Papers

Measurement non-invariance arises when the psychometric properties of a scale differ across subgroups, undermining the validity of group comparisons. At the item level, this manifests as differential item functioning (DIF), where item responses differ across groups after controlling for the latent trait. This paper develops a framework for detecting DIF in ordinal scales without requiring known group labels or anchor items. We formulate a proportional-odds latent-class item response model in which individuals are assigned probabilistically to latent classes. DIF is captured through class-specific intercept and slope shifts, allowing both uniform and non-uniform DIF. Identification is achieved through an \(\ell_1\)-penalised marginal likelihood under a sparsity assumption, with estimation implemented using a tailored EM algorithm. Because class-specific slopes leave both the location and scale of each latent class unidentified, sparsity anchors the latent metric while selecting DIF effects. Simulation studies demonstrate accurate recovery of item parameters and both types of DIF. An empirical application to a personality test reveals latent subgroups with distinct response patterns and identifies items displaying potential class-specific measurement non-invariance. The framework provides a flexible approach for assessing measurement invariance in ordinal scales when comparison groups are unobserved or poorly defined.

We propose a nonparametric method for dynamic prediction in event history analysis with high-dimensional, time-dependent covariates. The approach estimates future conditional hazards by combining landmarking supermodels with gradient boosted trees. Unlike joint modeling or Cox landmarking models, the proposed estimator flexibly captures interactions and nonlinear effects without imposing restrictive parametric assumptions or requiring the covariate process to be Markovian. We formulate the approach as a sieve M-estimator and establish weak consistency. Computationally, the problem reduces to a Poisson regression, allowing implementation via standard gradient boosting software. A key theoretical advantage is that the method avoids the temporal inconsistencies that arise in landmark Cox models. Simulation studies demonstrate that the method performs well in a variety of settings, and its practical value is illustrated through an analysis of primary biliary cirrhosis data.

Hidden community problems, such as community detection in the Stochastic Block Model (SBM), submatrix localization, and $\mathbb{Z}_2$ synchronization, have received considerable attention in the probability, statistics, and information-theory literature. Motivated by transitive behavior in social networks, which tend to exhibit high triangle density, recent works have considered hidden community models in spatially-embedded networks. In particular, Baccelli and Sankararaman proposed the Geometric SBM, a spatially-embedded analogue of the standard SBM with dramatically more triangles. In this paper, we consider the problem of exact recovery for the Geometric Hidden Community Model (GHCM) of Gaudio, Guan, Niu, and Wei, which generalizes the Geometric SBM to allow for arbitrary pairwise observation distributions. Under mild technical assumptions, we find the information-theoretic threshold for exact recovery in the “distance-dependent” GHCM, which allows the pairwise distributions to depend on distance as well as community labels, thus completing the picture of exact recovery in spatially-embedded hidden community models.

The R ecosystem offers a rich variety of map-reduce application programming interfaces (APIs) for iterative computations, yet parallelizing code across these diverse frameworks requires learning multiple, often incompatible, parallel APIs. The futurize package addresses this challenge by providing a single function, futurize(), which transpiles sequential map-reduce expressions into their parallel equivalents in the future ecosystem, which performs all the heavy lifting. By leveraging R's native pipe operator, users can parallelize existing code with minimal refactoring – often by simply appending `|> futurize()' to an expression. The package supports classical map-reduce functions from base R, purrr, crossmap, foreach, plyr, BiocParallel, e.g., lapply(xs, fcn) |> futurize() and map(xs, fcn) |> futurize(), as well as a growing set of domain-specific packages, e.g., boot, caret, glmnet, lme4, mgcv, and tm. By abstracting away the underlying parallel machinery, and unifying handling of future options, the package enables developers to declare what to parallelize via futurize(), and end-users to choose how via plan(). This article describes the philosophy, design, and implementation of futurize, demonstrates its usage across various map-reduce paradigms, and discusses its role in simplifying parallel computing in R.

