Papers

We propose a new neural network framework, termed Neural Network Machine Regression (NNMR), which integrates trainable input gating and adaptive depth regularization to jointly perform feature selection and function estimation in an end-to-end manner. By penalizing both gating parameters and redundant layers, NNMR yields sparse and interpretable architectures while capturing complex nonlinear relationships driven by high-order synergistic effects. We further develop a post-selection inference procedure based on split-sample, permutation-based hypothesis testing, enabling valid inference without restrictive parametric assumptions. Compared with existing methods, including Bayesian kernel machine regression and widely used post hoc attribution techniques, NNMR scales efficiently to high-dimensional feature spaces while rigorously controlling type I error. Simulation studies demonstrate its superior selection accuracy and inference reliability. Finally, an empirical application reveals sparse, biologically meaningful food group predictors associated with somatic growth among adolescents living in Mexico City.

We study the effect of high-order statistics of data on the learning dynamics of neural networks (NNs) by using a moment-controllable non-Gaussian data model. Considering the expressivity of two-layer neural networks, we first construct the data model as a generative two-layer NN where the activation function is expanded by using Hermite polynomials. This allows us to achieve interpretable control over high-order cumulants such as skewness and kurtosis through the Hermite coefficients while keeping the data model realistic. Using samples generated from the data model, we perform controlled online learning experiments with a two-layer NN. Our results reveal a moment-wise progression in training: networks first capture low-order statistics such as mean and covariance, and progressively learn high-order cumulants. Finally, we pretrain the generative model on the Fashion-MNIST dataset and leverage the generated samples for further experiments. The results of these additional experiments confirm our conclusions and show the utility of the data model in a real-world scenario. Overall, our proposed approach bridges simplified data assumptions and practical data complexity, which offers a principled framework for investigating distributional effects in machine learning and signal processing.

Simulating parameter-dependent stochastic differential equations (SDEs) presents significant computational challenges, as separate high-fidelity simulations are typically required for each parameter value of interest. Despite the success of machine learning methods in learning SDE dynamics, existing approaches either require expensive neural network training for score function estimation or lack the ability to handle continuous parameter dependence. We present a training-free conditional diffusion model framework for learning stochastic flow maps of parameter-dependent SDEs, where both drift and diffusion coefficients depend on physical parameters. The key technical innovation is a joint kernel-weighted Monte Carlo estimator that approximates the conditional score function using trajectory data sampled at discrete parameter values, enabling interpolation across both state space and the continuous parameter domain. Once trained, the resulting generative model produces sample trajectories for any parameter value within the training range without retraining, significantly accelerating parameter studies, uncertainty quantification, and real-time filtering applications. The performance of the proposed approach is demonstrated via three numerical examples of increasing complexity, showing accurate approximation of conditional distributions across varying parameter values.

This paper addresses the problem of designing efficient no-swap regret algorithms for combinatorial bandits, where the number of actions $N$ is exponentially large in the dimensionality of the problem. In this setting, designing efficient no-swap regret translates to sublinear – in horizon $T$ – swap regret with polylogarithmic dependence on $N$. In contrast to the weaker notion of external regret minimization - a problem which is fairly well understood in the literature - achieving no-swap regret with a polylogarithmic dependence on $N$ has remained elusive in combinatorial bandits. Our paper resolves this challenge, by introducing a no-swap-regret learning algorithm with regret that scales polylogarithmically in $N$ and is tight for the class of combinatorial bandits. To ground our results, we also demonstrate how to implement the proposed algorithm efficiently – that is, with a per-iteration complexity that also scales polylogarithmically in $N$ – across a wide range of well-studied applications.

Training predictive models across multiple centers typically assumes that all centers collect the same set of covariates. In practice, however, they may record different features of their observations, a setting we refer to as covariate mismatch. We study linear prediction under this challenging setting, assuming center-wise MCAR missingness patterns, and develop estimators that exploit information across centers despite heterogeneous feature sets. In the low-dimensional regime, we propose a plug-in estimator of the oracle linear predictor based on component-wise aggregation of covariance and cross-moment estimates. In higher dimensions, we study an impute-then-regress strategy that first completes the missing covariates using an exchangeability-preserving imputation procedure and then fits a ridge-regularized linear model. All proposed estimators are compatible with federated learning constraints: individual-level data remain local to each center, and only aggregated quantities are exchanged. We provide asymptotic and finite-sample learning rates for our predictors, explicitly characterizing their behaviour with the global dimension, the center-specific feature partition, and the distribution of samples across centers, and validate our approach through numerical experiments.

