Papers

Egocentric human data provide a principled source of supervision for learning dexterous robot manipulation. Unlike prior approaches that often collect such data in constrained or specially constructed environments, we collect in-the-wild egocentric demonstrations in real-world settings, including homes, factories, and pharmacies, etc., where people perform their ordinary tasks while wearing head-mounted cameras. This collection protocol captures diverse workflows and hand-object interactions across long-tailed object and skill distributions, but also yields visually challenging observations due to scene clutter and head-motion-induced viewpoint changes (a mean cumulative rotation of $15.93^{\circ}$/s). To address these issues, we introduce EgoWild2Dex, which transfers in-the-wild ego-human experience to dual-arm robots with dexterous hands by jointly aligning unstable egocentric views and human motions with robot observations and actions, respectively. This work offers three benefits. First, we introduce GeoFormer, a differentiable geometric transformer that warps noisy human observations toward robot observations. Second, we design a human-robot training scheme to bridge the embodiment gap, enabling high task success with limited robot supervision. Third, we release EgoWild, a 538.9-hour in-the-wild egocentric human dataset comprising 179,049 episodes, 125,961 unique task descriptions, and 1,282 object categories. On real robots, EgoWild2Dex achieves an average success rate of 96.7% across three long-horizon bimanual dexterous manipulation tasks and an average object-level zero-shot success rate of 33.3%. The data, models, and code will be released.

High-resolution in-beam ${\gamma}$-ray spectroscopy of $^{53}$Ca was performed at the Radioactive Isotope Beam Factory (RIBF) at RIKEN. Excited states in $^{53}$Ca were populated via one-neutron and one-proton removal reactions from a $^{55}$Sc radioactive beam, and the emitted ${\gamma}$ rays were detected using the HiCARI hybrid germanium detector array. The lifetime of the first excited $5/2^{-}$ state in $^{53}$Ca was measured for the first time through an analysis of Doppler-roadened ${\gamma}$-ray line shapes arising from relativistic recoil velocities. A lifetime of ${\tau} = 11.3_{-11.3}^{+5.3(stat)+3(sys)}$ ps was deduced, corresponding to a reduced transition probability of $B(E2;5/2^{-}{\rightarrow}1/2^{-}) > 2.6 e^{2}fm^{4}$ at the 1${\sigma}$ level, with a best-fit value of $4.5 e^{2}fm^{4}$. The measured transition strength is compared with state-of-the-art shell-model and ab initio calculations, providing a stringent benchmark for shell-model Hamiltonians and effective charges and modern ab initio descriptions of the neutron-rich calcium isotopes approaching the potential $N$ = 40 shell closure.

Generative world models aim to predict future states conditioned on actions, where action controllability is fundamental for reliable dynamics modeling. While recent efforts leverage simulator-generated data to enhance this capability, existing training pipelines face two fundamental limitations. First, static offline data collection leads to a distribution misalignment between training sets and the model's evolving error patterns, failing to resolve critical long-tail scenarios where dynamics predictions remain unreliable. Second, the standard objective of minimizing observational discrepancy often encourages the model to exploit spurious correlations instead of capturing the underlying action-effect causality. To address these limitations, we propose OnlineWM, an online training framework that continuously improves world modeling through active simulator interaction and causality-aware optimization. OnlineWM introduces two key innovations: (1) Active Online Learning: Instead of using fixed datasets, OnlineWM adaptively queries the simulator for new interaction sequences that target the model's current predictive weaknesses, ensuring high-utility data acquisition. (2) Causality-Aware Fine-Tuning: We propose a counterfactual learning strategy that contrasts the outcomes of different actions from identical states, forcing the model to attribute state transitions to specific actions rather than ambient environmental evolution, thereby grounding its predictions in reliable causal mechanisms. By integrating active data acquisition with causal optimization, OnlineWM establishes a closed-loop refinement process that ensures the model is both robust to diverse scenarios and precise in its causal attribution. Extensive experiments demonstrate that OnlineWM significantly enhances action controllability and generalizes effectively to unseen domains.

