Papers

We present a proof-of-concept study of neural ratio estimation (NRE) - a simulation-based inference algorithm - applied to WIMP direct detection using the XENONnT experiment as an example. A binary classifier is trained on simulated parameter-observation pairs to approximate the likelihood ratio for the WIMP mass $m_\chi$ and spin-independent proton coupling $c_p$, using realistic scintillation and ionization signals produced by the fuse detector simulation. We demonstrate that astrophysical uncertainties from halo model variations are marginalized over implicitly and at no additional inference cost by including multiple halo models in the training data. Using extreme $\pm5\sigma$ Standard Halo Model variations, we show that training on a combined dataset recovers well-calibrated posteriors even under severe misspecification, while more realistic halo models produce nearly indistinguishable results. The signal-plus-background classifier yields an exclusion limit that is both qualitatively and (after accounting for differences between frequentist and Bayesian approaches) quantitatively consistent with the XENONnT expected sensitivity. These results establish NRE as a scalable complement to the profile likelihood ratio approach, with particular advantages as experiments probe deeper into the neutrino fog.

We propose a parameter-reduced version of magnetic Moment Tensor Potential (mMTP) based on various matrix and tensor decompositions. The resulting compressed magnetic machine-learning potential reduces the number of parameters by a factor of approximately 1.5-3 without compromising predictive performance on the validation set. We evaluate the performance of the compressed potentials for magnetic, structural, and vibrational properties, as well as in molecular dynamics simulations of Fe-Al and CrN systems. We demonstrate that the simulation results obtained with the compressed mMTP are numerically consistent with those of the original uncompressed model and are in agreement with density functional theory calculations and experimental data.

We construct two superconducting pairing models on doped Kitaev honeycomb, whose microscopic kernels are derived rather than assumed. Both are driven by a bond-local glue neutral between the even- and odd-parity pairing channels; the chirality–parity linkage is set by bond-star geometry, not by the microscopic glue. (i) Model~1(odd-parity sector): the kernel follows from the flux-sector selection rule of the pure model, fixing its component-diagonal, nearest-neighbor separable, rank-one-per-bond form, with $w_\alpha=J_K^2\chi_\alpha\propto r_\alpha m_\alpha$, where $m_\alpha$ follows from the Hellmann–Feynman theorem and $r_\alpha\simeq0.24$ is computed numerically. The spin-vertex state is a unitary triplet with a bond-locked anisotropic $\mathbf d$ vector; its gap is a Gram sum of bond sines, so nodal directions are fixed by geometry. (ii) Model~2(even): the kernel is fixed by the $C_3$ little-group selection rule of a zone-corner chiral phonon, up to a Cooper-channel extension whose momentum constraint we state explicitly. The charge-vertex pairing is a $d+id$ singlet: the parity decomposition $f_\pm=g_\pm+p_\pm$ projects onto $g_\pm$, so chirality is slaved to the phonon polarization, with $C=\pm4$ and $\kappa_{xy}/T=2\kappa_0$. Reversing the pump helicity reverses the Chern number, Kerr rotation, and thermal Hall sign. (iii) General theorem: on odd-fold bond stars, the angular-momentum ladder of a single-star chiral form factor alternates in parity. Every chiral state selected by such a bond-local form factor is a geometrically enforced singlet–triplet composite with universal amplitude ratios ($n=3$: $d\oplus p$; $n=5$: $g\oplus p$ or $d\oplus f$); coherent doubled-star combinations can restore parity purity. The Chern number of the mixed state follows the dominant on-shell component across a topological boundary. The spin glue drives $C=\pm2$; the charge kernel $C=\pm4$.

