Developers increasingly construct multimodal large language models (MLLMs) by assembling pretrained components,introducing supply-chain attack surfaces.Existing security research primarily focuses on poisoning backbones such as encoders or large language models (LLMs),while the security risks of lightweight connectors remain unexplored.In this work,we propose a novel cross-modal backdoor attack that exploits this overlooked vulnerability.By poisoning only the connector using a single seed sample and several augmented variants from one modality,the adversary can subsequently activate the backdoor using inputs from other modalities.To achieve this,we first poison the connector to associate a compact latent region with a malicious target output.To activate the backdoor from other modalities,we further extract a malicious centroid from the poisoned latent representations and perform input-side optimization to steer inputs toward this latent anchor,without requiring repeated API queries or full-model access.Extensive evaluations on representative connector-based MLLM architectures,including PandaGPT and NExT-GPT,demonstrate both the effectiveness and cross-modal transferability of the proposed attack.The attack achieves up to 99.9% attack success rate (ASR) in same-modality settings,while most cross-modal settings exceed 95.0% ASR under bounded perturbations.Moreover,the attack remains highly stealthy,producing negligible leakage on clean inputs,and maintaining weight-cosine similarity above 0.97 relative to benign connectors.We further show that existing defense strategies fail to effectively mitigate this threat without incurring substantial utility degradation.These findings reveal a fundamental vulnerability in multimodal alignment: a single compromised connector can establish a reusable latent-space backdoor pathway across modalities,highlighting the need for safer modular MLLM design.

Advanced packaging and chiplet-based integration are increasingly adopted to build complex heterogeneous systems beyond the limits of monolithic scaling. While these architectures offer major benefits in terms of modularity, yield, and performance, they also introduce new physical attack surfaces. In this paper, we show that side-channel attacks can be mounted across chiplets within the same package or stack. Our key idea is that a communication-oriented chiplet, originally intended to interact with the external environment through an antenna, an RFID-like element, or another contactless coupling structure, can be repurposed as an internal observation platform. We formalize this threat through a realistic adversary model, describe the corresponding attack principle, and experimentally assess its feasibility. The obtained results demonstrate that signals captured through such a communication-oriented interface can reveal information correlated with the activity of a neighboring victim chiplet.

We establish a fundamental breakthrough in rank-one Drinfeld module arithmetic by deriving explicit formulas over the integral domain $\A = H^{0}(\mathbb{P}^1-P_ρ, \mathcal{O}_{\mathbb{P}^1})$, which generalizes the classical polynomial ring ($N=1$) to the projective line associated with an infinite place of degree $N \geqslant 2$. This fills a longstanding gap by developing a comprehensive parallel to Carlitz module theory foundational in positive characteristic arithmetic for the understudied case of infinite places of degree $>1$. We construct Anderson generating functions for these modules and link them to the Carlitz period via Pellarin's series, exponential torsion modules, and logarithmic deformations. These constructions provide powerful tools for studying such Drinfeld modules and their associated $L$-series, central to modern number theory. A key result reveals a critical distinction from Carlitz theory: the standard Anderson generating function residue formula fails due to Galois group action. We resolve this obstruction by introducing an exponential action, enabling simultaneous study of all twisted exponential functions a major methodological advance. We further show that Anderson generating function computation involves the dual of Drinfeld modules, leading to an appropriate residue formula modification. Notably, our natural approach generalizes to arbitrary Dedekind domains, extending our results beyond $\A$ and opening new avenues in Drinfeld module theory.

A mechanism for additional enhancement of SERS in the nanoparticle-on-mirror scheme is proposed. This new mechanism is based on the use of a substrate made of material with a near-minus-one permittivity. The setup involves a plasmonic nanoparticle in the form of an oblate ellipsoid positioned above the substrate and a Raman-active molecule located between them. In the conventional nanoparticle-on-mirror scheme, the plasmonic dipole resonance frequency coincides with the Stokes frequency of the Raman-active molecule. Consequently, due to the Purcell effect, the molecule's near fields mostly excites a dipole mode in nanoparticle. This dipole moment is many times greater than the dipole moment of the molecule by itself. If the real part of substrate permittivity is near minus one, the image of the nanoparticle dipole moment in the mirror-substrate is a dipole moment pointed in the same direction but approximately $ 1/{\rm{Im}} \varepsilon_{_{\rm{ENZ}}}$ times larger in magnitude. The simultaneous radiation of these two dipoles additionally increases the SERS intensity in $ 10^4$ times.