Ultrafast online learning is essential for high-frequency systems, such as controls for quantum computing and nuclear fusion, where adaptation must occur on sub-microsecond timescales. Meeting these requirements demands low-latency, fixed-precision computation under strict memory constraints, a regime in which conventional Multi-Layer Perceptrons (MLPs) are both inefficient and numerically unstable. We identify key properties of Kolmogorov-Arnold Networks (KANs) that align with these constraints. Specifically, we show that: (i) KAN updates exploiting B-spline locality are sparse, enabling superior on-chip resource scaling, and (ii) KANs are inherently robust to fixed-point quantization. By implementing fixed-point online training on Field-Programmable Gate Arrays (FPGAs), a representative platform for on-chip computation, we demonstrate that KAN-based online learners are significantly more efficient and expressive than MLPs across a range of low-latency and resource-constrained tasks. To our knowledge, this work is the first to demonstrate model-free online learning at sub-microsecond latencies.

We introduce SNAP (Self-coNsistent Agreement Principle), a self-supervised framework for robust computation based on mutual agreement. Based on an Agreement-Reliability Hypothesis SNAP assigns weights that quantify agreement, emphasizing trustworthy items and downweighting outliers without supervision or prior knowledge. A key result is the Exponential Suppression of Outlier Weights, ensuring that outliers contribute negligibly to computations, even in high-dimensional settings. We study properties of SNAP weighting scheme and show its practical benefits on vector averaging and subspace estimation. Particularly, we demonstrate that non-iterative SNAP outperforms the iterative Weiszfeld algorithm and two variants of multivariate median of means. SNAP thus provides a flexible, easy-to-use, broadly applicable approach to robust computation.

We introduce a general Bayesian framework for graph matching grounded in a new theory of exchangeable random permutations. Leveraging the cycle representation of permutations and the literature on exchangeable random partitions, we define, characterize, and study the structural and predictive properties of these probabilistic objects. A novel sequential metaphor, the position-aware generalized Chinese restaurant process, provides a constructive foundation for this theory and supports practical algorithmic design. Exchangeable random permutations offer flexible priors for a wide range of inferential problems centered on permutations. As an application, we develop a Bayesian model for graph matching that integrates a correlated stochastic block model with our novel class of priors. The cycle structure of the matching is linked to latent node partitions that explain connectivity patterns, an assumption consistent with the homogeneity requirement underlying the graph matching task itself. Posterior inference is performed through a node-wise blocked Gibbs sampler directly enabled by the proposed sequential construction. To summarize posterior uncertainty, we introduce perSALSO, an adaptation of SALSO to the permutation domain that provides principled point estimation and interpretable posterior summaries. Together, these contributions establish a unified probabilistic framework for modeling, inference, and uncertainty quantification over permutations.

Although diffusion models have successfully extended to function-valued data, stochastic interpolants – which offer a flexible way to bridge arbitrary distributions – remain limited to finite-dimensional settings. This work bridges this gap by establishing a rigorous framework for stochastic interpolants in infinite-dimensional Hilbert spaces. We provide comprehensive theoretical foundations, including proofs of well-posedness and explicit error bounds. We demonstrate the effectiveness of the proposed framework for conditional generation, focusing particularly on complex PDE-based benchmarks. By enabling generative bridges between arbitrary functional distributions, our approach achieves state-of-the-art results, offering a powerful, general-purpose tool for scientific discovery.

