Papers

Conformal prediction inference yields intervals for out-of-sample random outcomes with designated coverage probabilities, given exchangeable or independent-and-identically distributed random variables. For regression analysis, valid coverage can be achieved given finite sample sizes, despite unavoidable misspecifications of the regression function. We propose a novel method of conformal inference, aimed to produce confidence intervals of the unknown expectations of the in-sample outcomes with the designated coverage probabilities conditional on the realised sample. These conformal confidence intervals fill a gap between classical regression and conformal inference. The proposed approach is applied to small area estimation problems.

Comparing stochastically trained models requires estimating both a performance difference and its uncertainty from repeated runs. We study whether training logs from those same runs can make such comparisons more precise. Because training-log covariates are produced during training rather than measured before it, we use arm-specific covariate adjustment: each model is adjusted only with statistics from its own runs, and the raw mean difference remains the reported effect. In a vision study spanning three architectures and three datasets, simple adjustments based on early training logs often reduce uncertainty in model comparisons. The main limitation is covariate selection. Broadly searching the log pool for the most correlated statistic often adds more noise than it removes, even when useful statistics exist in hindsight. Training logs therefore appear useful for more precise model comparisons, but only when the adjustment avoids large selection noise.

Correlation matrices arise when marginal scales are removed from covariance matrices, yet a normalized likelihood must account for both quotient distance and quotient volume. We propose a Riemannian Gaussian model for full-rank correlation matrices under quotient-affine geometry. The distribution is proper and has finite radial moments. We derive exact score and profiled-scale equations and recover Fisher-transformed Gaussian inference for two-dimensional matrices. In higher dimension, a curvature calculation shows that the normalizing constant can vary with the center. Exact maximum likelihood and Fréchet estimation may therefore have different population targets. We develop chart-based methods for evaluating the normalizer, fitting the likelihood, and sampling. Numerical studies verify the analytic case and compare integration, estimation, and sampling procedures across dimensions and dispersion regimes. A rolling-finance application and a controlled prior study illustrate both the value and computational cost of the model. The method is most reliable in small to moderate dimensions, while proposal efficiency and numerical conditioning deteriorate near the boundary and at larger dispersion.

An overall treatment effect in a clinical trial may inadequately represent particular patient subgroups, creating uncertainty about whether a population-level efficacy conclusion can legitimately be transferred to them. Conventional interaction tests only investigate whether subgroup-specific treatment effects are exactly equal or not, but cannot detect whether the differences are small enough to be clinically negligible. We develop a formal framework for assessing subgroup treatment effect consistency for time-to-event outcomes within the Cox proportional hazards model. Consistency is formulated as an equivalence problem based on the (weighted) treatment-by-subgroup interaction coefficient. We consider detecting consistency between two complementary subgroups and consistency of subgroup-specific treatment effects with the overall treatment effect. For each setting, we develop a conventional two one-sided tests procedure (TOST) and a new test motivated by optimal equivalence testing for normally distributed parameters. We prove the asymptotic validity of all procedures and show empirically that the new tests are more powerful than their TOST counterparts. Finally, we apply the new methodology to a case study motivated by the CANTOS cardiovascular outcomes trial.

Stochastic learning objectives are typically written as expectations over abstract random variables. Actual training, however, uses concrete random inputs that enter both the realized loss and its gradient. Structure in these inputs that is accessible to the learning system can therefore be learned and exploited. Much prior structured-noise work asks how noise should be distributed or designed; we instead ask what structure in the concrete realized randomness becomes exploitable by the learner. We develop this general view and analyze its mechanism in diffusion models: in noise prediction, clean data and realized noise jointly form the noisy input, so the model can improve prediction by learning clean-data regularities or exploiting structure introduced through the noise, and the two routes can interact. Using pseudorandom streams as controlled, reproducible instances of structured noise, we provide mechanistic evidence on MNIST and CIFAR-10: random-role ablations localize the dominant effect to diffusion noise; in a diffusion probe, structured-noise training can reduce prediction loss below the IID reference, but replacing the test noise with IID reverses this advantage; and shuffling the same values largely removes the source-dependent loss reduction. This learned dependence can also affect generation. The same view offers a unified interpretation of data-dependent noise assignment, noise-based backdoors, and temporally correlated noise in video diffusion: although these methods introduce different structures, all alter what the model can exploit through noise and its interaction with clean-data learning. Our results indicate that diffusion noise is not merely a passive stochastic perturbation, but a learnable—and therefore potentially designable—input dimension.