We study tomography for quantum channels with input dimension $d_1$, output dimension $d_2$, and Kraus rank at most $r$, to within diamond norm error $\varepsilon$, using adaptive experiments that retain no quantum memory between channel queries. - For quantum channels whose non-zero Choi eigenvalues are bounded below by $\Omega(d_1/r)$, we establish optimal query upper and lower bounds $\Theta(d_1d_2r^2/\epsilon^2)$. The upper bound is achieved by a nonadaptive algorithm that uses the estimator from [Surawy-Stepney et al., Quantum (2022)], together with a new diamond-norm analysis. The lower bound applies to arbitrary adaptive incoherent protocols and follows from a new local family of channels and a uniform one-query Fisher-information bound. - For general channels, we establish an upper bound $O(d_1d_2r^2\log(2d_1)/\epsilon^2)$, nearly matching the above lower bound $\Omega(d_1d_2r^2/\varepsilon^2)$. To achieve this, we generalize the above nonadaptive algorithm by adapting the input state over $O(\log(2d_1))$ rounds with the Matrix Multiplicative Weight Update algorithm.

Lagrangian duality is fundamental to optimization, from the development of solver techniques to the interpretation of solutions; yet its geometric intuition can be difficult to grasp through mathematical derivation alone. This tutorial explains duality and its implications in a highly visual way for a broad audience of students and practitioners. It develops the motivation for the Lagrangian, the two-player game interpretation, and the connection to Fenchel duality. Throughout, it emphasizes the geometric meaning of the solution from both the primal and dual

We construct a static and spherically symmetric black hole (BH) spacetime surrounded by a depleted Dehnen dark matter (DM) halo. The model describes a final equilibrium configuration rather than dynamical accretion process: the initial halo mass inside a prescribed inner edge is incorporated into the central BH, leaving a vacuum gap between the dressed horizon and the surviving halo. Consequently, the horizon radius depends on the Dehnen cusp index $\gamma$ and the absorbed halo mass, while the Schwarzschild result is recovered exactly when the halo is removed. We derive general expressions for the metric functions and redshift phase and obtain explicit solutions for $\gamma=0,1,2,$ and $5/2$. The dominant energy condition imposes the universal local bound $\kappa\geq5/4$, whereas global admissibility depends additionally on the cusp index and halo compactness. A literal stable Einstein-cluster interpretation requires the stronger condition $\kappa\geq3$ when stable circular motion is imposed at the halo edge. We also analyze the photon sphere, shadow, and admissible parameter space, demonstrating how the halo cusp and depletion scale affect strong-field observables. Previously studied analytic BH–halo geometries and the Schwarzschild and weak-halo limits follow as special cases of the general construction.

We develop a theory of equivariant local systems which categorifies genuine equivariant cohomology theories – such as Atiyah–Segal $K$-theory, Lurie's tempered cohomology, and the equivariant elliptic cohomology of Grojnowski, Greenlees, and Gepner–Meier – analogously to how the category of ordinary local systems categorifies singular cohomology. More precisely, we introduce, for a global space $X$ and a coefficient system $\mathscr{A} \colon \mathrm{Orb}^{\mathrm{op}} \to \mathrm{Pr}^{\mathrm{L}}$, the categories $\mathrm{LS}^{\mathrm{glo}}(X,\mathscr{A})$ and $\mathrm{LS}^{\mathrm{gen}}(X,\mathscr{A})$ of globally equivariant and genuine local systems, the latter generalizing the genuine stable category $\mathrm{Sp}^G$ of a compact Lie group $G$ to non-constant coefficients. When the coefficients come from an oriented abelian group stack $A$ over a locally complex periodic base, we also construct the category $\mathrm{LS}^\mathrm{temp}(X,A)$ of tempered local systems, extending Lurie's theory beyond finite groups; the defining condition is justified by a tempered form of the Atiyah–Segal completion theorem. We show that global sections induce an equivalence $\mathrm{LS}^{\mathrm{temp}}(BU(1),A) \simeq \mathrm{QCoh}(A)$ and we prove that $\mathrm{LS}^\mathrm{temp}(X,A)$ is a smashing localization of $\mathrm{LS}^\mathrm{gen}(X,A)$ if $X$ is an orbispace.