We consider deep multivariate models for heterogeneous collections of random variables. In the context of computer vision, such collections may e.g. consist of images, segmentations, image attributes, and latent variables. When developing such models, most existing works start from an application task and design the model components and their dependencies to meet the needs of the chosen task. This has the disadvantage of limiting the applicability of the resulting model for other downstream tasks. Here, instead, we propose to represent the joint probability distribution by means of conditional probability distributions for each group of variables conditioned on the rest. Such models can then be used for practically any possible downstream task. Their learning can be approached as training a parametrised Markov chain kernel by maximising the data likelihood of its limiting distribution. This has the additional advantage of allowing a wide range of semi-supervised learning scenarios.

The ISO 5725 series frames interlaboratory precision through repeatability, between-laboratory, and reproducibility variances, yet practical guidance on deploying bootstrap methods within this one-way random-effects setting remains limited. We study resampling strategies tailored to ISO 5725 data and extend a bias-correction idea to obtain simple adjusted point estimators and confidence intervals for the variance components. Using extensive simulations that mirror realistic study sizes and variance ratios, we evaluate accuracy, stability, and coverage, and we contrast the resampling-based procedures with ANOVA-based estimators and common approximate intervals. The results yield a clear division of labor: adjusted within-laboratory resampling provides accurate and stable point estimation in small-to-moderate designs, whereas a two-stage strategy-resampling laboratories and then resampling within each-paired with bias-corrected and accelerated intervals offers the most reliable (near-nominal or conservative) confidence intervals. Performance degrades under extreme designs, such as very small samples or dominant between-laboratory variation, clarifying when additional caution is warranted. A case study from an ISO 5725-4 dataset illustrates how the recommended procedures behave in practice and how they compare with ANOVA and approximate methods. We conclude with concrete guidance for implementing resampling-based precision analysis in interlaboratory studies: use adjusted within-laboratory resampling for point estimation, and adopt the two-stage strategy with bias-corrected and accelerated intervals for interval estimation.

Constructing accurate and computationally efficient surrogate models (or emulators) for predicting dynamical system responses is critical in many engineering domains, yet remains challenging due to the strongly nonlinear and high-dimensional mapping from external excitations and system parameters to system responses. This work introduces a novel Function-on-Function Nonlinear AutoRegressive model with eXogenous inputs (F2NARX), which reformulates the recently proposed $\mathcal{F}$-NARX method from a function-on-function regression perspective. The proposed framework substantially improves predictive efficiency while maintaining high accuracy. By combining principal component analysis with Gaussian process regression, F2NARX further enables probabilistic predictions of dynamical responses via the unscented transform in an autoregressive manner. Such probabilistic prediction capabilities further facilitate active learning for first-passage probability evaluation. The effectiveness of the method is demonstrated through case studies of varying complexity. Results show that F2NARX outperforms state-of-the-art NARX model by orders of magnitude in efficiency while achieving higher accuracy in general. Meanwhile, the active learning approach enables accurate estimation of first-passage failure probabilities for dynamical systems using only a small number of training time histories.

Privacy preservation is critical in many high-stakes domains such as medicine and finance, where sensitive data must be analyzed without compromising individual confidentiality. At the same time, these applications often involve datasets with inherent missing values due to non-response or data corruption for example. Missing data is traditionally analyzed through its impact on statistical efficiency and model performance. In fact, it reduces the information available to analysts and can degrade the final utility of the model. In this work, we take an alternative approach and study missing data through the lens of privacy preservation. Intuitively, when features are missing, less information is revealed about individuals, suggesting that data missingness could inherently enhance privacy. We formalize this intuition within a novel framework that integrates missing data into differential privacy. In essence, our approach accounts for the de facto missingness in the data to refine existing privacy guarantees without modifying the underlying mechanism. Using this framework, we show for the first time that missing data can induce inherent privacy amplification for differentially private algorithms, highlighting a previously overlooked interaction between missing data and formal privacy guarantees..

Rapidly evolving market conditions call for real-time risk monitoring, but its online estimation remains challenging. In this paper, we study the online estimation of one of the most widely used risk measures, Value at Risk (VaR). Its accurate and reliable estimation is essential for timely risk control and informed decision-making. We propose to use the quantile regression forest in the offline-simulation-online-estimation (OSOA) framework. Specifically, the quantile regression forest is trained offline to learn the relationship between the online VaR and risk factors, and real-time VaR estimates are then produced online by incorporating observed risk factors. To further ensure reliability, we develop a conformalized estimator that calibrates the online VaR estimates. To the best of our knowledge, we are the first to leverage conformal calibration to estimate real-time VaR reliably based on the OSOA formulation. Theoretical analysis establishes the consistency and coverage validity of the proposed estimators. Numerical experiments confirm the proposed method and demonstrate its effectiveness in practice.