Boolean matrix factorization provides an interpretable framework for discovering latent binary patterns in high-dimensional data, yet existing methods typically analyze a single binary matrix or factorize multiple matrices independently, failing to exploit shared latent structure across related datasets. We propose Joint Bayesian Boolean Matrix Factorization (JBBMF), a model that simultaneously factorizes two related binary matrices through a shared latent Boolean pattern matrix and dataset-specific loading matrices. To capture dependence between paired datasets, we introduce a conditional prior linking the loading matrices, allowing latent factors to persist or change across conditions while preserving a common interpretable representation. The model combines Boolean matrix factorization with a Bernoulli observation model and conjugate priors, yielding closed-form full conditional distributions and an efficient Gibbs sampler for posterior inference,uncertainty quantification for latent factors, reconstructed matrices, and noise parameters. Simulation studies demonstrate that jointly modeling related binary datasets substantially improves recovery of shared latent factors compared with independently applying standard Boolean matrix factorization to each dataset, while maintaining high reconstruction accuracy. We apply JBBMF to paired chromosomal copy number alteration profiles from multiple myeloma patients collected at diagnosis and relapse. The analysis identifies recurrent chromosomal alteration signatures shared between disease stages and quantifies the uncertainty of these findings. JBBMF offers a flexible and interpretable Bayesian model for the joint analysis of related binary datasets in genomics and other application domains.

We present a simple Gaussian approximation to the finite-sample distribution of the classical ridge regression estimator. Our approximation captures the fact that, in finite samples, the ridge regression estimator trades off bias and variance to reduce estimation and prediction error. Our approximation is based on nonstandard asymptotics where $i)$ we let the estimator's regularization parameter grow proportionally to the sample size; and $ii)$ we treat the population regression coefficients as local to the reference vector that defines the estimator's direction of shrinkage. In contrast to other asymptotic approximations in the literature, we allow for general forms of heteroskedasticity and autocorrelation in the data generating process (at the cost of considering a low-dimensional model where the number of covariates is not allowed to grow with the sample size). We use our simple Gaussian approximation to propose two new strategies to select the regularization parameter for the ridge regression estimator. The suggested strategies select the regularization parameter to minimize either average or worst-case excess prediction risk, where risk is computed using our suggested Gaussian approximation.

This paper is concerned with one-bit mean estimation, where each independent sample is represented by a single binary message. We consider distributions on $\mathbb{R}$ with mean in $[-\lambda,\lambda]$ and absolute $k$-th central moment at most $\sigma^k$, where $k>1$ is fixed. For this class, previous work attained the optimal sample complexity for general queries using a two-stage protocol. The first stage localizes the mean. The second-stage queries are chosen after localization and refine the estimate around the decoded center. We show that this interaction can be avoided by constructing a randomized fully non-adaptive protocol that fixes all queries before observing the data and matches the optimal adaptive sample complexity. For target accuracy $\epsilon$ and confidence $1-\delta$, its sample complexity scales as \[ \log\frac{\lambda}{\sigma} + \begin{cases} (\sigma/\epsilon)^2\log(1/\delta), & k>2,\\ (\sigma/\epsilon)^2\log(\sigma/\epsilon)\log(1/\delta), & k=2,\\ (\sigma/\epsilon)^{k/(k-1)}\log(1/\delta), & 1<k<2, \end{cases} \] up to constants depending only on $k$. In the range covered by the known lower bound, this rate is minimax optimal even among fully adaptive protocols. This gives a negative answer to the COLT 2026 open problem asking whether interaction is necessary for order-optimal one-bit mean estimation with general queries \citep[Open Problem~1]{lau2026open}.

