Papers

The paper proposes a generalization of the fully implicit energy- and charge-conserving electromagnetic particle-in-cell method to the case where the lightest type of plasma particle (electrons) is described in the drift-kinetic approximation. This allows us to remove the very strict time step limitation of this method requiring to resolve the gyrorotation of electrons. Since the drift-kinetic model is only applicable to light particles whose contribution to plasma polarization is small, this model can be further simplified by neglecting the electron polarization drift and including in Maxwell's equations, in addition to the current of gyrocenters, only the magnetization current. In order for such a hybrid model to retain conservative properties in finite-difference form, a method of self-consistent interpolation of $\nabla B$ in the mirror force from grid to particle and the magnetization vector from particle to grid is proposed. Unlike existing drift-kinetic models, this model is not limited to considering small perturbations near a given equilibrium, and therefore allows one to study the formation of plasma equilibria in regimes with a finite ratio of plasma to magnetic field pressure. Testing of the parallel code implemented in C++ using the PETSc library confirmed the fulfillment of the finite-difference laws of energy and charge conservation, as well as the ability of the model (in the drift-kinetic version for all types of particles) to correctly reproduce the diamagnetic effect and longitudinal ion-acoustic wave.

Generalization in large language models (LLMs) is the ability to produce consistent and semantically stable outputs when the same input is expressed in different ways. Existing work typically evaluates generalization through aggregate accuracy on a single prompt format, task, or set of variations, which conflates robustness with overall benchmark performance. In this work, we show generalization evaluation at the level of individual examples, across multiple input variants, and across different aspects of model behavior, focusing on variability rather than reducing performance to a score that can be improved through narrow training or other ways that obfuscate generalization evaluation. Following this view, we introduce the Stability-Aware Generalization Objective (SAGO), a framework that measures how much model behavior changes for the same input under different variations and benchmarks, capturing variability across several dimensions including generation consistency, internal activations, confidence, and response mirroring. We show that many commonly used models exhibit statistically significant and consistent generalization instability: no model generalizes uniformly, behavioral axes capture independent failure modes, and cross-dataset variation can reverse model rankings.

Levantine Arabic (LA) is spoken by tens of millions of people, creating a pressing need for shared benchmarks to evaluate LA speech-language technologies. Evaluating such technology is particularly challenging given LA's internal diversity and its opaque and non-standardized orthography. We present SHAMS (SHami Annotated Multi-dialect Speech), a benchmark comprising 1,300 utterances drawn from open audio corpora, balanced across five LA varieties (Urban and Rural Palestinian, and Urban Jordanian, Lebanese, and Syrian). Each utterance is represented across four aligned tiers: audio, unvocalized orthography, diacritized text, and phonetic transcription. This structure supports evaluation of various downstream tasks such as diacritization, grapheme-to-phoneme conversion, automatic speech recognition, and audio-to-phoneme, grounded in audio and stratified by variety. We benchmark open and proprietary models across these tasks to demonstrate the utility of this benchmark for measuring progress across LA. We release SHAMS at https://shams-nlp.github.io .

Prescribing the speed of gradient flow on the risk itself, by the dynamics $\dot w=-u(E(w))\nabla E(w)/\abs{\nabla E(w)}^{2}$, makes the risk $e(t)=E(w(t))$ obey $\dot e=-u(e)$ exactly, whatever the landscape~$E$; the time needed to reach zero risk from $e_0$ is $\int_0^{e_0}\dd e/u(e)$. Minimizing this time alone is ill posed, and we study the regularized problem $\inf\{\int_0^{e_0}(\tfrac\lambda2\abs{u'}^{2}+1/u)\,\dd e:\ u\in H^{1}(0,e_0),\ u\ge0,\ u(0)=0\}$, $\lambda>0$. We prove that the minimizer exists, is unique, and is a linearly scaled cycloid, and we show that the optimal rate behaves like $u^{*}(e)\sim(9/(2\lambda))^{1/3}e^{2/3}$ near zero risk: the exponent $2/3$ is the one found in \cite{betti2026holder} by a power-law ansatz, and it lies in the Hölder window $(\tfrac12,1)$ where the arrival is in finite time with vanishing weight speed. The proof follows the classical route: existence by the direct method, uniqueness by strict convexity, positivity of the minimizer away from the origin, and the explicit integration of the Euler-Lagrange equation.