We investigate the phase-resolved X-ray spectral evolution of the high-mass X-ray binary Cen X-3 across eclipse ingress and egress, using two out-of-eclipse intervals as references, to constrain the absorbing material and Fe K emitting regions with high-resolution XRISM/Resolve spectroscopy. We model the spectra with SPEX, including interstellar and local wind absorption, a power-law and blackbody continuum, disk reflection, and Gaussian components for Fe XXV and Fe XXVI. We compare the orbital variation of the absorbing column with a smooth CAK wind model and investigate the variability of the Fe K alpha line width. The local absorbing column varies by nearly two orders of magnitude, from N_H^local = 4.25 x 10^22 cm^-2 during pre-ingress to 5.9 x 10^23 cm^-2 at ingress. The excess ingress absorption cannot be reproduced by a smooth wind model alone, suggesting a localized overdense structure along the line of sight. The Fe K alpha width varies from approximately 510 km s^-1 during pre-ingress to 1400 km s^-1 during egress, with intermediate, statistically indistinguishable values of approximately 900 km s^-1 during ingress and post-egress. Highly ionized Fe XXV and Fe XXVI emission is detected throughout the orbit, although fewer components are resolved during ingress. These diagnostics reveal a complex circumstellar environment with localized dense material and multiple reprocessing regions. The results demonstrate the capability of XRISM/Resolve to probe the geometry and kinematics of accretion environments in eclipsing high-mass X-ray binaries.

Estimating regression models with nonlinear predictor transformations is challenging when data are missing, because nonlinearity typically renders the conditional distribution of the missing values intractable. Previous methods require specific nonlinear forms, such as polynomials or interactions, or rely on approximations that can induce bias. We propose a stochastic EM algorithm that uses sampling-importance resampling (SIR-StEM) to handle missing data under arbitrary nonlinear transformations of predictors, either substantively motivated or incidental, such as spline basis expansions. Unlike approaches that require linearity or closed-form conditionals, SIR-StEM only requires evaluating the complete-data likelihood up to proportionality, making it applicable across a broad class of nonlinear regression models. We construct an algorithm that makes use of missing data pattern information for computational efficiency, and establish asymptotic normality of the estimator. We demonstrate the method in two simulation studies, one with parametric nonlinear transformations and another with a spline basis expansion based on real behavioral measures. Results show that SIR-StEM yields low bias and near-nominal confidence interval coverage, outperforming other common approaches. We conclude with limitations and directions for future research.

Convex spiral equicoverings were recently introduced by Espinosa-García, Martínez-Sandoval and Roldán-Pensado. They left open the question of whether every planar mass admits a convex $(3k,k+1)$-spiral equicovering. In this paper we give an affirmative answer to this question. To be precise, we prove the following: Given an integer $k \ge 2$, for every planar mass there is a fan consisting of $3k$ equal-mass sectors such that the union of every $k+1$ consecutive sectors is convex.

Autonomous multi-drone navigation requires fleets to service distributed objectives in cluttered environments under tight computational budgets. Efficient coordination depends on task bundling, where each drone visits multiple objectives along its route. Separate solvers for cost estimation, assignment, and execution incur redundant graph search and produce long, abrupt paths. We propose FlockDiffusion, a learned framework combining a scene graph encoder, an explicit allocation head, an assignment conditioned diffusion transformer, and a closed form trajectory decoder. An autoregressive teacher provides offline supervision for parallel fleet trajectory generation. PyBullet ablations show that bundling increases task completion from 50% to 100%, while our complete teacher further reduces route cost by 8.4% relative to MAGNNET with bundling. In the optimized scalability benchmark, evaluated on 100 scenes per density with ten drones, FlockDiffusion achieves 6.2 to 7.6 times faster inference and approximately 37% shorter routes than the classical pipeline. As nominal task counts increase from 20 to 40, latency rises from 7.8 to 11.1 ms, compared with 48.0 to 75.8 ms for the baseline. In a separate evaluation across five Gazebo environments, FlockDiffusion achieves 100% planner coverage and reduces planned route cost by 15.4% relative to the baseline with bundling. These results demonstrate efficient planning under increasing task density in configurations that are demanding to reproduce with physical drone fleets.

