MARCUS tackles the bottleneck of human interpretation in cardiovascular diagnosis by creating an agentic, multimodal vision-language model that jointly reasons over raw ECG signals, echocardiogram videos, and cardiac MRI. The core innovation is a hierarchical architecture where modality-specific expert encoders feed into an orchestrating agent that synthesizes findings while resisting 'mirage reasoning'—the tendency of VLMs to confabulate explanations without actually processing the image. Trained on 13.5 million clinical images and 1.6 million expert-curated Q&A pairs, MARCUS aims to bridge the gap between single-task diagnostic AI and interactive clinical reasoning.

The paper addresses functional Gaussian Process regression on compact Riemannian manifolds, proposing a time-adaptive Empirical Bayes framework that exploits invariance of covariance kernels under isometries and spectral decomposition via Laplace–Beltrami eigenfunctions. The core idea is to work in the time-varying angular spectral domain, truncating the infinite-dimensional expansion based on functional sample size (typically logarithmic) to balance computational cost with approximation accuracy. This matters because it extends GP regression to infinite-dimensional functional settings on non-Euclidean domains while attempting to maintain computational tractability through spectral truncation schemes.

Sodium-ion batteries need high-capacity anodes with fast ion transport, but hard carbon suffers from structural disorder and slow diffusion. This computational study uses the SpookyNet machine-learning force field with DFT to characterize aminobenzene-functionalized Janus graphene at room temperature. The work identifies a three-stage sodium storage mechanism and predicts a high capacity of ~400 mAh g$^{-1}$ with diffusion coefficients two to three orders of magnitude above hard carbon.

Time-dependent reliability analysis for nonlinear dynamical systems under stochastic loading is computationally prohibitive with Monte Carlo simulation. CoNBONet proposes a surrogate combining DeepONet operator learning with Variable Spiking Neurons (VSNs) for sparse computation, Bayesian variational inference for uncertainty, and split conformal prediction for calibration. The goal is fast, energy-efficient inference with theoretical guarantees on reliability estimates.

The paper establishes that a single binary operator $\operatorname{eml}(x,y)=\exp(x)-\ln(y)$, together with the constant $1$, suffices to generate all elementary functions—trigonometric, exponential, logarithmic, and arithmetic operations. This provides a continuous analog to the Sheffer stroke in Boolean logic, enabling uniform binary-tree representations of mathematical expressions and opening avenues for gradient-based symbolic regression using identical computational nodes.

The paper proposes a non-parametric classifier based on the Nadaraya-Watson (NW) estimator that achieves linear $O(n)$ computational complexity while providing frequentist uncertainty bounds on predictions. By reformulating kernel regression for multi-class classification and deriving error bounds under Lipschitz continuity or separability assumptions, the authors bridge the gap between efficient "black box" methods and computationally expensive approaches like Gaussian Processes that offer formal guarantees. The method achieves $>96\%$ accuracy on MIT-BIH ECG data with uncertainty intervals that flag low-confidence predictions, making it suitable for safety-critical applications.

The paper tackles Off-Policy Evaluation (OPE) for ranking policies when the logging policy is deterministic—a common industrial scenario where existing estimators fail due to lack of common support. The key insight is to replace action-propensity weighting with click-propensity weighting, yielding the Click-based IPS (CIPS) estimator that leverages intrinsic user stochasticity even when the logging policy has none. This shifts the support requirement from ranking-wise or position-wise action overlap to click-wise overlap, enabling low-bias estimation in previously intractable deterministic settings.

This paper attacks the expensive problem of annotating NLP test sets by importing Active Testing (AT) from computer vision into language tasks. Given a labeling budget $B$, the goal is to select a subset $X_A$ that minimizes the estimation error $|M(X_F) - M(X_A)|$ between full and sampled test-set metrics, potentially cutting annotation costs by up to 95% while keeping prediction error under 1%. The core mechanism couples importance-weighted unbiased estimators with acquisition strategies (including a novel Agreement strategy based on attention-head disagreement) and an adaptive stopping criterion that removes the need to pre-specify the budget.

SecureBreak introduces a response-level safety dataset designed to detect harmful LLM outputs that bypass alignment mechanisms. Unlike existing benchmarks that classify prompts, this work focuses on binary classification of generated responses (safe vs. unsafe) across 3,059 samples from multiple model families including Llama, Qwen, Gemma, and Mistral. The core value proposition is providing a 'last-line defense' layer for post-generation filtering and supervisory signals to guide security re-alignment, addressing the growing threat of jailbreak attacks.

This paper introduces BOOST-RPF, which tackles power flow analysis in distribution grids by reformulating voltage prediction from a global graph regression task into a sequential path-based learning problem. The key insight is leveraging the radial (tree) topology of distribution networks to decompose them into root-to-leaf paths, then using XGBoost to predict local voltage drops between parent-child bus pairs. This approach aims to combine the speed of machine learning with the size-agnostic, recursive inductive bias of classical solvers like DistFlow.