During fine-tuning, training accuracy often keeps improving after validation accuracy levels off, widening the train-validation gap. We propose a support-transfer account: early updates reuse structure shared by many examples, whereas later fitting relies on narrower structure whose benefits transfer less to held-out data. From fixed training examples, we normalize per-example gradient magnitudes over selected parameters and measure allocation dispersion across examples ($D_{\mathrm{den}}$) and within labels, weighted by class sample counts ($D_{\mathrm{within}}^{\mathrm{wt}}$). A local model gives sufficient conditions for dispersion and the cross-entropy gap to grow together, separating uniform score scaling; additional decision-boundary conditions characterize accuracy-gap growth. In a five-level hierarchy, increasing the share of private features, unique to individual examples, from 0.3 to 0.5 to 0.7 raises the final accuracy gap from 0.185 to 0.331 to 0.527 and increases unnormalized dispersion increments. Tests retaining one feature level at a time show a widening validation-minus-training difference in the private level's loss share. After the validation-loss minimum, the cross-entropy gap continues to grow without a decline in mean validation accuracy. Across 90 runs of RoBERTa, DeBERTa, and Qwen3-1.7B on six language tasks, both readouts correlate positively with the accuracy gap in every run; raw correlations of one-epoch changes are positive for $D_{\mathrm{den}}$ in 88/90 runs. Both readouts also retain positive associations in 15 ResNet-18 runs across three vision datasets. These results support a connection between evolving gradient allocation and train-validation separation.

The integration of Vehicle-to-Everything (V2X) communications is driving a profound transformation in vehicular connectivity, expected to significantly enhance traffic efficiency and safety. However, the stringent requirements of V2X services, particularly ultra-low latency and high reliability, present significant technical challenges. 5G's Network Slicing emerges as a key enabler by providing tailored virtual networks that ensure isolation and adaptability for heterogeneous services. This work proposes an intelligent Radio Access Network (RAN) slicing management framework specifically designed for scenarios where safety-critical V2X and high-capacity eMBB slices coexist. In such complex environments, harmonizing conflicting traffic requirements demands continuous, data-driven optimization. To achieve this, the proposed framework leverages an advanced Deep Reinforcement Learning (DRL) approach which dynamically optimizes resource allocation in real time. The framework is empirically validated on a real 5G Standalone (SA) network, where experimental results demonstrate that the DRL-driven approach successfully balances both objectives, outperforming traditional static and proportional allocation strategies by minimizing SLA violations while ensuring high resource utilization for eMBB slices.

Structured light can generate electronic dc currents with azimuthal winding, yet the rules governing their winding order $m$ remain unclear. Here we identify sector-resolved winding selection rules using graphene as a clean two-dimensional platform. By decomposing the current response into local and gradient sectors at the current-operator level, we show that the winding order is determined not by the optical orbital angular momentum $\ell$ alone, but by the angular structure of the corresponding current operators together with projection onto the azimuthal direction. For scalar Laguerre-Gaussian beams, linear polarization yields $m=|\ell\pm1|$ in the local sector and $m=|\ell|,|\ell\pm2|$ in the gradient sector, whereas circular polarization with helicity $\sigma=\pm1$ selects $m=|\ell-\sigma|$ in the local sector and $m=|\ell+2\sigma|$ in the gradient sector. Numerical time-evolution calculations verify these rules and further show that helicity can select an $m=0$ branch, producing an azimuthally uniform circulating current whose radial profile determines the axial magnetic field $B_z(z)$. Our results provide a sector-resolved organizing principle for classifying and controlling structured-light-driven dc currents with tailored winding structures.

Let $a_t(n)$ denote the coefficients of the Nekrasov–Okounkov hook length generating function $F_{t^2}(x)$, and let $b_t(n)$ denote the coefficients of the $t$-core partition generating function $C_t(x)$. We prove a sharp prime-power dichotomy: for every integer $t\ge 2$, the congruence $$ a_t(n)\equiv b_t(n)\pmod t $$ holds for all $n\ge 0$ if and only if $t$ is a prime power. When $t=p$ is prime, the congruence strengthens to modulo $p^2$. For $t=6$, and more generally for every integer that is not a prime power, the congruence fails. This identifies prime powers as the exact moduli for which these two partition-theoretic coefficient sequences are arithmetically equivalent.