In applied research, the decision to utilize multilevel modeling (MLM) is commonly guided by comparing a point estimate of the unconditional intraclass correlation coefficient (ICC) against some recommended threshold (e.g., 0.05). This naive approach, however, fails to account for sampling uncertainty and provides no formal inferential justification for a decision. The current work introduces negligible effect significance testing (NEST; or equivalence testing) for the ICC, proposing a framework that performs this test using a pivotal quantity of the ICC based on the F statistic. Furthermore, we offer guidance to help researchers define "negligible" that utilizes the design effect to characterize a tolerable level of variance inflation. R code illustrating the proposed procedure is provided.

We characterize the finite sample behavior of the log-likelihood ratio statistic in binary logistic regression, uniformly over both the design and the target parameter. For $n\geq d\geq 3$, we determine, up to universal constants, its worst case $(1-\delta)$ quantile over all fixed collections of design vectors and all target parameters: \[ d\log\left(\frac{e n}{d}\right)+\log\left(\frac{1}{\delta}\right). \] This is a nonasymptotic analogue of the Wilks $\chi^2_d$ phenomenon and requires no regularity assumptions on the design. The low dimensional cases exhibit unusual behavior. The worst case quantile in dimension $d=2$ is sharply of order \[ \log\log\log n+\log\left(\frac{1}{\delta}\right). \] The worst case quantile in dimension $d=1$ is of order $\log(1/\delta)$, with no dependence on $n$. Finally, i.i.d. Gaussian design vectors recover the classical Wilks scale. In the regime $n\gtrsim d+\log(1/\delta)$, we prove the sharp bound \[ d+\log\left(\frac{1}{\delta}\right). \] Unlike existing asymptotic results, our bounds are uniform over the target parameter, which may depend on $n$, $d$, and $\delta$.

Recently, rectified flow has emerged as a fundamental framework for large-scale image generation, powering state-of-the-art systems such as FLUX.1 and Stable Diffusion 3. Despite its remarkable empirical success, the computational and statistical guarantees of iterative rectified flow have remained largely unexplored. We address this problem by studying c-rectified flow, a cost-aware class of rectified flow that projects velocity fields onto a gradient class while preserving endpoint marginals. The ordinary rectified flow can fail to recover the optimal transport coupling: in a Gaussian case study, the iteration converges to the optimal coupling if and only if the source and target covariance matrices commute. In contrast, under suitable compactness and uniform-integrability assumptions, iterative c-rectified flow always converges to the optimal transport coupling. We further establish quantitative one-step contraction and exponential convergence guarantees under projection-stability assumptions for both quadratic and strongly convex displacement costs. Finally, under a Hölder ball assumption, we develop new minimax-optimal score estimation rates and show that, when combined with iterative c-rectified flow, they yield a rate-optimal estimator of the optimal transport for the dimension \(d \ge 3\) and a nearly parametric rate for \(d=1,2\).

Two active learning algorithms for hyperspectral image (HSI) classification are proposed that combine density-aware Fermat distances with Poisson-reweighted harmonic label propagation. Our methods actively query points using an uncertainty-based acquisition function, extending Poisson ReWeighted Laplace Learning (PWLL). Our first algorithm, Fermat Active Laplace Learning (FALL), builds an affinity matrix using Fermat distances between all data points. Then, PWLL is run with a diagonal perturbation using the minimum-norm acquisition function. In contrast, Approximate FALL (A-FALL) computes Fermat distances between each data point and landmark pixels selected via farthest-point sampling and constructs the affinity matrix using landmark multidimensional scaling. After several query rounds, A-FALL selects the Fermat exponent $p$ using a leave-one-out cross-validation variant. FALL and A-FALL leverage Fermat distances and subsequent harmonic label propagation to provide a density-aware estimation of the data manifold, improving labeling accuracy. Experiments on Salinas A and Pavia show the effectiveness of FALL and the scalability of A-FALL to large HSI scenes.