Organizations invest heavily in internal mechanisms such as incentive systems, governance structures, performance measurement, analytics, and repeated reorganizations. Yet realized performance gains are often weak or short-lived, particularly in large and mature organizations. Standard explanations emphasize allocative inefficiency arising from incentive misalignment, information problems, or bounded rationality. While relevant, these explanations struggle to account for environments in which effort is high, optimization capabilities are sophisticated, and aggregate performance remains far below apparent potential. This paper develops a theory of structural inefficiency inside organizations. The central mechanism is the presence of internal externalities: unpriced cross-effects of organizational mechanisms acting simultaneously on multiple performance objectives. When such effects are mixed in sign, organizational effort is partially canceled internally, creating a structural ceiling on realized performance. The paper introduces a simple representation of organizational primitives and key performance indicators and derives a realized-performance law in which internal conflict enters as a multiplicative discount factor. Under weak and realistic conditions, structural inefficiency dominates allocative inefficiency, explaining why optimization often fails before organizational coherence is restored.

On the real line, Bárány, Kolossváry, and Troscheit [5] made exponential separation of analytic self-conformal systems checkable through a dual iterated function system on a space of analytic functions. The quantity their condition separates is the pre-Schwarzian derivative, and this observation lets us carry the construction to conformal iterated function systems in every dimension, where the pre-Schwarzian is $T_f=\nabla\log\|Df\|$. A pointwise separation of the pre-Schwarzian cocycle implies the strong exponential separation condition, and in dimensions one and two separation of the Schwarzian cocycle implies the same condition modulo Möbius maps. Each hypothesis is equivalent to a uniform gap condition, and these are $\mathcal{C}^2$-open. On the line they are also dense, so both separation conditions are $\mathcal{C}^2$-generic there; this sharpens the genericity theorem of Bárány, Kolossváry, and Troscheit, whose open and dense set carries only the plain condition. In the plane they are dense on Jordan domains with simply connected extension domains, and both separation conditions are $\mathcal{C}^2$-generic there too. In dimensions at least three every conformal map is Möbius, so no system satisfies the condition modulo Möbius maps, and the pre-Schwarzian hypothesis becomes a pole-separation condition, dense when the generators contract strongly enough and stay away from the similarities, but not dense in general. On Jordan domains, the planar genericity and a theorem of Feng and Rapaport [13] settle the dimension drop conjecture for $\mathcal{C}^2$-generic planar systems with injective generators.

Small open-source large language models (LLMs) in the 0.6B-4B parameter range are increasingly deployed for structured output generation (JSON, function calling, data extraction), yet little is known about how constrained decoding (CD) interacts with model scale in this regime. We benchmark five models from three families across 14 structured-output tasks under three decoding conditions (native, Outlines, XGrammar). We introduce a two-axis evaluation that separates structural correctness (schema validity) from semantic correctness (content accuracy). We find that CD eliminates all structural failures across all models (schema validity: 78.6-92.9% to 100%), but content accuracy reveals a persistent semantic gap that is scale-dependent: type coercion failures are fully CD-rescuable, while instruction-semantic failures (e.g., multi-step function calling) remain CD-resistant. Schema conformance is necessary but not sufficient for semantic correctness; CD's reach ends exactly where schema conformance ends.

