Papers

Minimizing magnetic damping while understanding spin-lattice interactions remains a key challenge for magnonics. We show site-specific Al/Li ordering in spinel ferrites drives a ferrimagnetic insulating state, quenching the Fermi-level density of states. Ab initio calculations reveal collective acoustic modes alongside sub-lattice-selective optical modes on tetrahedral and octahedral networks, providing a microscopic understanding of substitution-driven spin dynamics. Crucially, the low-frequency magnon and acoustic phonon modes intersect, driving strong hybridization for low-loss technologies.

We propose a novel surrogate model for linear coupling in graded-index multimode fibers that is computationally more efficient than existing models, and can be tuned through a single measurable parameter. The model is benchmarked numerically and experimentally on a 90-polarization-mode fiber.

Antiferromagnetic oxides offer a compelling paradigm for magnonics. Combining ab initio and spin Hamiltonian modeling, we map Cr2O3 from bulk to monolayer limits, verifying stability via phonon spectra. Contrary to previously predicted half-metallic ferromagnetic monolayer phases, we conclusively demonstrate that monolayer and bilayer configurations preserve an antiferromagnetic semiconductor ground state despite systematic gap narrowing. Intersecting magnon and phonon branches drive strong hybridization, establishing low-dimensional Cr2O3 as an ideal platform for quantum spintronics.

Robots inspecting an assembly must determine which parts are present and whether they are correctly installed. During egocentric assembly assistance, head motion and workpiece handling reveal evidence for these checks, while spoken state confirmations link observations to procedural outcomes. We introduce INSPECT, which learns robot view preferences from records of a smart-glasses assistant that answers part queries and provides next-step guidance. Presence-Invariant TwinSwap (PI-TwinSwap) calibrates object evidence through paired identity interventions. Claim-indexed supervision separates evidence requirements from camera-reproducible observation changes. Object-centered calibration adapts relative view preferences to robot poses, while clause-level screening checks predicted evidence. The robot selects views using only its current observation and known poses, without candidate images. Evaluation uses annotated assistant-video replay to simulate state feedback, without target-domain view labels for policy training. On images of physical gearbox assemblies, INSPECT achieves the highest view utility among the compared non-oracle policies and raises human-rated full verifiability from 34.8% to 41.7% compared with keeping the current view. On commercial angle-grinder recordings in IMPACT, the transferred relative-view selector increases the correct decision rate from 50.6% to 54.3% with a frozen perception head. The source code is available at https://github.com/Kratos-Wen/INSPECT.

Frontier AI models are rapidly gaining the ability to exploit vulnerabilities in complex pieces of software. The risk is not theoretical, as evidenced by recent sandbox escapes performed by frontier models at OpenAI and Anthropic. Discussions of how to sandbox inference stack components often focus on components other than the inference engine itself (e.g., network proxies or code execution environments). However, the inference engine is an attractive target for a misaligned model. For example, if a model can trigger exploits in that engine merely by generating specially-crafted output tokens, the model can initiate a multi-step, to-the-bare-metal exploit chain in the engine, without relying on vulnerabilities in other components of the inference stack, and without assistance from externally-provided, maliciously-crafted input tokens. In this paper, we show that a misaligned model can perform inference engine fingerprinting to determine the specific engine (e.g., vLLM, SGLang) which executes the model. Once the engine has been fingerprinted, the model can leverage engine-specific exploits to take control of the engine using only carefully-selected output tokens. We provide concrete examples of model fingerprints in five popular engines, and demonstrate how realistic agentic harnesses allow a model to leverage those fingerprints to identify the local engine. We also describe a proof-of-concept, to-the-bare-metal exploit chain that originates from a fingerprinted (and subsequently compromised) inference engine. We conclude by discussing several ways that inference engines could be changed to make fingerprinting attacks more difficult.

We investigate a nonreciprocal quantum battery in which the charger and battery share one or two common reservoirs, with the charger driven by a degenerate optical parametric amplifier (DOPA). Reservoir-mediated dissipative interactions enable directional energy transfer, while the DOPA provides additional parametric am plification. We show that, in the stable below-threshold regime, the DOPA can enhance the battery energy and average charging power, with the nonlinear gain and pump phase providing effective control of the charging dynamics. Further optimization of the dissipative-coupling asymmetry improves the stored energy. These fea tures persist for both single- and two-common-reservoir configurations, demonstrating a tunable strategy for enhancing open-system quantum-battery charging.