A representation family is a distinct way of extracting features from time series. Ensemble algorithms that combine several representation families remain the most accurate approach to time series classification. Current state-of-the-art ensembles, most notably HIVE-COTE~2.0, pair a bespoke classification algorithm with each representation family and combine their predictions using a fixed, non-adaptive rule. We present TIGER (Time-series classification with In-context-learning Gated Ensemble of Representations), which instead applies the same small portfolio of three general-purpose classifiers (Ridge, Extra Trees, and Naive Bayes) to four representations from four distinct families, stacking the resulting twelve base learners' predictions into a meta-feature matrix. The final prediction is produced by an adaptive meta-classification rule that chooses, independently for each data set, between a weighted hard majority vote and TabICLv2, a pretrained tabular foundation model used in-context as a meta-classifier, based on the mean number of training samples available per class. On a 142-data-set benchmark drawn from the UCR time series classification archive, TIGER obtains the best mean accuracy, balanced accuracy, and F1-score among six compared algorithms, including HIVE-COTE~2.0, and significantly outperforms each of the other five individually. TIGER's adaptive rule also meaningfully outperforms either of its two constituent meta-classification methods used alone, and its single hyperparameter, tuned using only a twenty-data-set development subset, is shown to generalize to the full evaluation benchmark. We further characterize TIGER's design through an extensive set of ablation experiments and report the design alternatives that we investigated and ultimately discarded.

The design of high-quality nanoparticles requires the production of materials with narrow particle size distributions, as these strongly influence the functionalities of the final product. To achieve this goal, it is useful to consider a coupled process chain in which nanoparticles are first synthesized and subsequently fractionated by chromatographic separation. We introduce a novel mathematical optimization model for the integrated optimization of coupled nanoparticle synthesis and separation processes. The proposed model simultaneously accounts for process performance and economic objectives through a cost function that incorporates both production costs and achievable profit defined by the particle quality. For efficient evaluation of the synthesis model, we employ the exact method of moments. The resulting non-convex process-design problem is solved numerically using local optimization. Optimization results of our approach are demonstrated for the prototypical design of gold nanoparticles. The computational results show that optimized solutions of the coupled synthesis–separation process are capable of accurately representing both process stages while achieving optimized particle quality. The proposed model highlights the potential of mathematical optimization for the design of coupled nanoparticle production process-chains.

A relaxed maximum-based normal $S$-iteration method (RMNSI) is proposed for solving generalized absolute value equations (GAVEs). The method combines a maximum-based fixed-point formulation with a constant relaxation parameter and avoids the selection of an auxiliary matrix. Assuming that $A+B$ is non-singular, both stages require linear systems with the same coefficient matrix $A+B$, allowing a single factorization to be reused throughout the iteration. A single spectral condition is established that guarantees unique solvability of the GAVE and global convergence from an arbitrary initial vector. An admissible range of the relaxation parameter is characterized, and $R$-linear convergence and error estimates are derived. Since the global condition can be conservative, a local convergence analysis based on the solution sign pattern is also presented. Numerical experiments on complementarity-derived, dense mixed-sign, and asymmetric ridge-regression problems are presented, and comparisons with several recent methods demonstrate the competitive efficiency and accuracy of RMNSI. The results further indicate that the preferred relaxation parameter depends mainly on the matrix structure and diagonal shift, while its dependence on the problem dimension is generally weak.

Expressive humanoid interaction requires speech, facial animation, and body gestures to form a coherent response. However, many full-body humanoid robots produce speech and gestures without a visually expressive face, while talking-face animation and robot gesture generation are typically developed separately. We present Talk, Render, Act (TRABot), an agent-based framework comprising specialized agents for motion-atom construction, dialogue generation, motion planning, and facial animation. First, to produce natural and semantically meaningful gestures, we construct Robot-Ready Semantic Motion Atoms by segmenting long-form, G1-retargeted BEAT2 motion into units with natural gesture boundaries, human-verified communicative functions, and feasible trajectories. Second, to preserve semantic order and coordinate body motion with the spoken response, we introduce a Semantic-Conditioned Compositional Planner. Given an ordered semantic function sequence and an estimated response duration, the planner selects approved atoms to realize the longest feasible action sequence while accounting for transitions and neutral recovery. Finally, we deploy a Streaming Face-Speech-Body Integration system on a physical G1 humanoid, combining streaming dialogue audio, audio-driven facial animation, and semantically planned body motion in a unified real-time interaction loop. Quantitative and qualitative experiments demonstrate that TRAbot achieves the best overall performance among all compared conditions in terms of naturalness, expressiveness, and multimodal coherence.