We report a detailed analysis on the young stellar cluster Berkeley 59 using Fermi-LAT. Using up-to-date source catalog and background models, we found significant extended GeV emission around Berkeley 59, which can be modeled by a radial disk of 1.02 degree radius with a significance of the extension of 10.6 sigma. We investigated the molecular, neutral and ionized gas content and the hadronic origin. The gamma-ray spectrum of Berkeley 59 has a photon index of 2.88. The derived gas mass from H2 and HII around Berkeley 59 is about 289 solar mass. We derived the relationship between cosmic ray acceleration efficiency and diffusion coefficient. Our results suggest that the extended gamma-ray emission originates from cosmic rays accelerated by cluster winds interacting with surrounding gas.

Distributed computational substrates rely on two elementary operations: bundling, the act of populating a shared physical medium with independently retrievable components, and binding, the act of composing components into outputs whose identity depends on their relations. We study these two primitives on the simplest closed substrate carrying a continuous symmetry, a cycle graph of N nodes, in two parameter regimes of a single master equation of motion. The linear regime sorts a temporal input across the substrate's U(1)-organised eigenmodes, providing a feature representation that matches a windowed-FFT baseline at high signal-to-noise ratio and modestly outperforms it for transient signals at low SNR. The Duffing regime activates a cubic mode-mixing operation constrained by the substrate's symmetry into a sparse selection rule on integer wavenumbers, generating shape-dependent harmonic content that the linear regime cannot produce. We identify a single-number observable, $ϕ_0$, that summarises the bound representation's response to input shape, and we analyse its symmetry structure: a $π$-periodicity in the shape parameter is exact, while a time-reversal symmetry that would render $ϕ_0$ degenerate is broken by the substrate's dissipation. The asymmetric status of these two symmetries is what licenses $ϕ_0$ as a meaningful single-number observable; its trajectory across the quotient domain encodes the joint response of binding and dissipation to the input shape. Numerical experiments confirm that $ϕ_0$ retains its information content under additive band-limited noise, with seed-averaged means staying clearly above the symmetric-attractor value down to 0 dB input SNR. The framework is developed on synthetic signals only; extensions to richer substrates, more elaborate drives, and real biological signals are open questions for the work that follows.

We study which outcomes are implementable by disclosing coarse statistics of a data-generating process rather than its full distribution. Players observe data whose joint distribution is only partially known: they know the expectations of finitely many random variables and form beliefs by maximum-entropy inference. We obtain two characterizations. When message spaces are unrestricted, implementable outcomes coincide with jointly coherent outcomes, expanding the set of correlated equilibria. With canonical mechanisms, implementability reduces to a single cross-entropy condition: the target outcome must lie on the cross-entropy level set of some correlated equilibrium that passes through that equilibrium itself. Examples and several classes of games illustrate the reach of the framework.

We prove that every class of Eulerian directed graphs of bounded carving width (equivalently of bounded degree and treewidth) is well-quasi-ordered by strong immersion. In fact, we prove a stronger result, namely that every class of Eulerian directed graphs of bounded carving width, where every vertex is additionally labeled from a well-quasi-order, fixes a linear order on its incident edges, and may impose further restrictions on how the immersion is allowed to route paths through it, is well-quasi-ordered by an adequate notion of strong immersion. To this extent, we develop a framework seemingly suited to prove well-quasi-ordering for classes of Eulerian directed graphs by (strong) immersion and present a first meta theorem in that direction. We complement our results by observing that the class of Eulerian directed graphs of unbounded degree is \emph{not} well-quasi-ordered by \emph{strong} immersion, even if we assume the treewidth of the class to be at most two. We conclude with a dichotomy result, proving for a very restricted class of Eulerian directed graphs of unbounded degree that it is not well-quasi-ordered by strong immersion, but it is well-quasi-ordered by weak immersion.