Assuming post-quantum one-way functions, we construct device-independent oblivious transfer and bit commitment that tolerate independent honest faults at a constant rate and adaptive, bidirectional quantum leakage between the laboratories at a constant rate. Honest parties use trusted classical computation to operate untrusted quantum devices, which may share arbitrary entanglement and perform joint measurements without assumptions of independence or identical distribution. Security is simulation-based against quantum polynomial-time adversaries and composes sequentially with efficient simulators. For security parameter $\lambda$ and a prescribed total leakage budget $\ell$, one complete one-bit oblivious transfer uses $\Theta(\ell+\log^2\lambda)$ device coordinates in at most three single-use device windows, including preprocessing, and polynomial classical computation and communication. The leakage budget is shared across these windows. The commitment has efficient simulators against both parties and yields device-independent coin tossing with abort; each uses one window. Completeness of oblivious transfer gives device-independent secure computation with abort for every efficiently computable classical functionality on a fixed number of parties, against static corruption of any proper subset, provided each reached call to oblivious transfer satisfies the device and budget conditions. The construction combines short selectors with code-based choice masking, a global audit, and error correction across field symbols. The commitment extractor changes a classical code coset while leaving the device execution and its leakage unchanged. Sender security uses a global Fourier analysis and a selector-aware Magic Square repetition bound derived from the two-round threshold theorem of Kundu and Tan (2023).

We discuss mechanisms of $D \bar{D}$ pair production in $e^+ e^-$ collisions. We consider the $D^0 {\bar D}^0$-continuum mechanism with the $t/u$-channel $D^*$ vector-meson exchanges, where the $D^* D \gamma$ coupling constants are derived from the $D^* \to D \gamma$ decays. For the $D^+ D^-$ channel, we consider mainly the $D^{\pm}$ exchanges. We find a relatively large continuum contribution for the $D^0 {\bar D}^0$ channel and a significantly smaller one for $D^+ D^-$, concluding that the broad bump at $M_{D^0 {\bar D}^0} = 3.8$ GeV observed by the Belle and BaBar Collaborations has a continuum origin rather than corresponding to the broad $\chi_{c0}(3860)$ resonance. We also investigate the resonant production of $D \bar{D}$ pairs, including $\chi_{c0}(3860)$ and $\chi_{c2}(3930)$. The $\chi_{c2}(3930)$ meson is a good candidate for the first excited $P$-wave charmonium $\chi_{c2}(2P)$ state. This resonance can decay into both $D \bar{D}$ channels; however, its branching fractions are not well known at present. Comparing our model result with $\Gamma(\chi_{c2}(2P) \to \gamma \gamma) = 0.544$ keV (corresponding to the Buchmüller-Tye potential) to the BaBar data, we find $B(\chi_{c2}(3930) \to D \bar{D}) = 0.58 \pm 0.13$. We present results for differential distributions in several variables for Belle II kinematics. Finally, we shortly mention predictions for the ultraperipheral heavy-ion reaction $Pb Pb \to Pb Pb D \bar{D}$ for low invariant mass ($M_{D \bar{D}} < 4.3$ GeV) at the LHC.

The photostability of DNA nucleobases relies on ultrafast nonradiative relaxation following ultraviolet excitation. In thymine, the competition between internal conversion (IC) and intersystem crossing (ISC) determines the balance between ultrafast energy dissipation and potentially long-lived photochemistry. However, the precise molecular mechanisms linking early singlet-state electronic and structural evolution to the formation of long-lived triplet dark states remain unresolved. Here, we investigate the gas-phase photodynamics of thymine using time-resolved multi-center X-ray photoelectron spectroscopy (tr-mc-XPS) supported by state-of-the-art ab initio quantum calculations. By simultaneously probing multiple carbon 1s core levels, we track site-specific electronic and nuclear dynamics. We discover that the long-lived dark state is not initially a pure triplet; rather, it exists as a mixture of singlet and triplet states for the first 240 ps before transitioning fully to the $^3\pi\pi^*$ state. On the femtosecond timescale, we resolve atom-specific vibrational coherence at $\sim$730 cm$^{-1}$ (a ring-breathing mode with a period of 46 fs) that survives the initial $^1\pi\pi^*$ to $^1n\pi^*$ IC. Furthermore, we identify a critical dual role for the methyl group (C9H$_3$). During singlet relaxation, the methyl group acts as an inertia spectator, and its mass hinders direct internal conversion to the ground state. Upon $^3\pi\pi^*$ state formation, however, state-dependent hyperconjugation couples the methyl group to the pyrimidine ring, turning it into a highly sensitive electronic reporter of ISC. These findings establish tr-mc-XPS as a powerful approach for disentangling the complex, multi-timescale photodynamics governing nucleobase stability.