The goal of the present paper is to investigate the exact decay rate of the tail probability $\mathbb{P}(|X_1-x_0|>R)$ for large $R$, where $X_t$ is the solution to a multidimensional stochastic differential equation driven by a fractional Brownian motion with initial condition $x_0$. In the first place, under the assumption of uniform ellipticity, we establish a general $(2H+1)$-Weibull lower tail estimate in Young's regime of $H\in(1/2,1)$ and a general Gaussian lower tail estimate in the rough regime of $H\in(1/4,1/2)$. These two estimates are seen to be sharp in their respective regimes. In the second place, we prove a striking fact by constructing explicit examples that a uniformly elliptic system with $C_b^\infty$-coefficients could fail to have Gaussian lower tail in Young's regime. As a consequence, in our modest opinion, the multidimensional Gaussian lower estimate of \cite{BKT16} might not hold in its current form of generality without further assumptions on the vector fields. In the third place, we provide a simple nondegeneracy condition on the vector fields, under which a Gaussian lower tail can be established in Young's regime. In addition, given the existence of the aforementioned counterexamples, we prove another striking fact that any $2\times 2$ periodic, uniformly elliptic system always has Gaussian lower tail in Young's regime. Lastly, in the rough regime we establish a general $(2H+1)$-Weibull lower tail for a rich class of systems that satisfy a noncommutativity condition on the vector fields. This result provides a partially affirmative answer to a conjecture raised in \cite{BG24} on the genericness of the well-known Cass-Litterer-Lyons estimate for noncommutative systems.

We assess neutrino-assisted dark matter self-scattering in two scalar-mediator realizations by combining mediator-resolved potentials, numerical scattering calculations, thermal relic production, and the adopted observational constraints. In the $t$-channel case, the adopted laboratory and cosmological bounds suppress the neutrino force contribution at points reproducing the relic abundance and yielding sizable self-interactions, which are dominated by tree-level mediator exchange. Sommerfeld enhancement and early kinetic decoupling can change the relic abundance by up to 50% in the light mediator regime. In the $s$-channel case, thermal freeze-out fixes the coupling, leaving self-scattering well below the order of magnitude required for explanation of small-scale structure of the universe. An analytic upper envelope explains this shortfall of more than eleven orders of magnitude relative to the self-interacting dark matter benchmark. These results establish that, under the adopted constraints and thermal relic assumptions, the neutrino force cannot account for astrophysically relevant dark matter self-scattering in either realization.

Alternating projections (AP) near a clean intersection of smooth manifolds can converge slowly when the Friedrichs angle is small. We develop the Fixed-Point Manifold Acceleration (FPMA) framework for {local convergence and polynomial acceleration of} smooth maps that fix a manifold pointwise. {Under suitable regularity and uniform normal spectral stability, FPMA establishes local R-linear convergence and a smooth limiting map inducing a local retraction. At fixed points, consistent polynomial acceleration with $p(1)=1$ preserves the selected limit's first-order response to normal initial perturbations whenever the transformed normal blocks are uniformly spectrally stable. Under sufficient smoothness, we give a curvature criterion for second-order induced retractions and show that stable consistent polynomial acceleration preserves both the quadratic retraction expansion and this criterion. {As a special case, we establish new local R-linear convergence guarantees for fixed-degree Chebyshev-accelerated exact AP and second-order induced retractions. For uniform Friedrichs angle $\theta_F\in(0,\pi/2)$, acceleration strictly improves the local convergence rate per AP evaluation. The large-degree limiting spectral factor at fixed $\theta_F$ predicts $O(\theta_F^{-1})$ evaluations for fixed relative error reduction as $\theta_F\downarrow0$, versus $O(\theta_F^{-2})$ for ordinary AP.}} {The framework also covers} inexact, relaxed, and generalized AP {under their respective regularity and stability conditions.}