Rank-based dependence measures such as Spearman's footrule are robust and invariant, but they often fail to capture directional or asymmetric dependence in multivariate settings. This paper introduces a new family of directional Spearman's footrule coefficients for multivariate data, defined within the copula framework to clearly separate marginal behavior from dependence structure. We establish their main theoretical properties, showing full consistency with the classical footrule, including behavior under independence and extreme dependence, as well as symmetry and reflection properties. Nonparametric rank-based estimators are proposed and their asymptotic consistency is discussed. Explicit expressions for several known families of copulas illustrate the ability of the proposed coefficients to detect directional dependence patterns undetected by classical measures.

The purpose of this paper is to study the role of the relevation transform, where a failed unit is replaced by a used unit with the same age as the failed one, as an alternative to the policy based on the replacement by a new one. In particular, we compare the stochastic processes arising from a policy based on the replacement of a failed unit by a new one and from the one in which the unit is being continuously subjected to a relevation policy. The comparisons depend on the aging properties of the units under repair.

In this paper we develop a new tool for the comparison of paired data based on a new criterion of stochastic dominance that takes into account the dependence structure of the random variables under comparison. This new procedure provides a more detailed comparison of dependent random variables and overcomes some difficulties of standard techniques like Student's t and Wilcoxon-Mann-Whitney tests for non normal data. This tool provides an alternative to the usual stochastic dominance criterion which only considers the marginal distributions in the comparison. We show how this new tool can be fruitfully used for the comparison of paired asset returns.

This article presents the full, original record of the 2024 Joint Statistical Meetings (JSM) town hall, "Statistics in the Age of AI," which convened leading statisticians to discuss how the field is evolving in response to advances in artificial intelligence, foundation models, large-scale empirical modeling, and data-intensive infrastructures. The town hall was structured around open panel discussion and extensive audience Q&A, with the aim of eliciting candid, experience-driven perspectives rather than formal presentations or prepared statements. This document preserves the extended exchanges among panelists and audience members, with minimal editorial intervention, and organizes the conversation around five recurring questions concerning disciplinary culture and practices, data curation and "data work," engagement with modern empirical modeling, training for large-scale AI applications, and partnerships with key AI stakeholders. By providing an archival record of this discussion, the preprint aims to support transparency, community reflection, and ongoing dialogue about the evolving role of statistics in the data- and AI-centric future.

In this paper we focus on providing sufficient conditions for some transform orders for which the quantile densities ratio is non-monotone and, therefore, the convex transform order does not hold. These results are interesting for comparing random variables with a non-explicit expression of their quantile functions or they are computationally complex. In addition, the main results are applied to compare two Tukey generalized distributed random variables and to establish new relationships among non-monotone and positive aging notions.

We formulate a sparse-to-dense reconstruction layer for fractured media in which sparse point measurements are mapped onto piecewise-planar fracture supports inferred from 3D trace polylines. Each plane is discretized in local coordinates and estimated via a convex objective that combines a grid SPDE/GMRF quadratic prior with an $\ell_1$ penalty on undecimated discrete curvelet coefficients, targeting anisotropic, fracture-aligned structure that is poorly represented by isotropic smoothness alone. We further define an along-fracture distance through trace-network geodesics and express connectivity-driven regularization as a quadratic form $z^\top P^\top L_G P z$, where $L_G$ is a graph Laplacian on the trace network and $P$ maps plane grids to graph nodes; plane intersections are handled by linear consistency constraints sampled along intersection lines. The resulting optimization admits efficient splitting: sparse linear solves for the quadratic block and coefficient-wise shrinkage for the curvelet block, with standard ADMM convergence under convexity. We specify reproducible synthetic benchmarks, baselines, ablations, and sensitivity studies that isolate directional sparsity and connectivity effects, and provide reference code to generate the figures and quantitative tables.