The ages and masses of star clusters may leave systematic signatures in the second-order spatial structure of semi-resolved photometric images. To test this hypothesis, we analyze the images using the Hessian matrix and identify basic luminous structures, termed peak cores. Applying this analysis to six-band PHAT/AP images of 1249 M31 clusters with CMD-based ages of 10–300 Myr and masses of $10^{2.2}$–$10^{4.5}\,M_{\odot}$, we examine the correlations of age and mass with the peak-core structural parameters, $C_S$ and $Q$. Younger clusters show larger $C_S$ values in the ultraviolet bands, with the strongest age–$C_S$ anti-correlation in F336W ($\rho=-0.61$), whereas more massive clusters show smaller $Q$ values, most clearly in F336W ($\rho=-0.44$). These correlations suggest that the contrast of UV-bright peak cores decreases with age, whereas increasing mass alters the abundance and spatial organization of detectable structures across different bands.

We study the full angular dependence of the linearized Schrödinger form of the Calderón problem on the unit disk. We obtain an explicit continuous spectral diagonalization of its normal operator and hence a generalized singular system for the linearized forward map. To our knowledge, this is the first such diagonalization for the full angularly dependent, multiplicity-normalized linearized measurement map on the disk.

We study a contextual distributionally robust chance-constrained programming (C-DRCCP) model that controls violation probabilities conditional on a neighborhood of a target context. The ambiguity set is defined over the joint distribution of contexts and outcomes using the Wasserstein distance. For affine safety systems with right-hand-side uncertainty, we derive an exact mixed-integer programming (MIP) reformulation. The reformulation reveals that worst-case conditional risk is determined by two coupled uses of the transportation budget: moving empirical mass into the conditioning neighborhood and moving conditioned mass into the failure region. Equivalently, robust feasibility holds when the minimum transportation cost needed to reach the conditional risk boundary is at least the available budget. Calibrating the ambiguity set gives a finite-sample guarantee for the true conditional violation probability. This characterization leads to strong valid inequalities that exploit the contextual transport. Each contextual mass allocation determines both the conditional scenario weights and the budget available to move outcomes into failure, whereas conventional quantile cuts utilize only the induced weights. Accounting for the remaining budget generates strengthened quantile cuts that strictly dominate standard quantile bounds and yield tighter MIP coefficients. We also derive rank inequalities from an exchange property of contextual transportation costs, linking feasible failure patterns across contextual allocations. Extensive numerical experiments demonstrate that these inequalities close much of the linear-relaxation gap, speed up the solution process, and enable large-sample instances that are intractable for the baseline formulation.

In this contribution, a fluid flow problem on a general deforming domain for a Newtonian fluid in two and three space dimensions with artificial compressibility approximation is studied. We prove an estimate on the integral equicontinuity in time of the weak solutions under suitable domain regularity assumptions, which is independent on the compressibility parameter and serves as an alternative compactness argument for the convergence of weak solution sequences for vanishing compressibility. The corresponding estimate is obtained by remapping the problem onto a fixed reference domain and using appropriate divergence-preserving testfunctions involving the difference of two solutions at different points in time, thus defined with respect to different domains/coordinates.

This study examines the impact of processing conditions, such as thermalprocessing, on the resulting microstructure and mechanical properties. Spe-cial emphasis is placed on microstructural features obtained from three-dimensional phase-field simulations, which provide detailed insights into bai-nite morphology, phase distribution and retained austenite content. Thesesimulated microstructures are correlated with changes in yield strength undermultiaxial load, as represented in the yield surface of the material. The re-sults demonstrate that optimized processing routes can refine the microstruc-ture, enhance mechanical properties and significantly alter the yield surfacecharacteristics. These findings provide valuable insights for the design andapplication of bainitic steels, as well as for the development of predictivemodels linking process-structure-property relationships.