We extend a time frequency framework previously developed for directly modulated lasers to external modulation of the continuous wave output of a 1550 nm DFB laser described by Langevin rate equations. Four modulators are considered an InP electro absorption modulator, a LiNbO3 dual drive Mach Zehnder modulator, a dual parallel Mach Zehnder modulator and a silicon carrier depletion modulator

Power Delivery Networks (PDNs) are critical components of modern VLSI chips, providing stable voltage levels while satisfying electromigration (EM) and IR-drop constraints. Conventional PDN design methodologies typically rely on worst-case assumptions, often resulting in over-provisioned networks and inefficient use of resources. This paper presents a reinforcement learning-based framework for the optimization of workload-aware PDNs. The proposed methodology first generates workload-aware PDNs using architectural power traces obtained from system-level simulations. These power traces are mapped to spatial power density distributions, enabling adaptive allocation of PDN resources according to local current demand. A reinforcement learning agent then performs wire-width optimization to minimize PDN area while maintaining EM and voltage integrity constraints. Electrical and reliability metrics are obtained using SPICE-based circuit analysis and EM lifetime estimation. Experimental evaluation is performed on a dataset of workload-aware PDNs generated from 4-, 8-, and 16-core multiprocessor floorplans using PARSEC and SPLASH-2 benchmark workloads. Furthermore, the proposed Deep Q-Network (DQN)-based optimizer reduces the average normalized PDN area by 47% while satisfying all EM and IR-drop constraints. Compared to simulated annealing, the proposed approach achieves comparable optimization quality while providing approximately 26$\times$ faster optimization.

Vision-language models (VLMs) achieve strong zero-shot transferability but remain vulnerable to target-domain shifts at inference time. Test-time adaptation (TTA) offers a practical remedy, yet most existing VLM-TTA methods follow a prediction-side adaptation paradigm. They use test samples to adjust logits, prototypes, caches, priors, or feature statistics, often incurring additional computational overhead. In this paper, we take a different perspective and reframe VLM-TTA as candidate verification rather than prediction adjustment. We propose Test-Time Correction (TTC), a hypothesis-based correction framework guided by a simple principle: hypothesize, reconstruct, correct. Given a test feature and its top-k candidate labels, TTC treats each candidate label as a hypothesis, reconstructs the feature within the corresponding latent subspace stored in a memory bank, and measures the resulting divergence shift. This shift quantifies how much the candidate subspace and its relations to other candidates change after the hypothetical insertion of the test feature. A correct candidate hypothesis induces only a small shift, whereas an incorrect one perturbs the subspace more strongly. TTC therefore corrects the prediction by selecting the candidate with the minimum aggregated divergence shift. This training-free candidate-verification mechanism avoids iterative optimization and provides a favorable accuracy-efficiency trade-off. Across five TTA settings and 15 benchmark datasets, including zero-shot classification, domain generalization, few-shot classification, base-to-novel generalization, and cross-dataset evaluation, TTC consistently improves accuracy over state-of-the-art VLM-TTA methods while achieving up to 2x speedup, over 3x lower CPU memory usage, and up to 1.4x lower GPU memory usage than the lowest-memory training-free baseline.