It has been suggested that families of weakly-coupled AdS vacua with a large-$N$ holographic dual must satisfy non-trivial consistency requirements, which amount to the vanishing of certain cubic couplings, corresponding to (super-)extremal arrangements of scalar operators. While this constraint is known to hold in the simplest incarnation of the DGKT scenario in massive type IIA string theory, i.e. on the $\mathbb{Z}_3\times \mathbb{Z}_3$ orbifold, we find that it is generically violated for type II AdS$_3$ and AdS$_4$ vacua arising from $\mathbb{Z}_2 \times \mathbb{Z}_2 \times \mathbb{Z}_2$ and $\mathbb{Z}_2 \times \mathbb{Z}_2$ orbifolds respectively, including scale-separated solutions and DGKT-CFI-type models. In most cases, however, this can be cured by enlarging the orbifold group to a suitable (non-abelian) extension that projects out precisely those scalar operators that would otherwise participate in the constrained cubic couplings. Our results suggest that consistency of the putative holographic dual imposes a non-trivial restriction on the compactification geometry, indicating in particular that O-planes cannot wrap cycles in distinct homology classes.

A basic model in sequential decision making is the Markov decision process (MDP), which is extended to Robust MDPs (RMDPs) by allowing uncertainty in transition probabilities and optimizing against the worst-case transition probabilities from the uncertainty sets. The class of $(s, a)$-rectangular RMDPs with $L_p$ uncertainty sets provides a flexible and expressive model for such problems. We study this class of RMDPs with a discounted-sum cost criterion and a constant discount factor. The existence of an efficient algorithm for this class is a fundamental theoretical question in optimization and sequential decision making. Previous results only establish a strongly polynomial-time algorithm for $L_\infty$ uncertainty sets. In this work, our main results are as follows: (a)~we show that for any compact uncertainty set, the policy iteration algorithm for RMDPs is strongly polynomial with oracle access to solutions of Robust Markov chains (RMCs); (b)~we present strongly polynomial-time bounds on the policy iteration algorithm for RMCs with $L_1$ and $L_\infty$ uncertainty sets; and (c)~we establish hardness results for RMCs with $L_p$ uncertainty sets for integer $p$ satisfying $1<p<\infty$. Finally, motivated by our theoretical bounds, we present experimental results showing how fast policy iteration converges for RMDPs with $L_1$ and $L_\infty$ uncertainty sets.

Garment refitting, the task of adapting a garment from a source to a target avatar, must preserve the original design features and fine-scale wrinkles, a challenge exacerbated by significant shape variations and varying poses without registration to a shared canonical pose. Existing methods struggle to balance robustness, efficiency, and fidelity of detail: physics-based simulation is costly, data-driven approaches lack generalizability, and geometry optimization in the full vertex space is often ill-conditioned and prone to local minima with unsatisfactory quality. We identify that a fundamental limitation lies in the representation: deforming garments directly in global coordinates couples vertices non-locally, creating a complex and poorly-structured optimization landscape. Therefore, we introduce LoBoFit, a robust refitting method built upon a novel Local Bone Mapping Blending (LoBoMap Blending) representation. Instead of manipulating global vertex positions, LoBoMap Blending expresses garment geometry as a linear blend of its mappings into local bone coordinate frames. This representation is highly expressive and flexible: local bone mappings yield a pose-robust initialization and a well-conditioned parameterization, while blending weights smooth the optimization landscape and broaden the space of plausible solutions for stable convergence with fine-scale detail preservation. The subsequent refinement efficiently resolves collisions and preserves details by optimizing localized residuals, effectively decomposing the complex global deformation into manageable subproblems. Our experiments demonstrate that LoBoFit reliably refits high-resolution, single- and multi-layer garments across avatars with large shape and topological differences, while faithfully preserving intricate wrinkles and the intended fit style, outperforming state-of-the-art methods in robustness and output quality.