On-policy self-distillation (OPSD) lets a language model learn from a frozen copy of itself that sees an answer or a worked solution. Giving the teacher this extra information seems to offer the student more to learn, but how much does it add beyond distillation itself? To isolate that contribution, we construct AMPLE-Math, a reusable suite of 5,319 mathematical problems with six reasoning views that share the same answer, and compare each view with matched reference-free distillation. With a thinking-enabled teacher supervising direct-response rollouts, reference-free distillation accounts for much of Qwen3-1.7B's improvement under thinking-enabled evaluation, both in domain and on external benchmarks. Evidence for an additional reference benefit is modest in Qwen, strongest for a polished solution, whereas complete traces add two percentage points in SmolLM3-3B at step 50. These benefits depend on the student being trained. At the same checkpoint, replacing short direct-response rollouts with long thinking-enabled rollouts turns gains into losses in both families while the problems, references, and evaluation stay fixed. Teacher profiles and matched loss interventions in Qwen further show that changing token-level supervision can leave student behavior largely unchanged. Together, these findings suggest that OPSD can improve access to existing reasoning capabilities through parameters shared by direct-response and thinking-enabled inference. The value of a privileged reference is what it adds to this cross-mode transfer, not how much of the solution it reveals.

Recent observations have yielded the first measurements of young exoplanet demographics. Close-in sub-Neptune occurrence rates appear to rise from young ($10-100$ Myr) to intermediate (100 Myr $-$ 1 Gyr) ages and then decline sharply in the old ($\gtrsim 1$ Gyr) Kepler field population. In this paper, we test whether these observations can be explained by the effects of planetary cooling, atmospheric mass loss, and tidal orbital migration, which we model through a fully coupled evolution framework. The orbital evolution is assumed to operate exclusively through high-eccentricity migration, in effort to estimate the maximum possible contribution from this migration channel. In reality, only a subset of planetary systems are expected to undergo HEM. We find that high-eccentricity migration rapidly populates the close-in sub-Neptune bin, producing a sharp rise in occurrence within the first $\sim 15$ Myr. After this early phase, the occurrence evolves only weakly. While the modeled young to intermediate-age evolution is broadly consistent with the observed trend within the uncertainty limits, it does not support the idea of a sustained young-to-intermediate rise driven by tidal migration. It also fails to reproduce the sharp decline in occurrence seen from intermediate to old ages, motivating additional physics or formation channels to be included in future models.

We show an information-theoretic lower bound of $\Omega\left(\frac{\log n \log m}{\log \log n + \log \log m}\right)$ for online set cover against randomized algorithms, for all sufficiently large $m$ and $n$ satisfying $\log^2 n \leq m \leq 2^n$.

We investigate localization in a time-dependent contrained obstacle problem for the density of an active protein on a cell membrane. This free boundary problem has been derived in Logioti et al. [17] as a model for cell polarization under the influence of an external signal. We show that in the limit of small mass there is instantaneous localization, that is the mass of the active protein concentrates in the set where the signal is maximal for any positive time $t>0$. We give a precise a characterization of this concentration process on the appropriate slow time scale and illustrate this for a few examples of the external signal.

Astronomical observations of planet-forming circumstellar disks indicate that planetesimals form from 0.1-mm- to 1-cm-sized dust aggregates, commonly called “pebbles.” When such pebbles accrete beyond the water snow line, the resulting icy planetesimals undergo water–rock differentiation and develop porous pebble-pile cores whose voids are saturated with liquid water. Circulation of this water enhances heat transport in the core, suppressing the temperature rise caused by the decay of radionuclides. In this study, we constrain the accretion age and constituent pebble size of Ryugu's parent planetesimal by modeling the thermal evolution of icy planetesimals and comparing the results with the precipitation ages and temperatures of aqueously formed minerals identified in samples returned from asteroid Ryugu. Our numerical results suggest that Ryugu's parent planetesimal accreted within 2.0 Myr of the formation of calcium–aluminum-rich inclusions. The inferred early accretion age supports the hypothesis that Ryugu's parent planetesimal formed earlier than most chondrule-bearing carbonaceous chondrite parent planetesimals, potentially explaining the absence of chondrules in Ryugu samples. We also found that the core of the parent planetesimal was composed of pebbles no larger than a few millimeters. Such small pebble sizes are consistent with theoretical predictions for planetesimals formed beyond the CO$_{2}$ snow line.