Accurately quantifying the predictive uncertainty or improving model calibration plays an important role in medical image classification, in particular in melanoma diagnosis, where accurate uncertainty quantification can have significant implications for patient care. One of the methods for calibration improvement is data augmentation. In addition, data augmentation as a method for synthetically increasing the size of the dataset has been proven to improve the performance of models trained on imbalanced datasets. However, the impact of data augmentation, as a transformation of a part of the original data, on calibration of models trained on imbalanced datasets, in particular in melanoma classification is under-explored. We train neural networks on SIIM-ISIC 2020 melanoma classification dataset under two conditions: with and without data augmentation, and compare the differences in AUC and expected calibration error (ECE) in both scenarios. Our results shows improvements in uncertainty calibration using different augmentation methods.

We formulate equations of motion for uniform magnetization dynamics in multisublattice magnets, with emphasis on a recently proposed inertial damping torque. We include intra- and intersublattice contributions to Gilbert damping, reactive inertia, and inertial damping. For a two-sublattice uniaxial antiferromagnet in an easy-axis static magnetic field, we derive the dynamical magnetic susceptibilities and approximate eigenfrequencies across all field-dependent equilibrium phases. In particular, we show how inertial damping affects the relaxation of the nutational modes and propose an experimental protocol to determine the dynamical coefficients.

We study the influence of the ion magnetic-drift direction relative to the active X-point (favourable and unfavourable configurations) on edge turbulence, transport, and the radial electric field $E_r$ in ASDEX Upgrade and TCV. First-principles gyrokinetic simulations of the edge and scrape-off layer are performed in diverted X-point geometry using the spectral full-$f$ GENE-X code. Two matched pairs of L-mode discharges are simulated and validated against the available measurements in all four cases, showing good agreement, in particular for the $E_r$ profiles. A radial force-balance analysis shows that the deeper $E_r$ well in the favourable configuration is determined by the larger poloidal ion flow near the separatrix, which cannot be reproduced by neoclassical solvers and is identified in axisymmetric simulations as turbulence-driven. We then derive an exact evolution equation for the mean poloidal-flow acceleration from the full-$f$ gyrokinetic vorticity equation, which identifies the radial gradient of the total velocity stress as its nonlinear drive. The observed asymmetry is found to arise from stronger nonlinear energy transfer into the mean flow in the favourable configuration, associated with enhanced nonlinear $E\times B$ advection of vorticity near the separatrix and X-point. Flux-driven TCV simulations further show that the asymmetry persists at matched power more than six times the L-mode value. Assuming a critical $E_r$ at the L-H transition, this yields an estimated ratio of approximately 1.7 between the unfavourable and favourable L-H power thresholds, close to the experimental factor of two. Overall, the unfavourable configuration exhibits larger fluctuation amplitudes, a weaker $E_r$ well, stronger geodesic acoustic modes, and shorter energy confinement times for the same profiles, which may explain the observed difference in access to high-confinement regimes.

We develop a time frequency framework for the dynamics of directly modulated semiconductor lasers in which the optical field obtained from the singlemode rate equations is represented by distributions of Cohen class. Closed form small signal solutions are derived for RC filtered square current pulses and for sinusoidal modulation, including the exact chirp to power ratio with gain compression and spontaneous emission and a line expansion of the Wigner distribution of the sinusoidally modulated field. Kernels satisfying two known conditions return the laser chirp as the first conditional moment.

Quantum compilers synthesize high-level operations, such as the quantum Fourier transform and multi-controlled X gates, into gate-level circuits. Across compilation stages, a circuit may be serialized, exchanged as OpenQASM, converted between compilers, or lowered to gates. These representation changes can preserve computation while removing the high-level operation itself, leaving the receiving compiler unable to apply a requested synthesis method. We call this loss synthesis availability. We study synthesis availability in Qiskit, TKET, and Cirq through controlled experiments, repository analysis, circuits captured from project tests, and a Munich Quantum Toolkit pipeline. Synthesis availability survives serialization when deserialization restores the high-level operation and survives compiler conversion only when the receiving compiler supports that operation. It does not survive the evaluated OpenQASM routes. After a direct OpenQASM 3 round trip, requested synthesis has no effect in all 360 delivered conditions even though compilation succeeds. In the MQT pipeline, OpenQASM 2 interchange increases the two-qubit gate count by 37.2 percent for an eight-qubit Grover circuit. We also introduce Verified, which records high-level operations before a representation change and reconstructs them only when a check confirms that the record still matches the delivered circuit. Verified restores requested synthesis in all 830 recovery runs, with median verification times from 2.0 to 638 ms per operation family. The check rejects every stale record but also some valid records. These results show that preserving functional equivalence alone is insufficient when later synthesis choices depend on retained high-level structure.