We propose and analyze a hybrid qubit-qutrit quantum battery (QB) based on a mixed spin-1/2 and spin-1 system interacting via an anisotropic Heisenberg exchange coupling in the presence of a homogeneous magnetic field. The nonclassical properties of the system are characterized using the l1-norm of coherence and negativity, which quantify quantum coherence and entanglement, respectively. The performance of the quantum battery is evaluated through key indicators such as ergotropy, power, and capacity. Our results reveal that both ergotropy and power exhibit oscillatory dynamics, while the capacity remains constant over time. We further investigate the influence of system parameters and magnetic field strength on both quantum correlations and battery performance, demonstrating that nonclassicality plays a crucial role in enhancing energy-storage efficiency. Importantly, we establish a connection between the theoretical model and an experimentally realizable nickel-radical molecular complex, showing that quantum coherence, entanglement, and efficient energy storage persist even at room temperature. These findings provide a realistic pathway toward the implementation of hybrid qubit-qutrit quantum batteries in solid-state molecular platforms.

Tensor regression is an important tool for tensor data analysis, but existing works have not considered the impact of outliers, making them potentially sensitive to such data points. This paper proposes a low tubal rank robust regression method for analyzing high-dimensional tensor data with heavy-tailed random noise. The proposed method is based on a nonconvex relaxation of the tensor tubal rank within a general optimization framework, which allows for nonconvexity in both the loss and penalty functions. We develop an implementable estimation algorithm and establish its global convergence under some mild assumptions. Furthermore, we provide general statistical theories regarding stationary point, including the rates of convergence and bounds on the prediction error. These theoretical results cover many important models, such as linear models, generalized linear models, and Huber regression, and even encompass some nonconvex losses like correntropy and minimum distance criterion-induced losses. Supportive numerical evidence is provided through simulations and application studies.

Dislocation in (011)-oriented $β$-Ga$_2$O$_3$ substrates grown by the vertical Bridgman method was investigated using X-ray topography (XRT), combined with X-ray reticulography. Transmission XRT reveals dislocations lying on the (001) plane and extending along [010], forming arrays associated with domain boundaries. Dislocations on the (011) plane were also identified but differ from those responsible for line-shaped pits on (001) epilayers. Reflection XRT shows good agreement with transmission XRT and enables classification of dislocation types based on contrast features. Reticulography confirms domain boundaries with misorientation on the order of 1E-5 rad, providing insight into defect formation relevant to epi-growth and device performance.

Large Language Models (LLMs) are transforming recommendation from ranking into a generative task, but industrial deployment remains limited by the high latency of processing long, personalized prompts. Standard prefix caching provides limited benefit because reuse in recommendation workloads is often non-contiguous across user histories and item contexts. We present RcLLM, a distributed inference system for generative recommendation with Beyond-Prefix KV Caching. RcLLM decomposes prompts into reusable blocks and supports large item catalogs with a stratified distributed storage design: compact user-history caches are replicated for zero-latency retrieval, while massive item caches are sharded using similarity-aware placement. To reduce redundant quadratic attention computation, RcLLM combines an affinity-based global scheduler that improves data locality with a selective attention mechanism that corrects approximation errors. Experiments on real-world datasets show that RcLLM reduces Time-To-First-Token (TTFT) by 1.31x-9.51x compared with state-of-the-art prefix caching systems, enabling real-time serving with negligible impact on recommendation accuracy.

Both the level of conservativeness and the computational burden in robust optimization are critically influenced by uncertainty set design. However, contextual side information is rarely exploited in robust dispatch of power systems characterized by irregular data distributions, which hinders the explicit characterization of the relationship between covariates and uncertain parameters. To address this issue, a data-driven method for constructing contextual-aware uncertainty set is proposed in this letter. Based on a conditional Gaussian mixture model, a set of covariates is leveraged as side information to design uncertainty sets tailored to historical data exhibiting irregular distributions. The resulting set is formulated as a union-of-subsets formulation, and a mixed integer linear reformulation is adopted to describe the worst-case realization across all subsets. Finally, the effectiveness of the proposed method is demonstrated through numerical experiments applied to robust unit commitment.