In this paper we provide a geometric characterization of simply connected domains in the complex plane supporting an unbounded analytic function with bounded derivative, providing a partial answer to Problem 1.45 in the Hayman List. For a given simply connected domain this characterization is achieved by estimating above and below the norms of certain evaluation functionals on a suitable Banach space of analytic functions on that domain by geometric quantities.

We investigate the behavior of arboricity under two fundamental graph operations, namely wedges and joins, proving an exact formula for wedges and establishing general upper and lower bounds for joins. Using the characterization of the simplicial geometric category of connected graphs in terms of arboricity, we derive a wedge formula for simplicial geometric category and obtain corresponding estimates for graph joins. Finally, we illustrate these results through explicit computations for several classes of graphs by constructing forest decompositions and the associated covers by strongly collapsible subcomplexes.

Recent factor graph approaches to continuum robot state estimation have been successful for quasi-static applications and spatiotemporal estimation using white-noise kinematic motion priors. However, when inertial effects are significant, these approximations may fail to capture the underlying physics, limiting accuracy during dynamic motions. In contrast, our approach approximates the Cosserat rod dynamics of continuum robots. We write inertia and damping as equivalent applied loads, so that the dynamic balance retains the algebraic form of the static one from prior work with quasi-static robots. Without backbone observations, the framework reduces to a stochastic forward simulation of the robot's motion. Given observations, it jointly refines kinematic and dynamic states and infers external loads, among other states. We validate the approach through simulation and experiments, demonstrating stochastic forward simulation as well as state estimation on tendon-driven continuum robots.

This paper presents a real-time semantic world modeling framework specialized for precision agriculture using autonomous robots. The framework combines probabilistic mapping of objects and their semantic attributes, updated through Bayesian inference, with a graph-based Simultaneous Localization and Mapping (SLAM) approach implemented using $g^2o$, a general framework for graph optimization. This integration enables accurate mapping and localization without relying solely on GPS. By leveraging semantic information such as plant type, size, and health, the robot can perform tasks while mapping and localizing itself within a field of crops. The proposed framework was validated through Gazebo simulations and physical experiments on an indoor field with artificial plants using Boston Dynamics' robot dog Spot. A YOLOv8n object detection model was trained to extract object and semantic data from depth camera observations. These simulations and experiments demonstrate that the system can successfully perform real-time mapping of up to at least 400 plants.

We prove that the height-level path ensemble of a uniform alternating sign matrix converges to the Airy line ensemble near every interior point of the north-west arctic arc. Under the standard bijection, this is equivalently a limit theorem for the osculating paths of the domain-wall six-vertex model at the ice point. The proof has two main inputs. The integrable input is a new Pfaffian formula for rectangular frozen-corner probabilities, from which we obtain one-point convergence of the top path to the GUE Tracy–Widom distribution. The probabilistic input is an exact resampling law and accompanying bridge estimates for a strictly ordered midpoint encoding of the paths; these yield tightness and the Brownian Gibbs property for every subsequential limit. Aggarwal and Huang's strong characterization then identifies each such limit as the Airy line ensemble.

In this survey, we review certain types of special \(T_{1}\) spaces and their associated fixed-point theorems. Kupka introduced the notion of a feeble topological contraction, which naturally generalizes Lipschitz contractions defined on metric spaces. Specifically, Kupka established a fixed-point theorem for feeble topological contractions possessing a closed graph within the product of arbitrary \(T_{0}\) spaces. Furthermore, the \(T_{1}\) separation axiom is shown to guaranty the uniqueness of such fixed points. Finally, we discuss peripheral Hausdorff and locally Hausdorff spaces within the context of these fixed-point results.

