1007 papers
This collection explores the rapidly evolving intersection of statistics and artificial intelligence, where mathematical rigor meets machine learning innovation. These studies tackle how we can trust, interpret, and improve the algorithms that power modern technology, moving beyond black-box models to create systems that are both powerful and understandable. By bridging theoretical statistics with practical AI applications, researchers here are defining the next generation of intelligent tools.
Gist.Science processes every new preprint in this category directly from arXiv, ensuring you have immediate access to the latest breakthroughs before they appear in formal journals. For each paper, we provide both a detailed technical breakdown for experts and a clear, plain-language summary for anyone curious about the future of data and intelligence. Below are the latest papers in statistics and artificial intelligence, curated to help you stay ahead of the curve.
This paper introduces FinBench, a novel benchmark designed to evaluate the probabilistic calibration and uncertainty quality of agentic financial forecasting models by enforcing strict time-gating and utilizing proper scoring rules to penalize overconfidence and hallucinated certainty.
This paper proposes a two-phase latent Markov framework that decouples breast cancer risk from mammographic masking by modeling an underlying latent disease progression separate from observed density, thereby quantifying how irregular screening and masking effects can significantly underestimate risk in specific patient groups.
This paper proposes a distributionally robust covariance-steering framework that utilizes relative entropy constraints and computable upper bounds on risk-sensitive quantities to design stochastic guidance policies for nonlinear systems, ensuring the true state distribution remains close to a Gaussian surrogate while mitigating estimation errors caused by distribution ambiguity.
This paper proposes a distribution-free upper bound for the probability that a user encounters an undiscovered software defect, demonstrating that the fraction of defects reported exactly once during beta testing () serves as a conservative, model-independent estimate suitable for release decisions.
This paper demonstrates that while hierarchical Bayes (HB) credible intervals maintain near-nominal national coverage and significantly outperform classical direct estimators under stress-test scenarios involving unsampled strata, their domain-level coverage for Employment and Unemployment remains below nominal due to shrinkage bias that cannot be corrected by prior sensitivity or MSE adjustments.
This simulation study evaluates the performance of various missing data imputation methods for estimating average treatment effects under time-varying confounding, finding that multiple imputation generally outperforms other approaches in reducing bias and improving coverage, particularly when missingness mechanisms are complex.
The paper introduces PIONEER, a Bayesian joint modelling framework that integrates mechanistic tumour growth dynamics with multistate survival analysis to enable calibrated, uncertainty-quantified forecasting of clinical trial endpoints like PFS and OS using immature data and early longitudinal tumour measurements.
This study evaluates multi-season occupancy models with autocorrelation on heterogeneous datasets, finding that while they are robust to covariate overlap and uneven survey frequencies, they suffer from identifiability issues and produce biased predictions when spatiotemporal data gaps are present.
This paper derives explicit comparison bounds for multivariate max-stable distributions with unit--Fréchet margins, relating Kolmogorov and Wasserstein distances between these laws to metrics on their underlying de Haan representers, angular measures, and -functions, while also providing exact formulas and applications to Archimax and Brown-Resnick models.
This paper presents a methodology for developing aggregated combat models by using high-resolution discrete event simulations of a "meeting engagement" scenario to estimate attrition rates for a Markovian Lanchester process, while also evaluating various statistical estimation methods for this purpose.