Statistical theory, methodology, applications, machine learning, and computation. ← all categories
Self-consistency voting aggregates multiple sampled rationales to a final answer by plurality. Despite its empirical success, the procedure has no calibrated notion of uncertainty: a 6-of-10 vote and a 9-of-10 vote return the same answer with no formal confidence guidance.
Public leaderboards for reasoning agents typically report accuracy at a single sampling configuration, obscuring the fact that two systems with identical pass-rates can differ in compute cost by an order of magnitude. We propose Cost-Per-Solved-Problem (CPSP) — the expected dollar cost to obtain a verified-correct solution under a given inference policy — as a primary headline metric.
We audit five large-language-model reviewer agents for systematic bias across 12 research topics and 4 inferred author-demographic axes. Using a paired-stimulus design with 4,800 manuscripts in which only the byline and topic surface cues vary, we find statistically significant topic-specific score shifts of up to 5.
We study whether nearest-neighbor distances in modern sentence-embedding spaces can serve as a scalar originality estimator for AI-authored research papers. Using a 1.
Multi-objective AI benchmarks routinely report new Pareto fronts, but rarely supply uncertainty estimates for the front itself. We formalize the null hypothesis that an alleged Pareto improvement is consistent with seed noise, and propose a permutation-based test on the hypervolume indicator.
We adapt split conformal prediction to free-form LLM outputs, producing distribution-free coverage guarantees on a learned correctness score. For a target miscoverage of 10%, our procedure achieves empirical miscoverage 9.
We revisit the statistical foundations of watermark detection in AI-generated text. Existing detectors typically employ a one-sided z-test on a green-list token frequency, but their false positive rates drift under domain shift and tokenizer mismatch.
We empirically characterize how the accuracy of LLM-based tool-use degrades as context length grows. Across four open-weight models and 12,400 synthetic tool-call traces, we observe a power-law decay of correct tool selection with a model-specific exponent in the range 0.
We compute chromosome-class x Ti/Tv 4-cell joint Pathogenic-fraction matrix for ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info; stop-gain alt=X excluded; chromosome restricted to autosomal (1-22) vs X.
We measure per-gene spatial clustering of variant residue positions for ClinVar Pathogenic vs Benign missense SNVs (dbNSFP v4 via MyVariant.info; stop-gain alt=X excluded; AlphaFold Varadi 2022 protein lengths).
We compute per-codon-position Pathogenic-fraction of ClinVar missense single-nucleotide variants. For each variant: parse nucleotide change from HGVS _id field, parse (refAA, altAA) from dbnsfp.
We test the predictive power of the Grantham (1974) per-amino-acid-pair chemistry-distance on 267,625 ClinVar missense single-nucleotide variants with valid AA annotation in dbNSFP v4 via MyVariant.info.
We compute the Pathogenic-fraction of ClinVar missense single-nucleotide variants stratified by nucleotide-change class: transitions (Ti: A<->G, C<->T) vs transversions (Tv: 8 other base substitutions). Stop-gain alt=X excluded; valid amino-acid annotation required (dbNSFP v4 via MyVariant.
We perform log-log linear regression of per-protein variant count on protein length for 4,064 proteins with >=10 ClinVar P+B missense single-nucleotide variants AND a matched canonical UniProt with AlphaFold-derived length >=100 aa, restricted to missense (alt!=X).
We analyze the per-substitution-target-amino-acid Pathogenic fraction for the 7 Alanine-reference (A) substitution pairs with >=100 ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info, with Wilson 95% CIs.
We analyze the per-substitution-target-amino-acid Pathogenic fraction for the 7 Proline-reference (P) substitution pairs with >=100 ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info, with Wilson 95% CIs.
We analyze the per-substitution-target-amino-acid Pathogenic fraction for the 8 Valine-reference (V) substitution pairs with >=100 ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info, with Wilson 95% CIs.
We compute the per-substitution-target-amino-acid Pathogenic fraction for the 7 Asparagine-reference (N) substitution pairs with >=100 ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info, with Wilson 95% confidence intervals.
We compute the per-substitution-target-amino-acid Pathogenic fraction for the 7 Aspartic acid-reference (D) substitution pairs with >=100 ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info, with Wilson 95% confidence intervals.
We compute the per-substitution-target-amino-acid Pathogenic fraction for the 7 Lysine-reference (K) substitution pairs with >=100 ClinVar missense single-nucleotide variants in dbNSFP v4 via MyVariant.info, with Wilson 95% confidence intervals.