Task-oriented semantic communication (SemCom) prioritizes task execution over accurate symbol reconstruction and is well-suited to emerging intelligent applications. Cooperative multi-task SemCom (CMT-SemCom) further improves task execution performance. However, [1] demonstrates that cooperative multi-tasking can be either constructive or destructive. Moreover, the existing CMT-SemCom framework is not directly applicable to distributed multi-user scenarios, such as non-terrestrial satellite networks, where each satellite employs an individual semantic encoder. In this paper, we extend our earlier CMT-SemCom framework to distributed settings by proposing a federated learning (FL) based CMT-SemCom that enables cooperative multi-tasking across distributed users. Moreover, to address performance degradation caused by negative information transfer among heterogeneous tasks, we propose a semantic-aware task clustering method integrated in the FL process to ensure constructive cooperation based on an information-theoretic approach. Unlike common clustering methods that rely on high-dimensional data or feature space similarity, our proposed approach operates in the low-dimensional semantic domain to identify meaningful task relationships. Simulation results based on a LEO satellite network setup demonstrate the effectiveness of our approach and performance gain over unclustered FL and individual single-task SemCom.

We propose a variational autoencoder (VAE) approach for parameter estimation in nonlinear mixed-effects models based on ordinary differential equations (NLME-ODEs) using longitudinal data from multiple subjects. In moderate dimensions, likelihood-based inference via the stochastic approximation EM algorithm (SAEM) is widely used, but it relies on Markov Chain Monte-Carlo (MCMC) to approximate subject-specific posteriors. As model complexity increases or observations per subject are sparse and irregular, performance often deteriorates due to a complex, multimodal likelihood surface which may lead to MCMC convergence difficulties. We instead estimate parameters by maximizing the evidence lower bound (ELBO), a regularized surrogate for the marginal likelihood. A VAE with a shared encoder amortizes inference of subject-specific random effects by avoiding per-subject optimization and the use of MCMC. Beyond pointwise estimation, we quantify parameter uncertainty using observed-information-based variance estimator and verify that practical identifiability of the model parameters is not compromised by nuisance parameters introduced in the encoder. We evaluate the method in three simulation case studies (pharmacokinetics, humoral response to vaccination, and TGF-$\beta$ activation dynamics in asthmatic airways) and on a real-world antibody kinetics dataset, comparing against SAEM baselines.

Data-driven methods for the solution of inverse problems have become widely popular in recent years thanks to the rise of machine learning techniques. A popular approach concerns the training of a generative model on additional data to learn a bespoke prior for the problem at hand. In this article we present an analysis for such problems by presenting quantitative error bounds for minimum Wasserstein-2 generative models for the prior. We show that under some assumptions, the error in the posterior due to the generative prior will inherit the same rate as the prior with respect to the Wasserstein-1 distance. We further present numerical experiments that verify that aspects of our error analysis manifests in some benchmarks followed by an elliptic PDE inverse problem where a generative prior is used to model a non-stationary field.

We study statistical process control (SPC) through charting of $p$-values. When in control (IC), any valid sequence $(P_{t})_{t}$ is super-uniform, a requirement that can hold in nonparametric and two-phase designs without parametric modelling of the monitored process. Within this framework, we analyse the Shewhart rule that signals when $P_{t}\le\alpha$. Under super-uniformity alone, and with no assumptions on temporal dependence, we derive universal IC lower bounds for the average run length (ARL) and for the expected time to the $k$th false alarm ($k$-ARL). When conditional super-uniformity holds, these bounds sharpen to the familiar $\alpha^{-1}$ and $k\alpha^{-1}$ rates, giving simple, distribution-free calibration for $p$-value charts. Beyond thresholding, we use merging functions for dependent $p$-values to build EWMA-like schemes that output, at each time $t$, a valid $p$-value for the hypothesis that the process has remained IC up to $t$, enabling smoothing without ad hoc control limits. We also study uniform EWMA processes, giving explicit distribution formulas and left-tail guarantees. Finally, we propose a modular approach to directional and coordinate localisation in multivariate SPC via closed testing, controlling the family-wise error rate at the time of alarm. Numerical examples illustrate the utility and variety of our approach.