Spiking Neural Networks (SNNs) enable event-driven computation through biologically inspired dynamics at the neuronal scale, while Mixture-of-Experts (MoE) perform conditional computation through expert selection at the model scale. Integrating their strengths offers potential for flexible neural architectures. A key challenge, however, lies in designing an expert selection mechanism based on spiking activity. To address this, we introduce a spike-based k-WTA Router inspired by competition-inhibition observed in the hippocampal CA1 region. The router incorporates lateral inhibition and refractory period to select Top-K experts according to discrete spike counts. Building on this, we present SpikeMoE, a framework that integrates neuronal-scale spiking dynamics with model-scale expert selection. To address incomplete multisensory inputs in multimodal tasks, we further equip SpikeMoE with a two-stage missing-modality modeling module that combines empirical prototypes from an observed-modality pool with modality-specific learnable embeddings to construct missing-modality representations. Experiments on vision, language, and multimodal benchmarks demonstrate that SpikeMoE achieves state-of-the-art performance among the SNN baselines, matches or exceeds the performance of ANN counterparts, and maintains robustness across diverse missing-modality conditions. These results demonstrate a favorable trade-off between performance and energy efficiency, validating the integration of spiking dynamics with sparse expert computation and highlighting SpikeMoE as a promising approach to energy-efficient brain-inspired computing.

We find the optimal initial states of a linearized cavity-optomechanical system for estimating the single-photon coupling by quantum Fisher information. In the red-detuned (beam-splitter) regime the interaction exchanges excitations between the two modes, and for pure inputs the Fisher information is four times the variance of the excitation-exchange operator, reducing the optimization to variance maximization with a closed-form solution. With the mechanical mode in a Fock state the probe is degenerate: any state saturating the photon-number constraint is optimal, and a Fock reference with many phonons enhances the per-photon sensitivity linearly. For an arbitrary fixed mechanical reference it becomes a phase-matched squeezed vacuum, from an exact bound on that coherence obtained by a Lagrange-dual argument. Jointly optimizing both modes at fixed total excitation number, the quantum Fisher information obeys a Heisenberg bound quadratic in the energy, saturated by equal superpositions of the two extremal eigenstates of the generator, equivalently a two-mode NOON state after a balanced beam splitter; balanced product Fock states reach only a linear enhancement. Exact diagonalization in truncated Fock spaces confirms all the results, the familiar single-photon input being the simplest saturating case. Three extensions are worked out: the blue sideband has an unbounded two-mode-squeezing spectrum, so the entangled optimum becomes impossible and the balanced product Fock state is exactly optimal at definite energy; under cavity loss the information accumulation turns linear on the cavity-lifetime timescale, at a renewal-optimal duration of about two and a half lifetimes; and for a state-of-the-art quantum-coherent coupling experiment the per-hour relative precision lies between one part in a million and ten parts in a million, number and thermal references being the most robust.

Ringdown spectroscopy probes black hole (BH) dynamics, and exceptional points (EPs) can reveal underlying degrees of freedom. We develop a procedure to identify EPs in ringdowns using a time-weighted Jordan term and a splitting bound. For relative Jordan excitation proportional to separation, we derive a sixth-order squared distance between split and EP waveforms after fitting the resonance center. Independent constraints on these quantities can restore fourth order. Complete waveforms show that omitting ordinary responses can strengthen EP evidence and spuriously tighten splitting bounds.