The paper tackles Constrained Online Convex Optimization with Memory (COCO-M), where both losses and constraints depend on a window of past decisions, capturing realistic scenarios like smart-grid budgets and battery health limits. The authors propose the first algorithms achieving sublinear regret and cumulative constraint violation (CCV) under adversarial, time-varying constraints, both with and without unreliable predictions of future gradients. This work bridges the gap between classical constrained OCO and practical memory-dependent control problems.

This paper addresses personalized information retrieval for XML documents by representing users, queries, and documents as weighted concept vectors derived from a domain ontology. The core idea is a hierarchical weighting scheme that favors specific (deeper) ontology concepts combined with a dynamic profile update mechanism that reinforces concepts based on user interactions. The work targets the limitation of traditional keyword-based systems that return identical results regardless of user knowledge or preferences.

Interval uncertainty propagation typically requires solving expensive optimization problems for each input, making it infeasible for high-fidelity physics simulations. This paper proposes Direct Interval Propagation (DIP), reframing the task as interval-valued regression using neural surrogates to bypass optimization entirely. The authors extend DeepONet architectures to handle interval inputs and benchmark three distinct approaches---naive regression, bound propagation (IBP/CROWN), and interval neural networks---demonstrating orders-of-magnitude speedups on benchmark problems.

This paper investigates whether mechanistic interpretability findings from image-domain VAEs transfer to tabular data using 75 independent training runs across five architectures and four tabular benchmarks. It introduces posterior-calibrated Causal Effect Strength (CES) and Feature-Group Disentanglement (FGD) to compare circuit structures across modalities, finding that tabular VAEs exhibit ~50% lower modularity and that β-VAEs suffer catastrophic capacity collapse on heterogeneous tabular data (260× CES reduction) compared to images.

This paper argues that frictionless AI interfaces pose a systemic risk of "cognitive agency surrender"—the habitual abdication of human reasoning to algorithmic systems. Drawing on cognitive psychology, the authors theorize "Scaffolded Cognitive Friction" as a defense: intentionally injecting epistemic tension via Multi-Agent Systems (MAS) that expose structured disagreements (computational Devil's Advocates) to force System 2 activation. The work positions itself as a bridge between HCI, cognitive science, and AI governance.

Physics-informed neural operators enable rapid surrogate modeling of PDEs but incur substantial energy costs during repeated inference, limiting deployment on edge devices. This paper proposes SPINONet, which embeds Variable Spiking Neurons (VSNs) into the branch network of a separable DeepONet architecture to enable sparse, event-driven computation while preserving continuous coordinate pathways for derivative calculation. The core insight is that structural decoupling—spiking for input encoding and dense differentiability for coordinate encoding—allows physics-informed training without redundant multiply-accumulate operations.

The paper addresses the brittleness of in-context learning (ICL) to example ordering, an intractable $n!$ search problem. It proposes PLR, which reframes discrete permutation search as learning a Plackett-Luce distribution that concentrates probability mass on high-performing orderings. Using Gumbel perturb-and-sort for efficient sampling, PLR optimizes task-level metrics directly without requiring finite label spaces, extending naturally to open-ended reasoning tasks like mathematical problem solving.

Foundation models for Earth observation risk learning spurious correlations when pretraining with random masking. This paper proposes SpecTM (Spectral Targeted Masking), which deterministically masks pigment-sensitive spectral bands (phycocyanin, chlorophyll-a, red-edge) to enforce physics-based cross-spectral learning. Validated on microcystin concentration prediction using NASA PACE hyperspectral imagery over Lake Erie, the method achieves $R^2=0.695$ (current week) and $R^2=0.620$ (8-day-ahead), showing strong label efficiency but limited geographic validation.

This paper introduces TaigiSpeech, the first intent recognition dataset for Taiwanese Hokkien—a low-resource language spoken by 65% of Taiwanese elders. With 3,000+ utterances from 21 elderly speakers across emergency and smart-home scenarios, it addresses a critical gap in speech technology for aging populations. The authors also propose keyword-based and audio-visual mining strategies to bootstrap training data from unlabeled video sources.

Symbolic regression search spaces suffer from structural redundancy: expression DAGs with $k$ internal nodes admit $\Theta(k!)$ distinct node-numberings that encode the same mathematical expression. This paper proposes IsalSR, a representation framework that computes a pruned canonical string—a complete labeled-DAG isomorphism invariant—to collapse all equivalent forms into a single canonical representation. The approach promises to reduce effective search space size by $O(k!)$ and can be integrated into any existing SR algorithm as a preprocessing step.