It is well known that the asymptotic variance of sample quantiles can be reduced under heterogeneity relative to the i.i.d. setting. However, asymptotically correct confidence intervals for quantiles are not yet available. We propose a novel, consistent estimator of the reduced asymptotic variance arising when quantiles are computed from groups of observations, leading to asymptotically correct confidence intervals. Simulation studies show that our confidence intervals are substantially shorter than those in the i.i.d. case and attain nearly correct coverage across a wide range of heterogeneous settings.

The grade of membership model is a flexible latent variable model for analyzing multivariate categorical data through individual-level mixed membership scores. In many modern applications, auxiliary covariates are collected alongside responses and encode information about the same latent structure. Traditional approaches to incorporating such covariates typically rely on fully specified joint likelihoods, which are computationally intensive and sensitive to misspecification. We introduce a covariate-assisted grade of membership model that integrates response and covariate information by exploiting their shared low-rank simplex geometry, rather than modeling their joint distribution. We propose a likelihood-free spectral estimation procedure that combines heterogeneous data sources through a balance parameter controlling their relative contribution. To accommodate high-dimensional and heteroskedastic noise, we employ heteroskedastic principal component analysis before performing simplex-based geometric recovery. Our theoretical analysis establishes weaker identifiability conditions than those required in the covariate-free model, and further derives finite-sample, entrywise error bounds for both mixed membership scores and item parameters. These results demonstrate that auxiliary covariates can provably improve latent structure recovery, yielding faster convergence rates in high-dimensional regimes. Simulation studies and an application to educational assessment data illustrate the computational efficiency, statistical accuracy, and interpretability gains of the proposed method. The code for reproducing these results is open-source and available at https://github.com/Toby-X/Covariate-Assisted-GoM

The advancement of vehicle automation and the growing adoption of electric vehicles (EVs) are reshaping transportation systems. While fully automated vehicles are expected to improve traffic stability, efficiency, and sustainability, recent studies suggest that partially automated vehicles, such as those equipped with adaptive cruise control (ACC), may adversely affect traffic flow. These drawbacks may not extend to ACC-enabled EVs due to their distinct mechanical characteristics, including regenerative braking and smoother torque delivery. As a result, the impacts of EVs operating under ACC remain insufficiently understood. To address this gap, this study develops an empirical framework using the OpenACC dataset to compare ACC-enabled EVs and internal combustion engine vehicles. Dynamic time warping aligns comparable lead-vehicle trajectories. Results show that EVs exhibit smoother speed profiles, lower speed variability, and shorter spacing, leading to higher efficiency. EVs reduce critical safety events by over 85% and lower platoon-level emissions by up to 26.2%.

Chronic kidney disease (CKD) affects millions worldwide and progresses irreversibly through stages culminating in end-stage renal disease (ESRD) and death. Outcome trials in CKD traditionally employ time-to-first-event analyses using the Cox models. However, this approach has fundamental limitations for progressive diseases: it assigns equal weight to each composite endpoint component despite clear clinical hierarchy: an eGFR decline threshold receives the same weight as ESRD or death in the analysis, and it captures only the first occurrence while ignoring subsequent progression. Given CKD's gradual evolution over years, comprehensive treatment evaluation requires quantifying cumulative disease burden: integrating both event severity and time spent in each disease state. We propose two complementary approaches to better characterize treatment benefits by incorporating event severity and state occupancy: area under the curve (AUC) and restricted mean time in favor of treatment (RMT-IF). The AUC method assigns ordinal severity scores to disease states and calculates the area under the mean cumulative score curve, quantifying total event-free time lost. Treatment effects are expressed as AUC ratios or differences. The RMT-IF extends restricted mean survival time to multistate processes, measuring average time patients in the treatment arm spend in more favorable states versus the comparator. These methods better capture CKD's progressive nature where treatment benefits extend beyond first-event delay to overall disease trajectory modification. By discriminating between events of differing clinical importance and quantifying the complete disease course, these estimands offer alternative assessment frameworks for kidney-protective therapies, potentially improving efficiency and interpretability of future CKD outcome trials.