This paper deals with the historical development and physical mechanisms of pyroelectricity, a phe-nomenon whose roots date back over 2000 years to ancient times. This paper is aimed at students and engineers as a concise introduction to the subject area. While the attractive effect of heated tourma-line was already described by Theophrastus, scientific systematization did not occur until the 18th century by researchers, such as Aepinus and Canton, who identified the effect of electrical polarization resulting from temperature changes. This essay highlights the path from early analogies to magnetism to the modern crystallographic description by Haüy and Thomson. In the physics section, pyroelectricity is defined as the temperature dependence of the spontaneous polarization in anisotropic solids. At the microscopic level, the permanent dipole moment of an elementary cell is described by the vector sum of individual moments. It is mathematically demonstrated that the macroscopic spontaneous polarization correlates with the surface charge density and is linked to the temperature change via the pyroelectric cofficient. A distinction was made between the primary pyroelectric effect and the secondary effect resulting from the thermal deformation of the crystal. Finally, the renaissance of this field of research through the development of modern infrared detectors and ferroelectric materials in the 20th century was highlighted.

Scientific agents support a range of literature-based research tasks, such as retrieval, question answering, evidence-grounded generation, and claim assessment. Most existing systems, however, are organized around individual tasks: the same papers are repeatedly retrieved, segmented, and interpreted, and the understanding built in one task is difficult to reuse in the next. We present ScholarStack, a layered research asset framework that compiles a paper collection into reusable, versioned, and provenance-preserving assets at three complementary levels: source-grounded paper-level statements, domain-level organization, and evidence-grounded cross-paper syntheses. A common access interface returns task-specific views at the evidence granularity each task requires, preserving study conditions, source traceability, and verification status. We instantiate the framework on four task families spanning ten task settings, comparing agents that use the compiled assets with task-specific baselines under matched base models. Quality gains concentrate on tasks that require cross-paper evidence, such as multi-paper question answering and literature review generation, and query-time token cost falls on every task where it is measured, with assets compiled once and reused across tasks. These results suggest that layered research assets can serve as shared infrastructure for scientific agents, shifting literature-based assistance from isolated document processing toward cumulative, evidence-grounded workflows.

Two-round legislative elections are decided not only by first-round vote shares, but by how voters whose preferred party did not advance redistribute their votes among the surviving candidates. We develop a principled model of this transfer process, grounded in multi-dimensional ideological embeddings derived from the Chapel Hill Expert Survey, and apply it to French legislative elections. Calibrated on the 2017 and 2022 elections, the model is evaluated on the unusually complex 2024 contest, in which three ideologically distinct blocs reached the second round simultaneously, achieving 90.34 percent constituency-level accuracy. We show that ideological proximity alone explains the large majority of vote transfers, that a simple left-right axis is insufficient to capture the relevant distances, and that the structural configuration of the 2024 electorate was far less favorable to the far right than pre-election forecasts suggested.

Feed-forward visual geometry models such as the Visual Geometry Grounded Transformer (VGGT) have recently enabled direct 3D reconstruction from multi-view images. Despite their promising performance, these models scale quadratically with the number of input views due to their global attention mechanism, resulting in substantial latency for long sequence inputs. There have been some recent efforts to accelerate VGGT, but they primarily focus on reducing token redundancy through token merging or key/value sparsification. Our work resolves this bottleneck from a different perspective by investigating architectural redundancy in visual geometry transformers. We show that the multi-head attention modules in VGGT's global-attention layers contain substantial architectural redundancy, with only a subset of heads carrying critical geometric information. In light of this observation, we propose VGGT-Prime, a compute-adaptive mixture-of-heads model that resolves this redundancy to accelerate visual geometry transformers while maintaining competitive reconstruction quality. The key idea of VGGT-Prime is to estimate the appropriate computation level for each global-attention head using a lightweight router and then dynamically assign each head to different computation modes. Extensive experiments on multiple datasets demonstrate that VGGT-Prime can achieve an {$8\times$} inference speedup over VGGT while maintaining competitive performance on camera pose, depth, and point-cloud predictions. We further show that VGGT-Prime is complementary to existing acceleration methods, such as token merging, further improving inference speed by up to $14{\times}$ over VGGT. An overview of our work is available on our \href{https://vggt-prime.github.io}{project page}.