Connected vehicles use millimeter-wave (mmWave) sidelinks for the data rates cooperative driving demands, and emerging designs select the carrier band from sensed rainfall. We show that this weather awareness is an attack surface: an adversary who spoofs only the rainfall input dictates the victim's carrier frequency, and through it its communication range, without transmitting on the channel. We evaluate the attack in MilliCar, an ns-3 module that runs the selected band as the real 3GPP NR V2X carrier with per-band propagation, beamforming, and blockage. Forcing the band up to 73 GHz holds an eight-vehicle platoon's reliable range at 38 m while the honest baseline doubles it to 82 m; forcing it down to 5 GHz sustains 97% long-range reception but collapses the transport block to a third and quadruples long-range latency to 12.5 ms. We then implement the defense the mechanism implies. Rain loss grows linearly with distance while path loss grows logarithmically, so a receiver that tests measured SINR against the attenuation its reported weather predicts flags force-up with 98% probability within 1.5 s at a 1% false-alarm rate, and re-selection then restores long-range reception from 60% to 75%. The same test is structurally blind to force-down, because the 5 GHz fallback is nearly rain-immune. An advecting rain cell that swings the local rate from 15 to 81 mm/h leaves every result unchanged. Weather-aware band selection therefore requires an authenticated meteorological input; physical cross-checking covers one half of the threat.

The accessible information of any given quantum ensemble quantifies the maximum amount of Shannon information that can be extracted from the ensemble by any quantum measurement. Almost three decades ago, Shor conjectured that the accessible information of any quantum dichotomy, that is, an ensemble of two states, is attained by a projective measurement. Recently, a proof of this conjecture restricted to the qubit case was published by Keil. Here, we conclusively settle this longstanding open problem. First, we show that Shor's conjecture follows, in arbitrary dimension, from a recent result by Fang, Fawzi and Fawzi, and we provide a self-contained, elementary proof. Second, for any given dichotomy and measurement, we provide the explicit construction of a projective measurement that outperforms such a measurement in extracting information from the given dichotomy. Third, while the computation of the accessible information is known to be non-convex in general, we frame the computation of the accessible information of dihotomies as a convex problem.

We establish the existence and uniqueness of solutions to a $(1+1)$-dimensional fractional stochastic Burgers-type equation featuring a Caputo fractional time derivative, a fractional Laplacian, and space-time white noise forcing modified by a Riemann-Liouville fractional integral. From a physical perspective, equations of this type describe non-Markovian dynamics with long-range temporal and spatial dependencies. The main difficulty is that the solution to the linearized equation has low spatial Hölder regularity, rendering the singular nonlinearity $G(u)\partial_x u$ classically ill-defined. The proof relies on two main ingredients: first, combining the Da Prato-Debussche decomposition with rough path theory to rigorously interpret the nonlinear product; and second, deriving Schauder-type estimates for fractional heat kernels given in terms of slowly decaying Mittag-Leffler functions. The nonlocality of the fractional time derivative requires a strong assumption on the initial condition.

In this paper, we tackle the problem of jointly estimating the system states and partially unknown dynamics within distributed sensor-equipped networks, particularly in scenarios where only partial state observations are available. To address this issue, we propose an observer-based dynamic cooperative learning framework incorporating online distributed Gaussian Process (GP) regression, which enables accurate estimation despite incomplete in measurements and deficient GP models. In addition, a novel data collection strategy is introduced, with theoretical conditions ensuring feasible data acquisition. Moreover, we also derive an error upper bound encompassing state estimation and model estimation, leveraging the deterministic error bounds of GPs. Empirical simulations demonstrate the superiority of our approach compared to existing distributed GP-based methods.

A verified Kubo-Martin-Schwinger (KMS) relation, after a drive has ceased, is not sufficient evidence to prove equilibrium. We demonstrate this in a large-$N$ large-$q$ Sachdev-Ye-Kitaev (SYK) quench protocol. Although all post-quench fermion correlators are exactly thermal, a second quench back to the initial Hamiltonian reveals hidden memory of the initial state. It is encoded in correlators connecting to times before the first quench. This memory is an extensive nonequilibrium (NEQ) witness with its decay rate being the Lyapunov exponent $\lambda_L$. Thus $\lambda_L$ sets the rate at which the state becomes effectively indistinguishable from a Gibbs state. Despite being a unitary interacting many-body system, the complete NEQ real-time evolution is obtained exactly in the large-$N$, large-$q$ limit, making the setup ideal for analytically studying thermalization.