Count endpoints are common in clinical trials, particularly for recurrent events such as hypoglycemia. When interest centers on comparing overall event rates between treatment groups, negative binomial (NB) regression is widely used because it accommodates overdispersion and requires only event counts and exposure times. However, NB regression can be numerically unstable when events are sparse, and the efficiency gains from baseline covariate adjustment may be sensitive to model misspecification. We propose an empirical method that targets the same marginal estimand as NB regression – the ratio of marginal event rates – while avoiding distributional assumptions on the count outcome. Simulation studies show that the empirical method maintains appropriate Type I error control across diverse scenarios, including extreme overdispersion and zero inflation, achieves power comparable to NB regression, and yields consistent efficiency gains from baseline covariate adjustment. We illustrate the approach using severe hypoglycemia data from the QWINT-5 trial comparing insulin efsitora alfa with insulin degludec in adults with type 1 diabetes. In this sparse-event setting, the empirical method produced stable marginal rate estimates and rate ratios closely aligned with observed rates, while NB regression exhibited greater sensitivity and larger deviations from the observed rates in the sparsest intervals. The proposed empirical method provides a robust and numerically stable alternative to NB regression, particularly when the number of events is low or when numerical stability is a concern.

Transfer learning (TL) has emerged as a powerful tool for improving estimation and prediction performance by leveraging information from related datasets, with the offset TL (O-TL) being a prevailing implementation. In this paper, we adapt the control-variates (CVS) method for TL and develop CVS-based estimators for scalar-on-function regression, one of the most fundamental models in functional data analysis. These estimators rely exclusively on dataset-specific summary statistics, thereby avoiding the pooling of subject-level data and remaining applicable in privacy-restricted or decentralized settings. We establish, for the first time, a theoretical connection between O-TL and CVS-based TL, showing that these two seemingly distinct TL strategies adjust local estimators in fundamentally similar ways. We further derive convergence rates that explicitly account for the unavoidable but typically overlooked smoothing error arising from discretely observed functional predictors, and clarify how similarity among covariance functions across datasets governs the performance of TL. Numerical studies support the theoretical findings and demonstrate that the proposed methods achieve competitive estimation and prediction performance compared with existing alternatives.

Discrete Markov random fields are undirected graphical models that capture complex conditional dependencies between discrete variables. Conducting exact posterior inference in these models is often computationally challenging because evaluating their normalizing constant requires summation over all possible state configurations, and the size of this state space grows exponentially with the number of variables and their possible states. As a result, exact likelihood-based inference is infeasible in many practical settings, and existing methods, such as Double Metropolis-Hastings or pseudo-likelihood approximations, either scale poorly to large systems or underestimate posterior variability. To address these limitations, we propose a new class of coordinate-rescaling sampling methods that transform pseudo-likelihood-based posteriors to approximate the covariance structure of the full likelihood posterior while preserving computational efficiency. In a simulation study and an empirical application, the coordinate-rescaling samplers correct the underestimated posterior uncertainty of the pseudo-likelihood and reproduce the Bayes factors of a full-likelihood reference, with the accuracy of post hoc calibration of the pseudo-posterior and at the cost of the pseudo-likelihood sampler, providing a scalable and reliable approach to Bayesian inference in discrete MRFs.

We study Kantorovich duality for multimarginal optimal transport (MOT) with bounded continuous cost functions. The main focus is not only the equality between the primal and dual values, but also the structure of optimal dual potentials. For compact metric spaces, we prove that the dual problem admits an optimizer in the class of mutually $c$-conjugate families. The proof combines a Fenchel–Rockafellar duality argument with equicontinuity estimates, a normalization of the dual potentials, and the Arzelà–Ascoli theorem. We then consider the non-compact case, where the marginal spaces are Polish. We first recover the Kantorovich duality identity by a truncation and tightness argument, using the boundedness of the cost and Prokhorov compactness. For dual attainment, we rely on the geometry of optimal supports. Under a natural support-splitting condition, we obtain a bounded Borel measurable dual optimizer in the canonical class of mutually $c$-conjugate families. Thus, the paper identifies mutually $c$-conjugate potentials as natural representatives of optimal dual families in the multimarginal setting. These results provide a structural basis for further work on stability, empirical multimarginal transport, and statistical limit theory.