This work studies online episodic tabular Markov decision processes (MDPs) with known transitions and develops best-of-both-worlds algorithms that achieve refined data-dependent regret bounds in the adversarial regime and variance-dependent regret bounds in the stochastic regime. We quantify MDP complexity using a first-order quantity and several new data-dependent measures for the adversarial regime, including a second-order quantity and a path-length measure, as well as variance-based measures for the stochastic regime. To adapt to these measures, we develop algorithms based on global optimization and policy optimization, both built on optimistic follow-the-regularized-leader with log-barrier regularization. For global optimization, our algorithms achieve first-order, second-order, and path-length regret bounds in the adversarial regime, and in the stochastic regime, they achieve a variance-aware gap-independent bound and a variance-aware gap-dependent bound that is polylogarithmic in the number of episodes. For policy optimization, our algorithms achieve the same data- and variance-dependent adaptivity, up to a factor of the episode horizon, by exploiting a new optimistic $Q$-function estimator. Finally, we establish regret lower bounds in terms of data-dependent complexity measures for the adversarial regime and a variance measure for the stochastic regime, implying that the regret upper bounds achieved by the global-optimization approach are nearly optimal.

In many experimental settings, active learning can improve sample efficiency by sequentially selecting where to measure, which is particularly valuable when experiments are expensive. Gaussian processes with variance-based acquisition criteria are widely used for this purpose, but have two limitations. First, they are observation-independent: their posterior variance depends only on where samples are acquired, not on what is measured, impairing their sensitivity to the structure of the acquired data. Second, they inflate the variance near boundaries, leading to excessive sampling at the edges of the space compared to the interior. These limitations undermine the gains in sampling efficiency expected from sequential acquisition. We address both limitations. We derive a reconstruction-driven design density and use the posterior mean to build a training-free warp that places more measurements where the target function varies rapidly. A geometric equalizer separately corrects boundary bias. Across sixteen synthetic and two real-data benchmarks, the geometric equalizer consistently improves function reconstruction by correcting boundary bias, while the reconstruction warp provides further gains by concentrating measurements where the posterior mean varies rapidly.

Transformers excel through content-addressable retrieval and the ability to exploit contexts of, in principle, unbounded length. We recast associative memory at the level of probability measures, treating a context as a distribution over tokens and viewing attention as an integral operator on measures. Concretely, for mixture contexts $\nu = I^{-1} \sum_{i=1}^I \mu^{(i^*)}$ and a query $x_{\mathrm{q}}(i^*)$, the task decomposes into (i) recall of the relevant component $\mu^{(i^*)}$ and (ii) prediction from $(\mu_{i^*},x_\mathrm{q})$. We study learned softmax attention (not a frozen kernel) trained by empirical risk minimization and show that a shallow measure-theoretic Transformer composed with an MLP learns the recall-and-predict map under a spectral assumption on the input densities. We further establish a matching minimax lower bound with the same rate exponent (up to multiplicative constants), proving sharpness of the convergence order. The framework offers a principled recipe for designing and analyzing Transformers that recall from arbitrarily long, distributional contexts with provable generalization guarantees.

A/B testing has become a gold standard for modern technological companies to conduct policy evaluation. Yet, its application to time series experiments, where policies are sequentially assigned over time, remains challenging. Existing designs suffer from two limitations: (i) they do not fully leverage the entire history for treatment allocation; (ii) they rely on strong assumptions to approximate the objective function (e.g., the mean squared error of the estimated treatment effect) for optimizing the design. We first establish an impossibility theorem showing that failure to condition on the full history leads to suboptimal designs, due to the dynamic dependencies in time series experiments. To address both limitations simultaneously, we next propose a transformer reinforcement learning (RL) approach which leverages transformers to condition allocation on the entire history and employs RL to directly optimize the MSE without relying on restrictive assumptions. Empirical evaluations on synthetic data, a publicly available dispatch simulator, and a real-world ridesharing dataset demonstrate that our proposal consistently outperforms existing designs.