With AI systems gaining more access to individuals' information, it is important to protect privacy when reporting statistical answers. Equally important is to privatize the reporting of uncertainty in such answers. To this end, we adopt a Bayesian likelihood-free framework and make simulation from the posterior private. In particular, we propose a new private instantiation of the Bayesian bootstrap using a blocking strategy. Rather than assigning idiosyncratic random weights to each individual, we randomly group individuals and assign a single weight to each group. By concealing individuals' contributions within a group, we fortify differential privacy gates. We harness amortized inference that decouples private learning from posterior sampling. A push-forward map from observation weights to posterior samples is learned privately by adding calibrated noise during training. Subsequent posterior draws require no additional privacy and computation budget. We call the resulting method the Private Generative Bayesian Bootstrap (PGBB). We establish a differential privacy guarantee, analyze convergence to the non-private blocked-bootstrap target, and quantify the discrepancy between the ordinary and blocked Bayesian-bootstrap posteriors. In addition, we derive data-free tuning of the block Dirichlet concentration parameter that restores posterior dispersion asymptotically. We also show a single fit of PGBB can support a family of loss-based decision rules simultaneously without additional privacy cost. In simulations and in applications to U.S. Census returns to schooling and U.S. natality birthweight quantiles, PGBB gives competitive private uncertainty quantification and improves over private Bayesian alternatives that require a specified data-generating model in common settings.

Very often, accuracy of analysis and forecasting (multiple coefficient of regression and residual means) obtained for a sample used to formulate a regression model is not equal to the accuracy achieved for another homogeneous sample. Indeed, accuracy of analysis and forecasting based on another sample is much worse. This is explained by the discrepancy between a postulated and a real model. This discrepancy is caused by the redundancy of the number of terms of an approximating series. It describes a phenomenon under study with the sample noise. To filter this noise it is necessary to correctly choose the terms of an approximating series and to determine their number, which in turn depends on sample size. It is possible to include in this approximating series not only independent variables but also various functions of the value of the independent variables (e.g., squares, cubes, algorithms, etc.). This paper gives the procedure for selection of an approximating series.

Human-centered assessment tasks, which are essential for systematic decision-making, rely heavily on human judgment and typically lack verifiable ground truth. Existing approaches face a dilemma: methods using only human judgments suffer from heterogeneous expertise and inconsistent rating scales, while methods using only model-generated scores must learn from imperfect proxies or incomplete features. We propose Aggregate-then-Calibrate (AtC), a two-stage framework that combines these complementary sources. Stage-1 aggregates heterogeneous comparative judgments into a consensus ranking using a rank-aggregation model that accounts for annotator reliability. Stage-2 calibrates any predictive model's scores by an isotonic projection onto the order, enforcing ordinal consistency while preserving as much of the model's quantitative information as possible. Theoretically, we show: (1) modeling annotator heterogeneity yields strictly more efficient consensus estimation than homogeneity; (2) isotonic calibration enjoys risk bounds even when the consensus ranking is misspecified; and (3) AtC asymptotically outperforms model-only assessment. Across semi-synthetic and real-world datasets, AtC consistently improves accuracy and robustness over human-only or model-only assessments. Our results bridge judgment aggregation with model-free calibration, providing a principled recipe for human-centered assessment when ground truth is costly, scarce, or unverifiable.

We provide new estimates in Wasserstein distance for the asymptotic bias of the unadjusted Langevin algorithm, in the classical setting of log-smooth strongly log-concave measures. Our bound implies a Wasserstein mixing time of order $\kappa \sqrt{d}/\varepsilon$, where $\kappa$ is the condition number, $d$ is the dimension, and $\varepsilon$ is the target precision: this improves by a factor of $\sqrt{d}/\varepsilon$ over the previous state-of-the-art results.

We introduce a unified framework for the error analysis of generative models based on the entropy production rate of the forward-reverse diffusion process pair. For a pair of continuity equation flows, the rate admits a closed velocity form identity whose time integral decomposes the terminal Kullback–Leibler (KL) divergence into the sum of an initialization error, a score approximation error, and a time-discretization error. By analyzing the entropy production at the level of marginal distributions, rather than in path space, our framework yields a sharp convergence rate of $\mathcal{O}(h^2)$ for the Euler-Maruyama sampler, where $h$ is the step size. This improves upon the $\mathcal{O}(h)$ rates typically obtained from Girsanov's path-space analyses. Furthermore, our framework unifies the analysis of score-based SDEs, probability-flow ODEs, and stochastic interpolants by varying diffusion coefficients within a single inequality, revealing the trade-off between deterministic and stochastic sampling. Numerical experiments confirm the predicted scaling with step size and terminal time.