Single-Hessian Gaussian wavepacket dynamics (GWD) significantly reduces the computational burden of Heller's local harmonic GWD, while maintaining comparable accuracy in approximating vibronic spectra. Here, we provide a new, symplectic derivation of the equations of motion of single-Hessian GWD and show that, unlike the local harmonic version, this method conserves the non-canonical symplectic structure on the manifold of Gaussian wavepackets and$-$for bounded dynamics in smooth potentials$-$avoids the drift of energy. Our numerical results suggest that, despite being much more efficient than the local harmonic variant, the single-Hessian GWD exhibits the same $\mathcal{O}(\hbar)$ asymptotic error in averages of observables. To further accelerate numerical simulations, we implement high-order time-stepping geometric integrators that are time-reversible and conserve the norm and symplectic structure exactly, regardless of the time step. In addition, we present explicit expressions for the exact evolution of the width of a single-Hessian Gaussian wavepacket in a general potential, as well as for the exact evolution of the whole wavepacket in a global harmonic potential. Using on-the-fly ab initio Gaussian wavepacket dynamics on the first excited-state surface of ammonia, we numerically confirm the conservation of geometric properties by these integrators and demonstrate that high-order integrators can enhance both accuracy and computational efficiency. We also compute the photoelectron spectrum of the difluorocarbene anion and the absorption spectrum of methylamine, and find that, in comparison with experiment, single-Hessian GWD outperforms global harmonic models and matches the accuracy of local harmonic GWD. Finally, we identify which spectral features are sensitive to the choice of reference Hessian.

The success of machine learning in drug discovery hinges on learning the relationship between a chemical structure and its biological activity. While DNA-Encoded Library (DEL) technology can generate the massive datasets required for this task, its primary signal -- sequencing read counts -- is an indirect and often noisy proxy for true molecular binding affinity. To address the scarcity of public benchmarks for developing robust models that can overcome this data challenge, we introduce CA-DEL, a multi-dimensional public benchmark featuring screens against three homologous carbonic anhydrase isoforms. While recent benchmarks like KinDEL have introduced 3D poses for kinase targets, CA-DEL distinguishes itself by focusing on the selectivity challenge among homologous Carbonic Anhydrase isoforms (CAII, CAIX, CAXII). Unlike benchmarks relying solely on noisy enrichment scores, CA-DEL integrates a rigorous validation set of experimentally determined binding affinities ($K_i$) from ChEMBL, establishing a critical Sim-to-Real evaluation paradigm: training on noisy DEL screens and testing on high-fidelity biophysical data.

Hilbert algebras are the implicative subreducts of Heyting algebras. It is shown that having depth at most n is an equational condition in Hilbert algebras. This generalizes an analogous well-known result in the setting of Heyting algebras.

We present here the Robin boundary condition and its significance in mathematics and physics.

We theoretically study the non-Hermitian topology of magnons with finite lifetimes due to Gilbert damping. By incorporating the spin-wave theory and perturbation theory for the Landau-Lifshitz-Gilbert equation including nonlocal damping terms, we analytically evaluate the spectral winding number for point gaps, which indicates the existence of the non-Hermitian skin effect (NHSE). We find that the NHSE can occur even in the absence of nonlocal damping. In the presence of nonlocal damping along one direction, we show that the winding number for an energy band with a unique minimum is determined from the sign of the wave number at the band minimum. We demonstrate these results using a model that hosts a skyrmion-string lattice as a steady state. We further investigate spin-wave propagation dynamics excited by a magnetic-field pulse and show that the propagation direction changes drastically from band to band depending on the presence of local and nonlocal damping, consistent with the nontrivial winding numbers.

To solve the problem of detecting subspace signals in nonzero-mean clutter, we propose adaptive detectors, based on the strategies of generalized likelihood ratio test (GLRT), Rao test, Wald test, gradient test, and Durbin test. The results show that the detectors based on GLRT, Rao and Wald are structurally consistent with the subspace detectors in zero-means clutter. The analytic expressions for the probability of detection (PD) and probability of false alarm (PFA) of each detector are derived, and two major performance differences in the nonzero-mean clutter scenario are revealed. One is the loss of degree of freedom (DOF), which is reduced by 1 compared with the zero-mean clutter scenario. The second is the loss of signal-to-clutter (SCR) ratio. Simulation and measured data verify the effectiveness of the proposed detectors and demonstrate their practical value in real-world radar systems.