Learning a common policy from multiple source environments requires accounting for cross-source heterogeneity, and the offline setting introduces an additional challenge of limited data coverage. We formulate this problem as a distributionally robust Markov decision process under a shared linear MDP representation with source-specific transition and reward components. The uncertainty set is constructed through feature-wise mixtures of these components, yielding a $d$-rectangular structure that captures cross-source heterogeneity and admits a robust Bellman recursion. Building on this formulation, we develop Group-Robust Pessimistic Value Iteration (GR-PEVI), an offline policy-learning algorithm that combines feature-wise worst-case aggregation with a data-dependent pessimism penalty. Under a robust partial coverage condition, we establish finite-sample suboptimality guarantees for GR-PEVI and extend the analysis to clustered sources, characterizing the trade-off between the statistical gain from pooling similar sources and the approximation error induced by within-cluster heterogeneity. Simulation studies demonstrate the effectiveness of GR-PEVI, with empirical behavior consistent with the theoretical analysis. We apply GR-PEVI to a multi-source electronic health record study of multiple sclerosis treatment decisions to illustrate its use in a clinical setting.

Stochastic gradient methods are central to large-scale learning, but they treat mini-batch gradients as unbiased estimators, which classical decision theory shows are inadmissible in high dimensions. We formulate gradient computation as a high-dimensional estimation problem and introduce a framework based on Stein-rule shrinkage. We construct a gradient estimator that adaptively contracts noisy mini-batch gradients toward a stable estimator derived from historical momentum. The shrinkage intensity is determined in a data-driven manner using an online estimate of gradient noise variance, leveraging statistics from adaptive optimizers. Under a Gaussian noise model, we show our estimator uniformly dominates the standard stochastic gradient under squared error loss and is minimax-optimal. We incorporate this into the Adam optimizer, yielding SR-Adam, a practical algorithm with negligible computational cost. Empirical evaluations on CIFAR10 and CIFAR100 across multiple levels of input noise show consistent improvements over Adam in the large-batch regime. Ablation studies indicate that gains arise primarily from selectively applying shrinkage to high-dimensional convolutional layers, while indiscriminate shrinkage across all parameters degrades performance. These results illustrate that classical shrinkage principles provide a principled approach to improving stochastic gradient estimation in deep learning.

Approximate Bayesian Computation (ABC) methods often require extensive simulations, resulting in high computational costs. This paper focuses on multifidelity simulation models and proposes a pre-filtering hierarchical importance sampling algorithm. Under mild assumptions, we theoretically prove that the proposed algorithm satisfies posterior concentration properties, characterize the error upper bound and the relationship between algorithmic efficiency and pre-filtering criteria. Additionally, we provide a practical strategy to assess the suitability of multifidelity models for the proposed method. Finally, we develop a multifidelity ABC sequential Monte Carlo with adaptive pre-filtering strategy. Numerical experiments are used to demonstrate the effectiveness of the proposed approach. We develop an R package that is available at https://github.com/caofff/MAPS

Conformal Prediction (CP) provides a statistical framework for uncertainty quantification that constructs prediction sets with coverage guarantees. While CP yields uncontrolled prediction set sizes, Backward Conformal Prediction (BCP) inverts this paradigm by enforcing a predefined upper bound on set size and estimating the resulting coverage guarantee. However, the looseness induced by Markov's inequality within the BCP framework causes a significant gap between the estimated coverage bound and the empirical coverage. In this work, we introduce ST-BCP, a novel method that introduces a data-dependent transformation of nonconformity scores to narrow the coverage gap. In particular, we develop a computable transformation and prove that it outperforms the baseline identity transformation. Extensive experiments demonstrate the effectiveness of our method, reducing the average coverage gap from 4.20% to 1.12% on common benchmarks.