Accurate prediction of ionospheric Total Electron Content (TEC) during geomagnetically disturbed conditions remains challenging, particularly when empirical models perform poorly during storms and neural network (NN) approaches rely explicitly on geomagnetic indices such as Kp/Ap or Dst/SYM-H. In this study, we develop an NN-based model to predict TEC variations over the Indian longitude sector without incorporating geomagnetic indices as inputs. The model is trained on the full-year (2024) Global Ionospheric Map (GIM) TEC dataset at a 15-minute cadence, together with solar wind parameters: Interplanetary Magnetic Field (IMF) $B_z$, velocity ($V_{sw}$), and density ($N_p$), along with the F10.7 cm solar radio flux. Seasonal and diurnal variability are represented using sinusoidal transformations of Day of Year (DOY) and hour of day (HOD). An optimized architecture with 96 hidden neurons yields strong training performance ($R^2$ $\sim$ 0.96, RMSE $\sim$ 5.4, and MAE $\sim$ 3.8 TECU). Independent predictions for the year 2025 demonstrate robust generalization under both quiet and geomagnetically active conditions. Quiet-time residuals are generally confined within $\pm$10 TECU, while during moderate and strong storm-time conditions, the model maintains comparable error bounds across most of the Indian region, despite not being trained on geomagnetic indices. The results indicate that solar wind parameters, combined with the F10.7 solar flux and cyclical temporal encoding, contain sufficient information to reproduce geomagnetic storm-time TEC variability. This study therefore demonstrates the applicability of the proposed approach for regional TEC prediction and ionospheric nowcasting, with potential applications in GNSS positioning, satellite navigation, radio communication systems, and the future development of operational ionospheric forecasting tools.

Robotic systems are typically composed of multiple independently developed modules that work together to perceive, predict, and act in the environment. Although each module may perform reliably in isolation, composing them does not necessarily preserve uncertainty calibration at the system level. In this work, we show that well-calibrated component interfaces do not necessarily produce calibrated downstream behavior after composition. Using a moving-obstacle prediction pipeline, we demonstrate that position and velocity estimators can each appear well calibrated individually, yet differences in how their error are correlated lead to substantially different estimates of future-state uncertainty. Consequently, assuming independence can make the system either overly confident or unnecessarily conservative, directly influencing downstream planning decisions and safety. Through simulations, we show that modeling the joint covariance restores downstream calibration and improves system performance, whereas dependence-robust uncertainty bounds enhance safety at the cost of increased conservatism. Our findings reveal a fundamental limitation of independently validating robotic modules and highlight the need for interfaces that communicate dependence information or support direct system-level calibration.

We establish global well-posedness and a small-noise large deviation principle for the two-dimensional Dean-Kawasaki equation with Coulomb interaction $\partial_t\rho^\varepsilon=\Delta\rho^\varepsilon-\nabla\cdot(\rho^\varepsilon(V*\rho^\varepsilon))-\sqrt{\varepsilon}\nabla\cdot(\sqrt{\rho^\varepsilon}\circ\xi^{K(\varepsilon)})$, where $V=\lambda_V(\cos(\alpha_V)\nabla G+\sin(\alpha_V)J\nabla G)$. Here $\lambda_V\geq0$ is the interaction strength, $\alpha_V$ is the interaction angle, $G$ is the mean-zero Green function of $-\Delta$ on $\mathbb{T}^2$, $J$ is rotation by $\pi/2$, and $\xi^K$ is a finite-mode approximation of space-time white noise. For finite-entropy initial data of mass $M$ satisfying $\lambda_V\max\{\cos(\alpha_V),0\}M<8\pi$, the equation with the exact square-root coefficient is globally pathwise well posed at every finite Fourier cutoff in the class of stochastic kinetic solutions. Following Fehrman and Gess (arXiv:1910.11860), we prove a large deviation principle on $L^1((0,T)\times\mathbb{T}^2)$ under the scaling $K(\varepsilon)\to\infty$ and $\varepsilon K(\varepsilon)^4\to0$, with good rate function given by the quadratic control problem for the skeleton equation.