Real-time video surveillance systems store recorded video using digital video recorders (DVRs) and network video recorders (NVRs). To support continuous high-volume video storage, these devices employ specialized, nonstandard file systems that are often proprietary and undocumented. This lack of documentation significantly increases the time and effort required for forensic analysis. In this study, we analyze an undocumented proprietary file system used by Honeywell video surveillance devices-one that, to the best of our knowledge, has not been examined in prior work-and investigate its deletion mechanisms and demonstrate the feasibility of video recovery after deletion. We perform a file system analysis using a binary diffing technique and evaluate three deletion methods supported by the target device: 1) formatting-based deletion, 2) data expiration, and 3) overwrite. For each method, we investigate changes in file system metadata and on-disk data structures and demonstrate the feasibility of video data recovery. Our findings aim to support more efficient and accurate forensic investigations of Honeywell surveillance products and provide foundational insights into the analysis of proprietary file systems used in video recording devices.

Rotating structures are widely observed in engineering applications such as turbomachinary and wind turbine. These rotating structures, particularly for blades made by lightweight materials, can undergo large deformation in operations and display complex nonlinear dynamics under the coupling interaction of geometric nonlinearity, Coriolis effect and centrifugal force. Finite element (FE) methods provide a powerful and accurate modeling approach for capturing the complex nonlinear dynamics for realistic rotating structures, yet its high-dimensionality causes significant challenge to efficient prediction for the nonlinear vibration. Here, we present a non-intrusive spectral submanifold (SSM) model reduction for these FE models of rotating structures. We use COMSOL to establish FE models and simulate these FE models to verify the accuracy of SSM-based reduction. We first compute nontrivial static equilibrium induced by the centrifugal force and then construct non-intrusively SSM based reduced-order model (ROM) anchored at the equilibrium. These SSM-based ROMs enable efficient and accurate extraction of backbone and forced response curves. We use a suite of examples with increasing complexity to demonstrate the effectiveness of the SSM reduction, including a rotating beam, a twisted plate, a rotor with two disks, and an internally resonant fan with three blades. The obtained results also highlight the significant effects of Coriolis force on the nonlinear vibration of rotating structures.

Building on the information-theoretic perspective of P.~D.~Lax [\textit{Proc.\ Sympos., Math.\ Res.\ Center, Univ.\ Wisconsin}, 1978], we establish a two-sided quantitative compactness estimate for numerical solutions of scalar conservation laws with a uniformly convex flux, expressed in terms of Kolmogorov $\varepsilon$-entropy. We prove that, under specific grid constraints, conservative, monotone finite-difference schemes satisfying a discrete one-sided Lipschitz condition (OSLC) preserve the $1/\varepsilon$ Kolmogorov entropy scaling of the corresponding exact entropy solution set, matching the bounds obtained by De~Lellis and Golse [\textit{Comm.\ Pure Appl.\ Math.}\ \textbf{58} (2005)] and by Ancona, Glass, and Nguyen [\textit{Comm.\ Pure Appl.\ Math.}\ \textbf{65} (2012)]. Specifically, the upper bound follows from the discrete OSLC, while the lower bound relies on a uniform approximation argument on a bounded-variation precursor class. Our results show that prototypical first-order methods are high-resolution in Lax's sense. Finally, we abstract the lower bound mechanism into a general transfer principle, discuss implications for information recovery via post-processing, and indicate directions for future work.

We study unbiased estimation under Bregman losses and develop an extension of the classical theory of uniformly minimum variance unbiased estimators (UMVUEs). Exploiting bias--variance-type decompositions for Bregman divergences, we consider two natural loss functions, $D_φ(θ,\hatθ)$ and $D_φ(\hatθ,θ)$, and their corresponding notions of unbiasedness. We show that the latter formulation reduces to the classical setting, whereas the former yields a different framework in which unbiasedness is characterized in the dual space induced by $\nablaφ$. For the nontrivial case, we establish analogs of the Rao--Blackwell and Lehmann--Scheff{é} theorems, providing a systematic construction of type-I Bregman UMVUEs.