Motivated by the study of how daily temperature affects soybean yield, this article proposes a simultaneous functional quantile regression (FQR) model featuring a locally sparse bivariate slope function indexed by both quantile and time and linked to a functional predictor. The slope function's local sparsity means it holds non-zero values only in certain segments of its domain, remaining zero elsewhere. These zero-slope regions, which vary by quantile, indicate times when the functional predictor has no discernible impact on the response variable. This feature boosts the model's interpretability. Unlike traditional FQR models, which fit one quantile at a time and have several limitations, our proposed method can handle a spectrum of quantiles simultaneously. We tested the new approach through simulation studies, demonstrating its clear advantages over standard techniques. To validate its practical use, we applied the method to soybean yield data, pinpointing the time periods when daily temperature doesn't affect yield. This insight could be crucial for agricultural planning and crop management.

We study online inverse linear optimization, also known as contextual recommendation, where a learner sequentially infers an agent's hidden objective vector from observed optimal actions over feasible sets that change over time. The learner aims to recommend actions that perform well under the agent's true objective, and the performance is measured by the regret, defined as the cumulative gap between the agent's optimal values and those achieved by the learner's recommended actions. Prior work has established a regret bound of $O(d\log T)$, as well as a finite but exponentially large bound of $\exp(O(d\log d))$, where $d$ is the dimension of the optimization problem and $T$ is the time horizon, while a regret lower bound of $\Omega(d)$ is known (Gollapudi et al. 2021; Sakaue et al. 2025). Whether a finite regret bound polynomial in $d$ is achievable or not has remained an open question. We partially resolve this by showing that when the feasible sets are M-convex – a broad class that includes matroids – a finite regret bound of $O(d\log d)$ is possible. We achieve this by combining a structural characterization of optimal solutions on M-convex sets with a geometric volume argument. Moreover, we extend our approach to adversarially corrupted feedback in up to $C$ rounds. We obtain a regret bound of $O((C+1)d\log d)$ without prior knowledge of $C$, by monitoring directed graphs induced by the observed feedback to detect corruptions adaptively.

The doubly-robust (DR) estimator is popular for evaluating causal effects in observational studies and is often perceived as more desirable than inverse probability weighting (IPW) or outcome modeling alone because it provides extra protection against model misspecification. However, double robustness is an asymptotic property that may not hold in finite samples. We investigate how the finite sample performance of the DR estimator depends on the degree of covariate overlap between comparison groups. Using analytical illustrations and extensive simulations under various scenarios with different degrees of covariate overlap and model specifications, we examine the bias and variance of the DR estimator relative to IPW and outcome modeling estimators. We find that: (i) specification of the outcome model has a stronger influence on the DR estimates than specification of the propensity score model, and this dominance increases as overlap decreases; (ii) with poor overlap, the DR estimator generally amplifies the adverse consequences of extreme weights (large bias and/or variance) regardless of model specifications, and is often inferior to both the IPW and outcome modeling estimators. As a practical guide, we recommend always first checking the degree of overlap in applications. In the case of poor overlap, analysts should consider shifting the target population to a subpopulation with adequate overlap via methods such as trimming or overlap weighting.

With limited high-quality data and growing compute, multi-epoch training is gaining back its importance across sub-areas of deep learning. Adam(W), versions of which are go-to optimizers for many tasks such as next token prediction, has two momentum hyperparameters $(\beta_1, \beta_2)$ controlling memory and one very important hyperparameter, batch size, controlling (in particular) the amount mini-batch noise. We introduce a theoretical framework to understand how mini-batch noise influences the implicit bias of memory in Adam (depending on $\beta_1$, $\beta_2$) towards sharper or flatter regions of the loss landscape, which is commonly observed to correlate with the generalization gap in multi-epoch training. We find that in the case of large batch sizes, higher $\beta_2$ increases the magnitude of anti-regularization by memory (hurting generalization), but as the batch size becomes smaller, the dependence of (anti-)regulariation on $\beta_2$ is reversed. A similar monotonicity shift (in the opposite direction) happens in $\beta_1$. In particular, the commonly "default" pair $(\beta_1, \beta_2) = (0.9, 0.999)$ is a good choice if batches are small; for larger batches, in many settings moving $\beta_1$ closer to $\beta_2$ is much better in terms of validation accuracy in multi-epoch training. Moreover, our theoretical derivations connect the scale of the batch size at which the shift happens to the scale of the critical batch size. We illustrate this effect in experiments with small-scale data in the about-to-overfit regime.