Continuous-variable quantum key distribution (CVQKD) is compatible with telecommunication infrastructure, but implementing composable security with experimentally practical resources has remained challenging, particularly for discrete-modulated (DM) protocols. We report the first experimental demonstration of a DM CVQKD system that generates composable secret keys against general attacks with finite-size block lengths on the order of $\sim 10^{6}$ rounds via a quadrature phase shift keying (QPSK) system. Our implementation follows a variable-length, general security framework enabled by modern entropy accumulation techniques and conic optimization, whose experimental pipeline allows real-time operation on near-commercial hardware.

Computational models of auditory physiology commonly target specific responses or stages of the auditory pathway, limiting their ability to integrate findings across experimental paradigms and neural timescales. We present a foundation model of human auditory electrophysiology: a causal neural network trained to map binaural acoustic waveforms directly to high-sample-rate EEG. The model was trained on approximately 250 hours of EEG data from 92 subjects, with varied electrode montages and stimuli spanning tonebursts, speech, and music. We tested whether the model recovered effects of stimulus rate, frequency, and presentation method on auditory brainstem responses (ABRs); subcortical and cortical temporal response functions (TRFs) to continuous speech; and the click-evoked binaural interaction component (BIC). Predicted ABRs and TRFs reproduced established response morphology and stimulus-dependent effects, with model-grand-average correlations falling within the corresponding subject-level human distributions. The model-predicted BIC metrics closely resembled the values reported in the literature. These findings demonstrate that a single audio-to-EEG model can capture auditory physiology across paradigms and timescales, supporting future in silico experimentation and hearing technology applications.

Quantum long short-term memory (QLSTM) models extend recurrent sequence learning with variational quantum circuits, but their optimization behavior can vary substantially across random initializations and temporal contexts. This paper evaluates a recursive QLSTM architecture against a standard QLSTM for one-step-ahead prediction of daily minimum and maximum temperature. Using daily weather observations from Toronto and identical training settings, we compare convergence, predictive accuracy, and generalization across input windows of 8, 16, and 32 days over 20 random seeds. The recursive model consistently reaches a near-optimal test loss earlier, reduces mean absolute error and root mean squared error, and exhibits a smaller generalization gap. These results indicate that recursive quantum feature transformations can improve stability and out-of-sample performance for compact hybrid quantum–classical temporal models.

Word-in-Context (WiC) remains challenging for language models, despite recent progress on lexical-semantic tasks. We hypothesise that this difficulty arises not only from comparing two contextual uses of a word, but also from the absence of an explicit sense inventory that specifies the relevant level of semantic granularity. We evaluate open LLMs on WiC and traditional Word Sense Disambiguation (WSD) under similar settings. We find that providing candidate senses, similar to what is done in traditional WSD, improves WiC performance in all settings. In general, explicit sense information helps models make more consistent and targeted judgements. Human evaluation further shows that many apparent WiC errors reflect label ambiguity or mismatches between model and annotator sense boundaries rather than simple failures of lexical understanding. In particular, results show that LLMs overthink the sense distinction often leading to errors based on overly fine-grained distinctions.

Crystal structure refinement is a fundamental inverse problem in materials characterization, where structural parameters are optimized to reproduce experimental diffraction data. Conventional approaches, such as least-squares and likelihood-based optimization, rely on local search and often struggle with non-convex, noisy, and highly correlated parameter landscapes, particularly when integrating multiple diffraction modalities. Joint refinement of X-ray and neutron data is especially challenging due to their complementary but competing sensitivities, which are typically combined through scalarized objectives requiring manual weighting and leading to suboptimal solutions. We propose CrystalMO-TuRBO, a multi-objective trust region Bayesian optimization architecture for joint crystal structure refinement. The method models X-ray and neutron discrepancies as separate objectives and transforms the problem into a normalized maximization setting. A two-phase optimization strategy is introduced: Phase 1 performs global exploration using parallel trust-region Bayesian optimization across multiple scalarizations to identify promising regions of the parameter space, while Phase 2 conducts localized refinement within a shrinking region to achieve high-precision solutions. This design explicitly separates global search from fine-grained optimization, addressing the unique accuracy requirements of refinement tasks. We evaluate the proposed method on experimentally collected X-ray and neutron diffraction data from single-crystal Ho2Ti2O7. Results demonstrate improved convergence, robustness, and parameter precision compared to classical refinement methods and Bayesian optimization baselines on refinement of a single-crystal pyrochlore material system.