Channel state information (CSI) is critical for multi-input multi-output (MIMO) orthogonal frequency division multiplexing (OFDM) system. Pilot-based channel estimation methods suffer from high pilot overhead and low channel acquisition quality, while pilot-free approaches typically impose impractical demands on positional or environmental information precision. This paper proposes geometry-aided channel deduction (GCD), which leverages readily available geometric information to assist channel acquisition. The environmental map and base station position together constitute the scenario geometry, which can provide geometric channel features through ray tracing. To obtain the complete channel, the user first retrieves approximate geometric features by performing neighborhood searching within a pre-extracted geometric feature set, and then converts them into pseudo channels through a priori designed feature alignment. These pseudo channels serve as contextual prompt, providing supplementary channel features beyond those derived from pilot-based estimate. Finally, a neural network fuses these pseudo channels with partial estimate to generate the complete channel. Comprehensive experiments validate the superiority of our method, which achieves the leading accuracy in channel acquisition under sparse pilot conditions, demonstrates strong generalization capabilities in new scenarios and dynamic environments, and exhibits robust resilience against user position errors and non-ideal environmental information.

Transition edge sensors (TESs) are superconducting energy-resolving microcalorimeters that have demonstrated low background rates as well as quantum efficiencies close to unity for photons at optical and near-infrared wavelengths. This makes these detectors well suited for rare-event searches. We report on the comprehensive characterization of a two-channel detector module consisting of two tungsten TESs optimized for the detection of photons with a wavelength of 1064nm. The devices achieve a system detection efficiency of $(86\pm1)$%, an energy resolution better than 7%, and a background dark-count rate of photon-like events below 6mHz when coupled to an optical fiber. Using an unbinned likelihood framework, we find the dark count rate to be compatible with blackbody radiation from the room-temperature laboratory environment. Thanks to the energy resolution of the TESs, we show that it is possible to detect monochromatic signals at 1064nm with photon rates $\geqslant 2.7_{-0.6}^{+0.8} \times10^{-5}$Hz, which corresponds to a power of $\geqslant(5.0_{-1.1}^{+1.4})\times10^{-24}$W, within 20 days of measurement time at the 5$σ$ confidence level. This makes our detectors well suited for searches for hypothetical axions and axion-like particles with experiments such as the Any Light Particle Search II (ALPS II) or axion interferometers. The developed methodologies are not only applicable to axion searches, but are also relevant for rare-event searches with TESs in general.

This paper discusses wrongful convictions in a medical setting, focusing on nurses. Common features are lack of strong direct evidence: the nurse was never seen doing anything wrong. There is no DNA evidence of tampering of apparatus or medications by the nurse. There is no CCTV footage showing suspicious actions. Analysis of medical records at the time led coroners to issue certificates of natural deaths, and most events were not, at the time, thought suspicious by hospital staff. There is no confession and the nurse consistently asserts they are completely innocent. There is no evidence of earlier psychopathic behaviour. Instead, private writings (e.g., in a diary) are interpreted by the prosecution as a confession; mundane behaviour is given a sinister interpretation. Motive remains speculation. The main evidence is statistical: a spike in deaths or collapses and a statistical association with a particular nurse. There is forensic evidence which suggests one or two patients might have been harmed by administration of medication much used in the hospital, and even legitimately used earlier in the care of the alleged victims. Police investigations are driven by the hospital consultants who were clinically responsible for the patients allegedly killed or harmed by the nurse.

The continued improvement of large language models (LLMs) increasingly depends on eliciting high-quality, user-generated data, yet such data are costly to provide and often withheld due to privacy and effort concerns. This creates a fundamental design challenge: how to incentivize data contribution when model improvements require coordinated, threshold-level inputs, while contributions remain privately costly and partially reversible. We develop and theoretically analyze incentive mechanisms for user data contribution that explicitly account for threshold effects and reversibility, focusing on how subsidies and withdrawal rights can be jointly designed to overcome coordination failure. As a natural benchmark, we first consider subsidy-based incentives, under which users respond to posted payments with privately optimal floor contributions. These decentralized responses may fall below the improvement threshold, resulting in subsidy expenditure without model improvements. We then analyze mechanisms with withdrawal rights, in which users report costs, the provider centrally assigns contribution burdens, and users may withdraw before training. We prove that combining cost reporting with personalized assignment can eliminate inefficient provision by ensuring that data are collected only when improvement is sustainable, converting infeasible instances into a null outcome rather than subsidy leakage. Finally, we compare two withdrawal protocols. The simultaneous protocol can achieve lower total cost, while the small-first sequential protocol better incentivizes participation, encouraging greater data provision and thereby increasing the probability of crossing the improvement threshold.