Estimating causal effects under interference, where the stable unit treatment value assumption is violated, is critical in fields such as regional and public economics. Much of the existing research on causal inference under interference relies on a pre-specified "exposure mapping". This paper focuses on difference-in-difference and proposes a nonparametric identification strategy for direct and indirect average treatment effects under local interference on an observed network. In particular, we proposed a new concept of an indirect effect measuring the total outward influence of the intervension. Based on parallel trends assumption conditional on the neighborhood treatment vector, we develop inverse probability weighted and doubly robust estimators. We establish their asymptotic properties, including consistency under misspecification of nuisance models under some regularity conditions. Simulation studies and an empirical application demonstrate the effectiveness of the proposed method.

Differentially private synthetic data enables the sharing and analysis of sensitive datasets while providing rigorous privacy guarantees for individual contributors. A central challenge is to achieve strong utility guarantees for meaningful downstream analysis. Many existing methods ensure uniform accuracy over broad query classes, such as all Lipschitz functions, but this level of generality often leads to suboptimal rates for statistics of practical interest. Since many common data analysis queries exhibit smoothness beyond what worst-case Lipschitz bounds capture, we ask whether exploiting this additional structure can yield improved utility. We study the problem of generating $(\varepsilon,\delta)$-differentially private synthetic data from a dataset of size $n$ supported on the hypercube $[-1,1]^d$, with utility guarantees uniformly for all smooth queries having bounded derivatives up to order $k$. We propose a polynomial-time algorithm that achieves a minimax error rate of $O_{k,d}(n^{-\min \{1, \frac{k}{d}\}})$, up to a $\log(n)$ factor. This characterization uncovers a phase transition at $k=d$. Our results generalize the Chebyshev moment matching framework of (Musco et al., 2025; Wang et al., 2016) and strictly improve the error rates for $k$-smooth queries established in \citep{wang2016differentially}. Moreover, we establish the first minimax lower bound for the utility of $(\varepsilon,\delta)$-differentially private synthetic data with respect to $k$-smooth queries, extending the Wasserstein lower bound for $\varepsilon$-differential privacy in (Boedihardjo et al., 2024).

Time Series Foundation Models (TSFMs) leverage extensive pretraining to accurately predict unseen time series during inference, without the need for task-specific fine-tuning. Through large-scale evaluations on standard benchmarks, we find that leading transformer-based TSFMs exhibit redundant components in their intermediate layers. We introduce a set of tools for mechanistic interpretability of TSFMs, including ablations of specific components and direct logit attribution on the residual stream. Our findings are consistent across several leading TSFMs with diverse architectures, and across a diverse set of real-world and synthetic time-series datasets. We discover that all models in our study are robust to ablations of entire layers. Furthermore, we develop a theoretical framework framing transformers as kernel regressors, motivating a purely intrinsic strategy for ablating heads based on the stable rank of the per-head projection matrices. Using this approach, we uncover the specific heads responsible for degenerate phenomena widely observed in TSFMs, such as parroting of motifs from the context and seasonality bias. Our study sheds light on the universal properties of this emerging class of architectures for continuous-time sequence modeling.

Aligning large language models (LLMs) to diverse human preferences is fundamentally challenging since criteria can often conflict with each other. Inference-time alignment methods have recently gained popularity as they allow LLMs to be aligned to multiple criteria via different alignment algorithms at inference time. However, inference-time alignment is computationally expensive since it often requires multiple forward passes of the base model. In this work, we propose inference-aware meta-alignment (IAMA), a novel approach that enables LLMs to be aligned to multiple criteria with limited computational budget at inference time. IAMA trains a base model such that it can be effectively aligned to multiple tasks via different inference-time alignment algorithms. To solve the non-linear optimization problems involved in IAMA, we propose non-linear GRPO, which provably converges to the optimal solution in the space of probability measures.