We prove that every subspace $V\subset\ell_\infty^N$ of codimension one or two with absolute projection constant $\lambda(V)<4/3$ satisfies $$ d(V,\ell_\infty^{\dim V}) \leq\frac{\lambda(V)}{4-3\lambda(V)}. $$ We conjecture that the codimension assumption can be removed and that the same estimate holds for every subspace $V\subset\ell_\infty^N$ with $\lambda(V)<4/3$.

A Lifshitz transition is a class of non-symmetry-breaking transitions typically associated with metallic systems wherein the Fermi surface changes its connectivity as the chemical potential sweeps through a band extremum or a saddle point. The phenomenology naturally adapts itself to the superconducting state, where the object that reconstructs is the manifold of minimum quasiparticle energy rather than the Fermi surface. This paper concerns Lifshitz-type transitions in multi-band superconductors. Thus, we focus on the attractive Hubbard–Kondo model on the two-dimensional Lieb lattice, with classical core spins coupled to the itinerant electrons, by laying recourse to a real-space Bogoliubov–de Gennes mean-field theory and an analytic Green's function calculation. The competition between the Kondo and pairing channels yields a magnetic superconductor throughout, with $(\pi,\pi)$ order near half filling giving way to $(0,\pi)$ and then, at stronger Kondo coupling, to an incommensurate spiral $(0,q)$. In such a multiband superconductor hosting a flat band, we identify two inequivalent Lifshitz-type transitions. With the core spins decoupled (the Kondo coupling set to zero), the minimum energy contour deforms with chemical potential, a hole pocket about $M$ opening into a perfectly nested square at $\mu=-2t$ and reconnecting into pockets about $\Gamma$; this nesting fixes where $(\pi,\pi)$ order sets in. At finite Kondo coupling, for certain chemical potential, the dimension of the manifold itself drops. We obtain it analytically for each magnetic order and show that only the spiral carries the minimum energy through zero, nucleating nodes and leaving a gapless magnetic superconductor, with clear signatures in the density of states and the spectral function.

Melanoma is the deadliest type of skin cancer, whose early diagnosis is crucial for patients' survival. Image classification using deep learning models has shown promising results for melanoma diagnosis. However, the performance of these models on the melanoma datasets such as SIIM-ISIC melanoma classification dataset is a challenge due to the class imbalance. One of the methods to deal with this challenge is using loss function modifications. In this work, we have investigated the effect of different loss functions on the performance of deep neural networks. We trained these networks using focal loss, logit-adjusted softmax cross-entropy (CE) loss, and weighted softmax CE loss, and we report different metrics for evaluating performance and uncertainty calibration. Our results suggest that focal loss delivers a good combination of performance in terms of AUC and uncertainty calibration in terms of expected calibration error (ECE) simultaneously.