A/B testing remains the standard for rolling out new features in the technology industry. Each experiment, however, consumes real traffic, engineering effort, and weeks of wall-clock time. Can AI agents—conditioned on behavioral profiles and contextual descriptions of the intervention—simulate outcomes accurately enough to vet candidate treatments before committing live traffic? We formalize this question as a Simulated Randomized Controlled Trial (S-RCT) and derive a two-layer error decomposition that separates agent approximation error from subsampling error, enabling targeted improvements to each. The framework is agent-agnostic: any behavioral model—from a fine-tuned specialist to a general-purpose foundation model—can serve as the simulation engine. Validated on 67 historical marketing A/B tests, a baseline S-RCT using an off-the-shelf foundation model captures directional signal (sign overlap 0.70) but systematically overshoots effect magnitudes. A two-phase pre-period calibration protocol reduces the squared prediction error (after removing irreducible measurement noise) by ${\sim}77\times$; a within-subject design—where each agent is exposed to both arms—reduces standard errors by ${\sim}2.4\times$. We discuss limitations of the current approach and identify applications where experimenters stand to benefit from agentic signals.

Creator economy platforms face a strategic dilemma: allocating traffic to established stars for immediate ad revenue versus nurturing emerging creators to build a follower base for future monetization. We develop a continuous-time dynamic optimization model to characterize the optimal traffic allocation policy for a platform managing heterogeneous creators with dual revenue streams (advertising and direct follower contributions). We characterize the optimal policy analytically, revealing a “most-valuable-creator-first" rule driven by a forward-looking activation set. Under Bass diffusion dynamics, this policy exhibits a sophisticated “conditional reversal" strategy, where the platform temporarily prioritizes lagging creators to capitalize on word-of-mouth effects. Regarding the ecosystem structure, we find the optimal policy acts as a selective gatekeeper. Unlike myopic policies that lead to a harsh “winner-take-all" market, or naive fairness-driven heuristics that foster inefficient “indiscriminate growth," the optimal policy imposes a strict capability threshold that screens which creators receive platform traffic. Furthermore, it enforces disciplined growth for successful entrants, capping their follower bases at an optimal ceiling to prevent over-investment. Finally, we demonstrate that simple heuristics can lead to significant revenue losses (up to 25%) and propose a practical “follower-growth adjusted" heuristic that achieves near-optimal performance by leveraging observed growth momentum.

Distinguishing long memory behaviour from nonstationarity can be very difficult as in both cases the sample autocovariance function decays very slowly. Available stationarity tests either do not include long memory or fare poorly in terms of empirical size, especially near the boundary between long memory and nonstationarity. We propose a testing procedure based on evaluating periodograms at different epochs. Limiting distributions established here are easily tractable as sum of weighted independent $\chi^2$ random variables. Moreover, numerical studies are provided to show that the proposed approach seems to outperform existing methods.

Mediation analysis investigates whether part of the treatment effect is channelled through one or more mediators along the causal pathway between the treatment and the primary outcome. However, the presence of multiple, potentially dependent mediators raises substantial challenges, particularly when the outcome is binary, the mediators are measured on different scales, and interactions are present. We consider on causal mediation analysis with a binary treatment and a binary outcome, defining natural direct, indirect, and total effects on the relative-risk scale, thereby avoiding the interpretational difficulties associated with non-collapsible effect measures. Under a sequential ignorability assumption, we develop a unified regression-based framework that accommodates multiple continuous, binary, or mixed mediators. The proposed framework accounts for dependence among mediators and allows for both exposure-mediator and mediator-mediator interactions. We derive closed-form expressions for the causal effects across the different mediator settings and develop a likelihood-based inference procedure for estimating the causal effects and quantifying their uncertainty. The methodology is illustrated through two empirical applications.