In this paper, we study the modified Korteweg-de Vries-Burgers (mKdVB) equation \[u_t + (u^3)_x = \mu u_{xx} - \kappa u_{xxx}\] with $\mu>0$ and $\kappa>0$. Since the flux is non-convex, the (inviscid) Riemann problem may be solved by a composite wave of a degenerate Oleinik shock and a rarefaction wave. We prove the global existence of solutions and the time-asymptotic stability of the corresponding viscous-dispersive composite wave, up to a time-dependent shift, under small $H^1$ perturbations. The shock strength $\delta_S$ and the rarefaction strength $\delta_R$ need not be small and are restricted only by the explicit ratio $\delta_R\le\frac{5}{18}\delta_S$ and the quantity $\kappa\delta_S^2/\mu^2$ governing the monotonicity of the shock. The stability estimate is uniform with respect to the dispersion strength $\kappa$. As a key ingredient of the stability analysis, we establish structural properties of the degenerate viscous-dispersive shock profile in the monotone regime, including two-sided pointwise bounds on its derivative and the rates at which it converges to its two end states. These decay rates coincide with those of the corresponding purely viscous profile, with constants independent of the dispersion coefficient. In the absence of dispersion, namely, when $\kappa=0$, the time-asymptotic stability of the corresponding viscous composite wave was established by Huang, Wang and Zhang [26]. The present work extends their result to the viscous-dispersive setting and relaxes the restriction on the rarefaction strength. The proof relies on the method of $a$-contraction with shifts (for viscous conservation laws) developed by Kang and Vasseur [32,33,35].

Let $f(n,t)$ be the minimum number of triangles in a tripartite graph with $n$ vertices in each part and minimum degree at least $n+t$. In 1975, Bollobás, Erdős and Szemerédi proved that $f(n,1)=\min\{4,n\}$. They further remarked that it is “very likely” that $f(n,t)\ge4t^3$ for $n\ge5t$. They also proved that $f(n,t) \ge t^3$ for all integers $n \ge t \ge 1$. We construct graphs showing that, for all integers $t\ge1$ and $n\ge3t+2\lceil(1+\sqrt5)t/2\rceil$, \[ f(n,t)\le(1+\sqrt5)t^3+\left(1+\frac1{\sqrt5}\right)t^2. \] Here $1+\sqrt5\approx3.236<4$, and the displayed upper bound is strictly less than $4t^3$ for every $t\ge2$, disproving their proposed bound. We also improve their lower bound $t^3$ by showing that $f(n,t)\ge\frac{12}{5}t^3$ for all integers $t\ge2$ and $n\ge18t^6$.

Current dense visual SLAM systems rely almost exclusively on 8-bit tonemapped Low Dynamic Range (LDR) inputs, limiting their robustness in extreme lighting where shadows and highlights trigger tracking drift and mapping collapse. Conversely, existing raw and High Dynamic Range (HDR) reconstruction pipelines operate strictly offline. They depend on Structure-from-Motion preprocessing and are not suited for large inter-frame motion. We present, to the best of our knowledge, the first online Gaussian SLAM framework that tracks and maps directly on single-exposure 16-bit linear HDR imagery. Our method rests on three core components: an architecture-agnostic HDR Gaussian Splatting module featuring an MLP-free logarithmic parameterization of Gaussian color features; a Reinhard range-compressed photometric objective; and structure-guided spatial gradient weighting. Combined, these components allow our approach to outperform a direct HDR adaptation of MonoGS in both trajectory and reconstruction accuracy, while rendering natively in linear scene radiance for post-rendering processing. The same formulation runs unchanged on standard 8-bit inputs, roughly halving the MonoGS baseline error. Furthermore, our HDR Gaussian module transfers seamlessly to SplaTAM, Gaussian SLAM, and DROID-W, eliminating all tracking failures these systems suffer on challenging illumination sequences. To enable this research, we introduce RawSLAM: a dataset of 10 real-world indoor sequences featuring 16-bit RAW imagery, aligned depth, IMU measurements, and external OptiTrack poses. Code and dataset will be made publicly available soon.