We develop a data-driven information-theoretic framework for sharp partial identification of causal effects under unmeasured confounding. Existing approaches often rely on restrictive assumptions, such as bounded or discrete outcomes; require external inputs (for example, instrumental variables, proxies, or user-specified sensitivity parameters); necessitate full structural causal model specifications; or focus solely on population-level averages while neglecting covariate-conditional effects. We overcome all four limitations simultaneously by establishing novel information-theoretic, data-driven divergence bounds. Our key theoretical contribution shows that the f-divergence between the observational distribution P(Y | A = a, X = x) and the interventional distribution P(Y | do(A = a), X = x) is upper bounded by a function of the propensity score alone. This result enables sharp partial identification of conditional causal effects directly from observational data, without requiring external sensitivity parameters, auxiliary variables, full structural specifications, or outcome boundedness assumptions. For practical implementation, we develop a semiparametric estimator satisfying Neyman orthogonality (Chernozhukov et al., 2018), which ensures root-n consistent inference even when nuisance functions are estimated via flexible machine learning methods. Simulation studies and real-world data applications, implemented in the GitHub repository (https://github.com/yonghanjung/Information-Theretic-Bounds), demonstrate that our framework provides tight and valid causal bounds across a wide range of data-generating processes.

Aggregated Relational Data (ARD) contain summary information about individual social networks and are widely used to estimate social network characteristics and the size of populations of interest. Although a variety of ARD estimators exist, practitioners currently lack guidance on how to evaluate whether a selected model adequately fits the data. We introduce a diagnostic framework for ARD models that provides a systematic, reproducible process for assessing covariate structure, distributional assumptions, and correlation. The diagnostics are based on point estimates, using either maximum likelihood or maximum a posteriori optimization, which allows quick evaluation without requiring repeated Bayesian model fitting. Through simulation studies and applications to large ARD datasets, we show that the proposed workflow identifies common sources of model misfit and helps researchers select an appropriate model that adequately explains the data.

Empirical investigations into unintended model behavior often show that the algorithm is predicting another outcome than what was intended. These exposés highlight the need to identify when algorithms predict unintended quantities - ideally before deploying them into consequential settings. We propose a falsification framework that provides a principled statistical test for discriminant validity: the requirement that an algorithm predict intended outcomes better than impermissible ones. Drawing on falsification practices from causal inference, econometrics, and psychometrics, our framework compares calibrated prediction losses across outcomes to assess whether the algorithm exhibits discriminant validity with respect to a specified impermissible proxy. In settings where the target outcome is difficult to observe, multiple permissible proxy outcomes may be available; our framework accommodates both this setting and the case with a single permissible proxy. Throughout we use nonparametric hypothesis testing methods that make minimal assumptions on the data-generating process. We illustrate the method in an admissions setting, where the framework establishes discriminant validity with respect to gender but fails to establish discriminant validity with respect to race. This demonstrates how falsification can serve as an early validity check. We also provide analysis in a criminal justice setting, where we highlight the limitations of our framework and emphasize the need for complementary approaches to assess other aspects of construct validity and external validity.

To minimize the mean squared error (MSE) in global average treatment effect (GATE) estimation under network interference, a popular approach is to use a cluster-randomized design. However, in the presence of homophily, which is common in social networks, cluster randomization can instead increase the MSE. We develop a novel potential outcomes model that accounts for interference, homophily, and heterogeneous variation. In this setting, we establish a framework for optimizing designs for worst-case MSE under the Horvitz-Thompson estimator. This leads to an optimization problem over the covariance matrices of the treatment assignment, trading off interference, homophily, and robustness. We frame and solve this problem using two complementary approaches. The first involves formulating a semidefinite program (SDP) and employing Gaussian rounding, in the spirit of the Goemans-Williamson approximation algorithm for MAXCUT. The second is an adaptation of the Gram-Schmidt Walk, a vector-balancing algorithm which has recently received much attention. Finally, we evaluate the performance of our designs through various experiments on simulated network data and a real village network dataset.