Graph Convolutional Networks (GCNs) have emerged as a powerful framework for learning from graph-structured data, yet their deployment on resource-constrained edge platforms remains challenging due to the computational and memory demands of sparse graph aggregation. This work presents an FPGA-based GCN accelerator that combines DSpar graph sparsification, 8-bit quantization, and approximate multipliers on the AMD Kria KV260. Evaluated on Cora, LastFM Asia, and Amazon Photo, the design explores the interaction between sparsification and approximation across graphs with widely varying densities. Results show that the effectiveness of approximate arithmetic is governed by accumulation depth within GCN computations. Approximate multipliers are most effective when applied to sparse aggregation operations, while graph sparsification further improves their viability by reducing aggregation depth. The combined approach achieves up to 9.88$\times$ speedup while maintaining 86.6% classification accuracy on Amazon Photo, and 1.52$\times$ speedup with 77.0% accuracy on Cora, with total power consumption below 1 W. These results demonstrate that graph sparsification and approximate computing are complementary techniques whose co-optimization enables efficient low-power GCN inference on edge FPGA platforms.

Human videos offer scalable manipulation data, but the embodiment gap between human hands and robot manipulators limits their direct use. Existing video-editing methods replace hands with rendered robots, yet inaccurate interaction reconstruction and compositing can produce inconsistent grasps and implausible robot-object occlusions. We address these failures from two complementary physical aspects: interaction geometry and scene visibility. First, an interaction-aware contact reconstruction module combines hand-object segmentation with mesh-level contact prediction to recover dense 3D contacts, then converts them into temporally stabilized grasps for parallel-jaw grippers. Second, a depth-aware compositing module uses scene and robot depth to enforce physically consistent robot-object occlusions. The resulting videos preserve the interaction structure of human demonstrations in a robot-compatible form and are co-trained with robot demonstrations. Using identical human videos and robot data, we compare against robot-only training and the original Masquerade pipeline. Across four RoboTwin tasks and two Diffusion Policy visual encoders, our method achieves the highest average success rates, with especially strong gains under out-of-distribution scene variation. Real-world deployment further shows that the proposed co-training approach improves robustness to visual distractors when the task geometry is observable, while performance on depth-sensitive grasps remains limited by the single-camera setup.

In this paper, we introduce and study a special function, referred to as a variant of the fractional gamma-regularized zeta kernel, in which the lower incomplete gamma function appears in the numerator, contrasting with the fractional hypergeometric zeta function, where it appears in the denominator. We investigate its analytic properties, convergence behavior, and its connection to the Riemann zeta function. A representation is derived in terms of a multiplicative modification of the Riemann zeta function, and comparisons are made with known zero-free regions and deformation phenomena. We demonstrate that this function defines a deformation of the Riemann zeta function, which interpolates to the alternating zeta function in a specific limit. Additionally, a Dirichlet-type series representation is provided. Potential extensions, including a functional equation and asymptotic analysis, are proposed for future work.

Engineering design often starts from a 2D sketch that fixes style and proportions, yet the subsequent 3D shape optimization relies on learned generative priors to keep the geometry valid. However, these priors are agnostic to the sketch: while they admit a valid design by correcting a drifted proposal back to its training distribution, they often correct it towards the high-density region, ignoring the specified design. We introduce AnchorGen, a rectified-flow framework trained unconditionally on the concatenated shape and sketch latents of paired data. The learned manifold represents the joint distribution of shape-sketch pairs, so constraining the sketch component restricts the iterate to the sub-manifold of shapes consistent with a target style. Since training employs no conditioning signal, the constraint is imposed at inference: gradient descent optimizes the shape latent to minimize a differentiable drag surrogate, while constraining the sketch latent to remain close to the target sketch via a token-wise cosine penalty. A single model thereby supports design-preserving optimization, dimensionally explicit design edits, and sketch-only synthesis.

This paper proves propagation of energy moments for the three-dimensional plasma-charge model, namely, the repulsive Vlasov-Poisson system with a moving point charge. For every \(K>2\), an initially finite \(K\)th energy moment remains finite on every bounded time interval, without any higher-order moment assumption, smallness condition on the plasma mass, or initial separation from the point charge. The constructed global weak solution conserves total energy for almost every time. The proof relies on a critical space-time density estimate and a direct control of the electric field along the moving charge.