Large language model (LLM) agents can now undertake increasingly complex tasks, but the way they organize interaction history into memory does not ensure a coherent understanding of the current world. We introduce PoS, an inference-time framework that constructs and continually maintains explicit belief states as the agent's decision context. Each belief combines an estimate of the current world state with unresolved task requirements, making explicit what the agent still needs to learn and accomplish. To keep this belief reliable and actionable, PoS validates its consistency and monitors task progress to detect Belief Trapping, where the agent continues to act without making meaningful progress toward the goal. Recovery is then tailored to both the trapping pattern and the type of unresolved task requirement. Experiments on four benchmarks spanning execution and diagnosis show that PoS achieves the highest overall performance on every benchmark with all three LLM backbones. Ablations demonstrate the importance of consistency validation and recovery, while context-scaling experiments show resilience to context growth. Together, these results support belief construction and continual maintenance as a foundation for long-horizon context management beyond history retention and compression.

We analyze thermalization and eigenstate thermalization in the random free fermion model, a quadratic Hamiltonian with a Gaussian Orthogonal Ensemble hopping matrix, by an exact random-matrix calculation. For a single-particle observable $A$ we compute the ensemble average and fluctuations of $\operatorname{Tr}(A\rho(t))$: we derive the relaxation envelope $g^2(t/\tau_\lambda)$ with $\tau_\lambda = \hbar/(2\eta\sqrt{N})$ and its oscillatory $|t|^{-3}$ tail, both properties of the semicircular limiting density, and we show that the variance of the fluctuations is $O(1/N)$, so that the relaxation profile self-averages at fixed scaled time. Incorporating particle-number conservation, we prove a factorization theorem for the energy-shell statistics of the many-body eigenstates: the energy-independent ETH structure follows from the statistical independence of the eigenvector matrix and the spectrum, and at fixed realization the energy constraint enters only through a covariance subtraction of relative order $1/N$. A controlled eigenvector–eigenvalue correlation deformation $H(\varepsilon)=H_0+\varepsilon H_1$ then generates, at first order in $\varepsilon$, the energy dependence of the diagonal energy-resolved slope and of the off-diagonal spectral function, a single-particle diagnostic of the ETH structure in the allowed particle-hole sector. Correlation-function fluctuations and the single-particle spectral form factor finally identify the subleading $O(1/N)$ late-time fingerprints of two-level spectral correlations. The model thus separates eigenbasis chaos from spectral chaos and identifies the statistical origin of energy-independent versus energy-dependent ETH.

Reinforcement learning (RL) algorithms frequently compare probability distributions, such as state visitation distributions induced by policies and experts, action distributions from learned policies and offline datasets, or transition distributions from learned models and environments. However, commonly used divergences may become ineffective when these distributions overlap weakly, which is frequently encountered in imitation learning, offline RL, and deployment under distribution shift. Optimal transport (OT) offers an alternative by measuring the cost of moving probability mass from one distribution to another under a ground cost that encodes task geometry. This survey covers how OT is used inside RL objectives and algorithms. For each method, we identify: the role OT plays, the distributions compared, the OT formulation used, and the treatment of temporal structure. Beyond categorising existing methods, we discuss the motivations behind different OT choices, practical considerations such as cost design and computational challenges, and highlight open problems including scalable trajectory-level transport, principled handling of mass mismatch, and theoretical analysis for OT-regularised RL.

We show that there exists a theoretically well-motivated Higgs sector in which the conventional heavy-Higgs production mechanisms are suppressed. In this limit, the $b$-associated production channel assumes the role of the primary production mechanism. We demonstrate that the democratic three Higgs-doublet model (3HDM) provides a viable beyond the SM (BSM) framework in which this limit can be realized. We further show that existing $b$-associated searches provide the most sensitive LHC probe of this 3HDM scenario in the up-phobic limit and derive the corresponding constraints on its parameter space. We also present projections for future collider searches, demonstrating their potential to probe further regions of the model's up-phobic parameter space.