The rapid maturation of zero-shot Text-to-Speech (TTS) models has turned high-quality voice cloning into a widely available capability, raising acute concerns over unauthorised replication, fine-tuning and resale of proprietary speech models. Yet ownership verification for TTS remains largely open: speech is a continuous waveform whose perturbations are easily destroyed by routine signal processing, and the human auditory system imposes a much tighter perceptual budget than vision. We present TTS-Guard, a black-box ownership verification framework for TTS models built on adversarial speaker-pair fingerprints. TTS-Guard(i) selects key speaker pairs in a dual embedding space for architecture-agnostic stealth;(ii) optimises a perturbation through an adaptive curriculum of shadow models covering fine-tuning, pruning, quantisation and distillation; and (iii) aggregates black-box queries into a calibrated Verification Confidence Score. On five mainstream TTS systems, TTS-Guard reaches an average Fingerprint Success Rate of $96.4\%$ at a False Positive Rate of $5.8\%$, while preserving intelligibility and naturalness. The fingerprint remains effective against ten audio attacks, six model modifications, and two state-of-the-art adversarial purifiers.

A central problem in nonlinear elasto-plasticity theory is to formulate a geometrically nonlinear, large-strain model of elasto-plasticity in single crystals in which plastic flow is driven directly by the motion of dislocations and which allows for a homogenization procedure from discrete dislocation lines to dislocation densities. In this work, such a model is introduced based on two hypotheses: (1) The Small-Distortion Hypothesis posits that the total plastic distortion may be arbitrarily large but its dominant part admits a representation as a gradient, meaning that the distorting effect of dislocations is small relative to the specimen size. Concretely, this hypothesis is realized here through a "multiplicative Helmholtz-type decomposition", which splits an arbitrary matrix field into the product of a gradient and a matrix field with controlled curl. (2) The Line-Tension Approximation treats all dislocations as infinitesimally thin lines whose local stress fields are individually negligible compared to specimen-scale stresses and only matter in the aggregate. The resulting model, termed the Small-Distortion Geometric Model, furthermore employs the so-called Space-Time Framework, which furnishes a geometric language to precisely describe the advection of dislocation lines. As such, it provides a physically grounded account of crystal plasticity in an idealized mesoscopic setting that bridges dislocation mechanics and continuum elasto-plasticity.

Score-level fusion based on logistic regression is widely used in speaker recognition to combine complementary systems. However, conventional approaches assign fixed system-dependent coefficients and do not explicitly account for variations in the reliability of individual enrollment and test utterances. Drawing on recent research on the entropy of deep learning-based speaker recognition models, this study incorporates an uncertainty component into the fusion process. By exploiting both system-level complementarity and utterance-dependent uncertainty, the method achieves robust performance in large-scale speaker recognition tasks that involve highly variable characteristics of the speech signal. These results demonstrate that model-entropy information provides a valuable complementary cue in large-scale scenarios.

Large language models have shown strong performance across a range of legal tasks, but existing benchmarks rarely evaluate the ability to take and defend a legal position, reason under incomplete information, or synthesize multiple statutory provisions. This gap is particularly pronounced for Canadian law, which remains underrepresented in legal NLP. We introduce GRACE (Grounded Reasoning Adversarial Canadian LEgal examples), a dataset of 1,915 question-reasoning-answer instances grounded in Canadian federal legislation. GRACE covers three reasoning modes: adversarial advocacy, uncertainty, and applied reasoning. We develop a pipeline that partitions raw statutory text, generates scenario-based questions and reasoning, and filters examples through model-free citation verification and LLM-based quality auditing. As a proof of concept, we fine-tune CLeAR-4B (Canadian Legal Adversarial Reasoning), a lightweight model for grounded legal reasoning, and evaluate it against the unmodified Qwen3-4B base model in open- and closed-book settings. CLeAR-4B substantially improves agreement with teacher outputs and statutory citation behavior when the relevant act text is provided, while its grounding degrades sharply when the statute is withheld. These results suggest that GRACE can support the development of lightweight legal models that reason more effectively from supplied statutory text.