Family-based study designs allow the investigation of gene mutation effects on a disease risk by considering related family members. Some methods have been developed for testing sets of genetic variants in family studies but only very few can handle right-censored time-to-event data. We propose here a correlated frailty model for the analysis of a survival outcome related to cancer in presence of familial correlations. These familial correlations are explained by a residual familial component specified by a kinship matrix and a region- or gene-based specific correlation structure modeled via identical-by-descent (IBD) probability matrix. The proposed approach is used to quantify and evaluate the association between a set of common single nucleotide polymorphism (SNPs) or rare variants (or both) from the same genomic region and a survival outcome, e.g. time to disease onset. The model's marginal likelihood is maximized using the Marquardt algorithm. We evaluated the method by simulations under various scenarios where we varied the family size, the strength of genetic associations from multiple rare variants and the presence or not of residual familial correlation. The results indicate that the correlated frailty model can be valuable in family cancer studies, for example to identify genomic regions significantly associated with the time to cancer onset.

ALL-IN meta-analysis was developed and first applied during the COVID-19 pandemic. While this setting inspired its name, ALL-IN can also benefit non-pandemic circumstances. Conventional meta-analysis loses its coverage when updated repeatedly over time and when the decisions to initiate new trials and synthesize them depend on the results within the meta-analysis (accumulation bias). ALL-IN meta-analysis is anytime-valid. In its simplest form, ALL-IN meta-analysis stays familiar to run and read based on forest plots with confidence intervals that are wider than standard ones. Within collaborative prospective meta-analysis, the payoff is flexibility and speed, with fast sharing of individual participant data or harmonized aggregate data. Outside of this setting, the payoff is validity, when the decisions that shape the evidence (when to stop trials and whether new ones start) are typically outside the control of the meta-analyst. On the one hand, ALL-IN meta-analysis becomes inefficient when there is a maximum sample size or a stopping rule that the meta-analyist can control. On the other hand, it enables adaptations for any evidence synthesis to halfway become living, prospective or even real-time on interim trial results, without complicating the statistics.

When the target of inference is a group mean rather than an individual score, long assessments can be statistically inefficient. From the marginal score of the likelihood, we derive the asymptotic standard error of the marginal maximum likelihood estimator of a latent IRT mean under random item sampling, and obtain a closed-form approximation via a Gaussian-convolution treatment of the item information. The formula makes the test-length–sample-size trade-off explicit and includes an inflation factor for jointly estimating the population variance when the item pool is mistargeted. We validate it against exact benchmarks, Monte Carlo simulation under pool–population mismatch, and a NAEP mathematics resampling study; group means remain approximately unbiased even with single-item forms.

Given an algebraic statistical model, a challenging problem is classifying the data according to the number of positive critical points of the likelihood function. The positive critical points are the positive solutions to an algebraic system, say likelihood equations. So, identifying the number of positive critical points is a real root classification problem for the likelihood equations. A discriminant variety of a likelihood-equation system geometrically describes the data for which the number of real solutions becomes unusual. As an essential component of the discriminant variety, the nonproperness set collects the data such that the likelihood-equation system has a solution at infinity. So, the number of real solutions varies when the data passes the nonproperness set, and identifying the nonproperness set plays a crucial role in the real root classification. In this work, we develop a novel method for computing nonproperness sets of likelihood-equation systems. We prove the correctness of this method. We show experimentally that it is far more efficient than the known methods in the literature.

Predicting drought risk is essential for anticipating impacts on water resources, agriculture, ecosystems, and climate adaptation planning. Yet drought forecasts remain uncertain because variability can substantially alter regional precipitation and evaporative demand. Treating this variability as unstructured noise ignores the fact that internal variability has spatial, seasonal, and temporal structure and thus contains information that can be used to improve drought forecasting. We propose a deep-learning-based forecasting framework for European drought prediction and extend it with an uncertainty-aware drought bound that explicitly incorporates internal forecast variability from a large climate model ensemble. This bound represents a physically plausible lower-tail trajectory of future drought conditions and marks how severe drought could plausibly become under an unfavourable realisation of internal variability, giving adaptation planning a conservative, risk-averse reference. We compare the proposed bound with a lower bound derived from reanalysis data only and show that our proposed ensemble-informed bound is better calibrated across most regions and seasons. This is specifically true during anomalously dry conditions, when historical reanalysis alone underestimates lower-tail drought risk. Our results show that internal variability should be treated as a forecast quantity in its own right. More broadly, large ensembles provide a practical way to transfer physically plausible climate variability into machine-learning drought forecasts, yielding risk-aware bounds that are more informative for drought assessment under shifting climate conditions.