Journal entry anomaly detectors are commonly evaluated on the full population with ROC-AUC, precision and recall, ignoring the review budget and which anomaly types are found. We propose a type-aware evaluation combining per-type recall, fair-share type recall (FSR), which caps each type's credit at its budget share, type coverage and first-hit rank. We evaluate nine unsupervised detectors, a supervised reference and feedback-driven Deep Semi-Supervised Anomaly Detection (DeepSAD) on four real client ledgers with injected typed anomalies and a public synthetic ledger. On the largest client ledger, principal component analysis (PCA), an autoencoder (AE) and a variational autoencoder (VAE) each place on average 98 anomalies among the first 100 postings, but at least 95.8 belong to one type. FSR instead favours a nearest-neighbour (kNN) detector and changes the top-ranked detector on three of four client ledgers. Representation also matters: one-hot encoding exposes unseen accounts, whereas frequency encoding leaves unseen contra accounts largely undetected. On the public ledger, the Histogram-Based Outlier Score (HBOS) and Empirical Cumulative Distribution-Based Outlier Detection (ECOD) reach all eight markings within 1,386 entries, whereas kNN, the hit leader at 1,000 entries, first reaches cross-linked clearing at rank 4,641, and the supervised row-level reference misses this marking within 1,000 entries. There, the adaptive DeepSAD review protocol raises mean hits per 100 reviews from 40.0 to 68.3 but type coverage only from 2.7 to 3.0. These findings show that high hit rates can conceal systematic blind spots and suggest that feedback can reinforce existing detection patterns without broadening anomaly coverage.

Implementing security features—functionalities that protect sensitive data or prevent malicious actions by attackers—is important for ensuring the security and integrity of software systems. Correctly implementing access control, cryptography, or other security features is challenging as they require substantial domain expertise, careful design, and custom implementation to integrate them within the software system. Previous work has thoroughly studied security awareness, perception, practices, and expertise via surveys, interviews, and experiments. While they have shown that the implementation of security features are often supported by security libraries and frameworks, they have also shown that developers rarely are security experts, and make mistakes that introduce vulnerabilities into software when using them. Despite this extensive knowledge of security practices and developer behavior, we still lack a comprehensive understanding of how security features actually manifest at the code level across full software systems. We close this gap by conducting a mining study of security features in 9 popular and large open-source Java systems. We manually inspected 2,127,761 lines of code across 19,121 files, identifying and annotating 183,395 lines implementing 561 security features corresponding to 54 security features in our taxonomy. We analyzed the characteristics of the identified security features, such as their size, scattering, tangling, common implementation patterns, and the use of internal and external functionalities. Our findings show that security features are far more than mere calls to external libraries. Security features, such as access control, can grow large in size, scatter across the whole codebase, and frequently tangle with each other. External security libraries are commonly used, but they require substantial amounts of code for integration.

In this contribution we provide a compact model for the description of inelastic x-ray scattering of spatially localized excitons. We combine our analytical theory with first principles calculations and apply it to the material lithium fluoride (LiF). A comparison of two different approaches for the underlying single-particle wave functions to calculate the spectrum yield a qualitatively good agreement with experimental data but also shows that a detailed description of the wave functions is essential to obtain qualitative results.

The dynamics of chemical reaction systems with two chemical species can be described (under mass-action kinetics) by planar autonomous systems of ordinary differential equations (ODEs) with right-hand sides containing polynomials. While their behaviour close to hyperbolic critical points is well understood, their dynamics has not been fully characterized for strongly non-hyperbolic critical points, where the linearization close to the equilibrium is given by the zero matrix. In this paper, the behaviour of such chemical reaction systems is investigated. Considering second-order chemical reaction networks with a positive equilibrium point, we show that only nine geometric equivalence classes are possible out of the sixteen classes for general ODE systems. Examples of chemical reaction networks in each class are presented. All sixteen geometric equivalence classes can be realized by chemical systems if the non-hyperbolic equilibrium point is located at the origin.