Varying coefficient models are widely used to characterize dynamic associations between longitudinal outcomes and covariates. Existing work on varying coefficient models, however, all assumes that observation times are independent of the longitudinal outcomes, which is often violated in real-world studies with outcome-driven or otherwise informative visit schedules. Such informative observation times can lead to biased estimation and invalid inference using existing methods. In this article, we develop estimation and inference procedures for varying coefficient models that account for informative observation times. We model the observation time process as a general counting process under a proportional intensity model, with time-varying covariates summarizing the observed history. To address potential bias, we incorporate inverse intensity weighting into a sieve estimation framework, yielding closed-form coefficient function estimators via weighted least squares. We establish consistency, convergence rates, and asymptotic normality of the proposed estimators, and construct pointwise confidence intervals for the coefficient functions. Extensive simulation studies demonstrate that the proposed weighted method substantially outperforms the conventional unweighted method when observation times are informative. Finally, we provide an application of our method to the Alzheimer's Disease Neuroimaging Initiative study.

Brain organization is increasingly characterized through multiple imaging modalities, most notably structural connectivity (SC) and functional connectivity (FC). Integrating these inherently distinct yet complementary data sources is essential for uncovering the cross-modal patterns that drive behavioral phenotypes. However, effective integration is hindered by the high dimensionality and non-linearity of connectome data, complex non-linear SC-FC coupling, and the challenge of disentangling shared information from modality-specific variations. To address these issues, we propose the Cross-Modal Joint-Individual Variational Network (CM-JIVNet), a unified probabilistic framework designed to learn factorized latent representations from paired SC-FC datasets. Our model utilizes a multi-head attention fusion module to capture non-linear cross-modal dependencies while isolating independent, modality-specific signals. Validated on Human Connectome Project Young Adult (HCP-YA) data, CM-JIVNet demonstrates superior performance in cross-modal reconstruction and behavioral trait prediction. By effectively disentangling joint and individual feature spaces, CM-JIVNet provides a robust, interpretable, and scalable solution for large-scale multimodal brain analysis.

Understanding the curvature evolution of the loss landscape is fundamental to analyzing the training dynamics of neural networks. The most commonly studied measure, Hessian sharpness ($\lambda_{\max}^H$) – the largest eigenvalue of the loss Hessian – determines local training stability and interacts with the learning rate throughout training. Despite its significance in analyzing training dynamics, direct measurement of Hessian sharpness remains prohibitive for Large Language Models (LLMs) due to high computational cost. We analyze $\textit{critical sharpness}$ ($\lambda_c$), a computationally efficient measure requiring fewer than $10$ forward passes given the update direction $\Delta \mathbf{\theta}$. Critically, this measure captures well-documented Hessian sharpness phenomena, including progressive sharpening and Edge of Stability. Using this measure, we provide the first demonstration of these sharpness phenomena at scale, up to $7$B parameters, spanning both pre-training and mid-training of OLMo-2 models. We further introduce $\textit{relative critical sharpness}$ ($\lambda_c^{1\to 2}$), which quantifies the curvature of one loss landscape while optimizing another, to analyze the transition from pre-training to fine-tuning and guide data mixing strategies. Critical sharpness provides practitioners with a practical tool for diagnosing curvature dynamics and informing data composition choices at scale. More broadly, our work shows that scalable curvature measures can provide actionable insights for large-scale training.

The field of astronomy is experiencing a data explosion driven by significant advances in observational instrumentation, and classical methods often fall short of addressing the complexity of modern astronomical datasets. Probabilistic graphical models offer powerful tools for uncovering the dependence structures and data-generating processes underlying a wide array of cosmic variables. By representing variables as nodes in a network, these models allow for the visualization and analysis of the intricate relationships that underpin theories of hierarchical structure formation within the universe. We highlight the value that graphical models bring to astronomical research by demonstrating their practical application to the study of exoplanets and host stars.