Scoring open-ended music analysis responses is time-consuming and requires nuanced judgments of harmonic knowledge and formal understanding. This study evaluates the validity and repeatability of GPT-4o-mini for rubric-based scoring of music analysis essays, using teacher mean scores as the benchmark. A dataset of 300 university-level student responses was scored by teachers on four dimensions: Harmony, Form, Reasoning, and Terminology. GPT-4o-mini scored the same responses using three prompting strategies: few-shot prompting with chain-of-thought reasoning (Fs+CoT), retrieval-augmented generation (RAG), and self-consistency based on five internal generations per administration (SC). Each strategy was administered three times with the model, prompt, rubric, and response held constant. Single-pass scores represented an operational scoring condition, whereas median aggregation across three runs was used to examine robustness. Agreement with teacher mean scores was evaluated using correlation, intraclass correlation, Krippendorff's alpha, quadratic weighted kappa, and scoring error indices. Fs+CoT showed the strongest agreement with teacher mean scores in both single-pass scoring and median aggregation. RAG showed systematic over-scoring, whereas SC produced highly repeatable scores but weaker individual-level agreement. Dimension-level analyses showed that scoring performance varied across rubric components, with Terminology generally showing weaker agreement than Reasoning. These findings indicate that GPT-4o-mini can generate stable scores for complex music analysis responses, but prompting strategies produce distinct scoring profiles. Operational use therefore requires strategy-specific calibration, dimension-level validation, and continued human oversight.

A Gaussian mixture density can have more modes than components. It has been conjectured that the maximum number of modes of a $d$-variate $k$-component Gaussian mixture density is $\binom{d+k-1}{d}$, which equals six for $(d,k)=(2,3)$. We construct an explicit family of equally weighted heteroscedastic three-component bivariate Gaussian mixture densities with at least seven distinct nondegenerate modes, showing that this conjectured upper bound fails for $(d,k)=(2,3)$. To the best of our knowledge, this provides the first counterexample to the conjecture across all pairs $(d,k)$.

Non-parametric statistical testing is based on surrogate data generation that randomizes chosen features in the empirical data. In the graph setting, graph signal processing (GSP) brings forward versatile schemes; e.g., to preserve smoothness of graph signals as measured by the Dirichlet energy. However, how to deal with directed graphs remains an active area of research. We begin by revisiting the definition of directed graph wide-sense stationarity. The surrogate signals preserve covariance under the stationary assumption. We demonstrate the feasibility of the scheme to detect irregular node covariance and benchmark our method against conventional schemes using the symmetrized graph. We also investigate how the level of asymmetry affects the detection performance, thus assessing the advantages of the presented approach. Finally, we show results for a real-world graph extracted from the Freeman EIES social network dataset.

Detecting a change in a multivariate series answers only the first of two questions; the operational question is which coordinates changed. Existing answers are incomplete. Block-level procedures certify predefined groups of coordinates under an additive union bound, high-dimensional variable-selection methods return interpretable rankings without error guarantees, and the post-detection inference literature controls error along the time axis rather than across coordinates. We propose ARM (Attribution by Rank Maxima), a wrapper that accepts a changepoint located by an arbitrary detector and returns the set of coordinates certified to have changed, each carrying a location or scale type label. ARM scores each coordinate by a max-over-splits rank statistic. Because this statistic dominates the corresponding statistic at the estimated split, the resulting certificate is invariant to the manner, and to the accuracy, of the changepoint estimate. Three finite-sample guarantees follow from within-coordinate ranks alone: per-coordinate validity under any detector; exact family-wise error control through a Westfall–Young joint permutation that preserves cross-coordinate dependence, with a fully distribution-free Holm fallback; and false discovery rate control under arbitrary coordinate dependence in high dimensions through Benjamini–Yekutieli and e-BH. In simulations, naive per-coordinate testing at the estimated changepoint inflates its family-wise error beyond $0.66$ as the dimension grows, whereas ARM maintains the nominal level while retaining validity under heavy tails, power in high dimensions, and accurate type labels. On five financial series surrounding the 2008 collapse, ARM attributes a scale change to every asset class and excludes injected control coordinates.