intervals, paired significance tests, power analysis, clustered standard errors, multiple-comparison correction -- is well established, but no standard, pip-installable tool packages it in the shape an evaluation actually takes: a per-item results table. We present evalci, a pure-Python library (numpy/scipy/pandas only) that turns a per-item results table into a publication-ready claim -- e.g.,"Model A beats Model B, $\Delta=3.1$ pts, 95% CI [1.2, 5.0], paired permutation $p=0.002$, $n=1{,}319$"-- in one function call, with adapters for lm-evaluation-harness and HELM output. Every routine is validated against an independent reference (statsmodels, or brute-force exact enumeration) rather than only against itself. As a case study, we re-analyze a public comparison of nine language models'MMLU accuracy and find that 3 of the 8 adjacent leaderboard-rank gaps are not statistically significant after correcting for the 36 pairwise comparisons the ranking implies. evalci is available at https://pypi.org/project/evalci/ (source: https://github.com/Shreyaskc/evalci, DOI: https://doi.org/10.5281/zenodo.21201815)
Thousands of R packages hold statistical methods with no native Python equivalent. Runtime bridges require an R installation; hand-written ports do not scale. Translation fails silently where the languages diverge, as in transform normalization, integer width, and argument evaluation. We present r2py, a framework that converts an R package into a native Python library using orchestrated language-model agents under human supervision, with correctness established by numerical comparison against the original at declared tolerances. The compiled code is retained unmodified, so any divergence lies in the translation. Seven phases decompose the work for independent invocations: structural analysis fixes conversion order, every base-R construct's rendering is settled in reviewable guides before code generation, and four verification methods each expose defects their predecessors miss. Packages reaching compiled code through .Call() add a five-phase prologue reconstructing the R C API they use. Conversions of KernSmooth and rpart reproduce R across 518 and 846 tests.
LangChoiceBench is introduced, a project-level code-generation benchmark for measuring Python preference, recommendation-implementation consistency, and language diversity, and it is found that Python remains heavily over-selected, recommendation-implementation consistency is low, and smaller open-weight models generally show stronger Python preference and lower language diversity.
Lukas Twist, Twm Stone, Helen Yannakoudakis et al.· 0 citations
Under cross-benchmark transfer evaluation across twelve models completed by all four readouts, difference-in-means attains the highest model-level mean on ten models, while logistic regression wins the most capability-model cells.
Yanshi Li, Xue Bai, Shuman Liu et al.· 0 citations
Empirical comparisons between systems are a standard form of evidence in computer science research, but few are checked for statistical validity: most are never framed as statistical tests at all. Existing multiple-comparison procedures could control the resulting error, but need inputs (what an analysis examined, and how its observations are arranged) that are not recoverable from a list of p-values. We introduce Tacet, a language in which an analysis declares what it generated, states what it expects to find, and is refused any claim it cannot afford or cannot properly test. Its core calculus T pairs a free estimation sublanguage, carrying a reported footprint and a purity bit that records whether any outcome was consulted in building a value, with a priced claim sublanguage, carrying a wealth transformer, connected only by a mechanism that prices a comparison. A sample selected by reading outcomes sets the purity bit and is recorded as having examined everything it read, permanently, so it can never be granted a one-sided or confirmatory price, without the system ever asking whether the analyst intended to cherry-pick. Whether a comparison is paired or clustered is computed statically from the artifact schema, from declared functional dependencies between key fields alone and before any data is read, and a mechanism that assumes that structure away is refused rather than priced. Because the wealth transformer is antitone in the realized p-value, affordability can be checked before the analysis runs too, turning pre-registration into a typing rule. We prove the metatheory machine-checked in Lean 4 with no admitted gaps, and demonstrate the approach on a reference implementation and two case studies on published artifacts, the SWE-bench Verified leaderboard and BIG-Bench Hard.
An automated, multi-dimensional evaluation framework for C# code generation, applying it to four state-of-the-art LLMs: GPT, Gemini, Claude, and Grok is presented and a substantial gap between correctness and quality attributes is revealed.
Seyed Mohammad Mahdi Ghalandarian, Majid Bazargani, Masoumeh Taromirad· 0 citations