Weak perturbations of the same size can lead to terminal Wasserstein errors of different orders. For infinitely divisible laws smoothed by a Gaussian, we differentiate their distribution functions with respect to variance-weighted Lévy characteristics. This includes distributional directions caused by moving atoms. We obtain first- and second-order expansions that are uniform over Lipschitz test functions. When the first response is nonzero, Gaussian analyticity gives the signed second coefficient of the Wasserstein–1 distance. When the first response vanishes, a finite second displacement moment is enough to obtain a quadratic response. Its coefficient is strictly positive if the displacement dispersion is nonzero. For local balanced remeshing, the error is comparable to the grid-alignment variance. This gives sharp quadratic grid rates for compactly supported densities and nonaligned atomic sequences. We derive metric speed and length along admissible non-atomic curves, and use the response to build a linear program with a proved bound on how much its error exceeds the infimum over the allowed class and consistent quadrature. We also prove a sharp transition as smoothing vanishes and give a multivariate second-order expansion. The appendix gives state-dependent response formulas and further stability estimates.

We study the critical points $U_\varepsilon$ of the two-dimensional Ginzburg–Landau energy on domains with corners $\Omega$ and imposed Dirichlet boundary condition which vanishes precisely at the vertices $V$. We isolate the forced scalar boundary layer $\rho_\varepsilon$ by the weighted Lassoued–Mironescu splitting. Assuming the reduced logarithmic energy bound and vertex logarithmic tightness, we prove that after passing to a subsequence, the normalized maps $v_\varepsilon = U_\varepsilon/\rho_\varepsilon$ converges in $C^k_{\mathrm{loc}}(\Omega\setminus A)$ $\forall k\ge0$ away from a finite set of interior vortices $A$. The limiting map is a canonical $S^1$-valued harmonic map $u_*$ containing integer interior vortex factors and fixed fractional corner factors. The renormalized energy is defined by removing discs around the interior vortices and sectors around the vertices and subtracting the corresponding logarithmic divergences. The interior vortex locations of the limiting map are critical points of this renormalized energy.

Reliable uncertainty estimation is essential for deploying object detectors when distribution/covariate shift and adversarial attacks may occur. Existing approaches often require detector retraining, architectural modification, or repeated inference, which may be infeasible or incur significant overheads. We introduce a lightweight post-hoc evidential meta-model that learns when object localisations should be considered uncertain while keeping the base detector frozen. Our approach automatically identifies localisation-relevant features and uses saliency-guided modification to construct an increasingly challenging curriculum. Detection-level targets combine localisation error, modification level, and prediction instability to guide an evidential meta-model to estimate uncertainty for each predicted bounding box. Our approach requires no changes to the detector and preserves its original localisation outputs. Across adversarial attacks and evaluated strengths, GRACE improves TP-FP AUROC by 22% relative to the strongest comparator in some cases while maintaining in-distribution detection performance.

Trajectory crossing remains a critical bottleneck in Flow Matching (FM), and previous works typically view these crossings from a theoretical optimization perspective causing velocity averaging. They attempt to address it indirectly by post-hoc distillation or endpoint coupling, without explicitly regulating the intermediate trajectories. In this paper, we introduce a new network learning perspective: crossing points inherently induce large local Lipschitz constants in the target velocity field, leading to two drawbacks. First, high Lipschitz constants correspond to high-frequency signals in the velocity field that neural networks struggle to fit due to spectral bias. Second, they also imply drastic velocity variations, leading to severe numerical integration errors in few-step inference. To alleviate this, we propose CoFlow, a framework that introduces the contrastive learning paradigm into FM to explicitly repel trajectories during training, thereby lowering the local Lipschitz constants of the velocity field. Specifically, we formulate CoFlow from a Stochastic Differential Equation (SDE) perspective by injecting a repulsive drift term. This drift actively guides the forward process of positive samples away from negative trajectories, effectively reducing the local Lipschitz constant. Furthermore, we derive an equivalent stochastic interpolant formulation from this SDE, providing a simple and tractable design space to control the influence of negative samples. Extensive experiments on ImageNet 256x256 demonstrate that CoFlow significantly reduces FID compared to standard FM in few-step inference (e.g., 20 steps), with no added training overhead. The code can be accessed at: https://github.com/HKUST-LongGroup/CoFlow

The Starobinsky model provides a successful realization of inflation and serves as a benchmark for alternative inflationary models. Motivated by its observational success, we investigate the dynamical evolution of the early Universe within this framework. Incorporating radiation production from scalaron decay, we formulate a coupled system of dynamical equations describing scalaron depletion and radiation buildup while ensuring exact energy conservation in the expanding background. In the Friedmann equation, both scalaron and radiation components gravitate, thereby accounting for the backreaction of the generated radiation. Starting with a radiation-free Universe and initial conditions from the standard slow-roll approximations, exact numerical evolution shows that inflationary dynamics is predominantly governed by the scalaron. Despite continuous radiation production, the radiation component remains subdominant, and the slow-roll conditions persist throughout inflation, yielding approximately 60 e-folds. Following inflation, the scalaron enters a damped oscillatory phase accompanied by rapid radiation growth. This transition provides a graceful exit from inflation and a smooth entry into reheating. During reheating, radiation grows continuously as the scalaron energy is depleted, with the radiation energy density ultimately becoming substantially larger than the residual scalaron energy density. This indicates the onset of a hot, radiation-dominated phase, satisfying the Kofman-Yi criterion for successful reheating. Our results demonstrate that the coupled scalaron-radiation dynamics provides a self-consistent realization of graceful exit and the transition to radiation domination in the Starobinsky model.

A novel analytical framework, designated as the JO&Minna method, is presented for interpreting the depth profiles of residual activation induced by heavy-ion beams in thick structural materials. By combining the limiting fragmentation and factorization principles, the JO-factor formalism rigorously links the spatial distribution of projectile-like fragments to the range derivatives and electronic stopping powers. Furthermore, the Minna procedure establishes a mathematical link via profile differentiation and convolution analysis to reconstruct the original energy spectrum and energy spread of the fragments from the measured activity profiles. The methodology is benchmarked using experimental data for $^{46}\text{Sc}$ production from $^{238}\text{U}$ irradiation in aluminum, demonstrating a reliable and straightforward approach for extracting fragment production kinematics and energy distributions in accelerator materials research.

This paper presents a collection of audio transforms that introduce perceptually irrelevant distortions and demonstrates their use as perceptual stress tests for audio quality models. We refer to these methods as Perceptual Audio Data Perturbation (PADP). PADP exploits the insensitivity of the human auditory system to certain fine-grained signal variations to substantially alter the waveform while preserving perceived audio quality and content. The audibility of PADP and the selection of its parameters are evaluated through controlled listening tests, ensuring that the transformations achieve transparent or near-transparent quality for both non-critical and critical items. We further probe the robustness of state-of-the-art (SOTA) perception-motivated objective audio quality models and foundation models. The results reveal a misalignment between model responses and human auditory perception, highlighting the limited perceptual awareness of these models for certain proposed transforms.

Graph clustering aims to partition the vertices of a graph into groups with dense intra-cluster connections and sparse inter-cluster connections. In this paper, we study a quadratic $K$-medoids formulation for graph clustering from the perspective of CM optimization. By identifying the binary feasible set as a discrete CM set, we establish a connection between the quadratic $K$-medoids model and CM optimization. Rather than solving the original binary problem directly, we consider the convex hull of its feasible set and introduce a negative squared-norm penalty to promote extreme-point solutions, resulting in an extreme point pursuit model over a simple convex set. The efficient projection onto this convex set enables the use of projected-gradient-based optimization for large-scale medoid selection. Based on this formulation, we propose a two-stage clustering method. In Stage~1, a penalty-continuation projected gradient solver is used to approximately solve the penalized medoid-selection problem. In Stage~2, the selected medoids are converted into a graph partition through a unified two-level assignment rule, where Jaccard distance is used as the primary criterion and shortest-path distance is used as a secondary criterion when needed. Numerical experiments on synthetic PPM and SBM graphs, together with real-world graph data, demonstrate the effectiveness and computational efficiency of our method. The experiments also show that a smaller objective value in the medoid-selection problem does not necessarily lead to a more accurate final graph partition, highlighting the importance of evaluating both optimization performance and clustering quality after assignment.

Spiking Transformers merge the energy-efficiency of spiking neural networks (SNNs) with the representational power of self-attention, creating a promising architecture for high-performance, energy-efficient computation. However, a performance gap persists versus its counterparts in artificial neural networks (ANNs). Unlike prior works attributing this to binary activations, we reveal that both spiking neurons and spiking self-attention (SSA) act as low-pass filters through multiscale spectral analysis. This characteristic leads to the dissipation of high-frequency components. To address this issue, we propose the Spiking Contrastive Attention (SCA) paradigm, which draw inspiration from the edge-detection and differential sensing properties of biological visual system. By extracting contrast prototypes via global contrastive aggregation and applying local differential refinement, SCA effectively enhances high-frequency information. Extensive experiments show that SCA is a general module that consistently boosts Spiking Transformers across image classification, semantic segmentation, and event-based tracking. Furthermore, it achieves lower complexity, offering superior efficiency over original SSA. These results establish its potential as a fundamental building block for energy-efficient Spiking Transformers.

Co-propagating chiral Majorana edge modes are predicted to exist in proximity-coupled quantum anomalous Hall systems. They combine the exciting prospects of edge vortices as flying non-Abelian Ising anyons, with their fully electrical detection through anyon fusion processes. In this setup, electrical transport arises from the interference of a pair of Majorana edge states. By bosonizing the model with an additional replica sector, we generalize previous treatments to fully asymmetric configurations. We analyze a Mach-Zehnder interferometer and show that while the DC conductance does not exhibit signatures of edge-vortex interference, at low frequencies, this interference manifests as an effective capacitance. This quantity is sensitive to the non-Abelian anyon statistics of edge vortices and, in particular, carries signatures of their topological spin and nontrivial conformal dimension. We demonstrate that the thermal length $l_T = \hbar v/(\pi k_B T)$ is the relevant scale at low temperatures $T$, analogous to quasiparticle transport in conventional quantum Hall edge states with velocity $v$. As long as the interferometer arm lengths match within this length scale, transport signatures of the non-Abelian statistics of edge vortices remain visible.

We define a symplectic structure on singular 1-form densities and a presymplectic structure on augmented triples, a compressible analog of the Marsden–Weinstein symplectic structure on vortex membranes in the incompressible case. Equivalence classes of augmented triples can be regarded as special vorticities in compressible fluids, i.e., singular elements of the dual to the Lie algebra of the group of all diffeomorphisms of a manifold, while Marsden–Weinstein symplectic structure is related to singular vorticities for the group of volume-preserving diffeomorphisms. This structure on singular 1-form densities occupies an intermediate position between two other settings, which we revisit, the Weinstein symplectic structure on weighted isotropic submanifolds of symplectic manifolds and the Marsden–Weinstein symplectic structure on vortex membranes.

The small scale density distribution of the cold circumgalactic medium (CGM) plays a central role in galaxy evolution, yet it is still poorly understood. We provide new constraints on the cold CGM density distribution by comparing the CGM emission from MUSE observations of quasar (QSO) nebulae at $z \sim 3.8$ with mock observations generated from the new DaLya simulation suite under the assumption of maximal fluorescence. These are further complemented by mock observations generated from the COLIBRE (m5 and m6 resolutions) and HELLO simulations. We first investigate the distribution of the observed and simulated Ly$\alpha$ surface brightness (SB) values, which directly depend on the cold gas density distribution under our assumption, finding stark differences between observations and upper limits provided by simulations with the exception of the COLIBRE higher-resolution run (m5). The observations generally show a much larger covering fraction of regions with relatively large Ly$\alpha$ SB values and suggest that densities reaching at least 10 cm$^{-3}$ must be present in the CGM of QSOs within a radial distance of 50 to 200 ckpc. To understand the origin of this discrepancy, we study the cold CGM density distribution finding in all cases simulated PDFs consistent with skewed lognormal ones with medians $\sim 10^{-2}$ cm$^{-3}$ and clumping factors between 16 and 102. We compare the broadness of these distributions to the one implied by the observed HeII/Ly$\alpha$ CGM ratio, which is independent on the assumption of maximum fluorescence. Assuming lognormal PDFs with median values similar to the simulations, the observed ratio requires a broader PDF with a clumping factor at least 8 times larger than predicted by simulations. These results suggest that higher resolution and/or additional physical mechanisms are needed to reproduce the observed CGM of QSOs at these redshifts.