Papers

The IO500 benchmark provides a common basis for evaluating high-performance computing (HPC) storage systems, while its growing submission repository also offers an opportunity to study performance behavior across systems and time. This work presents a repository-scale characterization of 294 post-reset IO500 submissions from 116 sites spanning 2019–2025. We combine descriptive and temporal analysis with phase-level correlation, scale-sensitivity analysis, and benchmark-log examination to investigate how performance characteristics represented by individual benchmark phases relate to composite rankings. The results show strong relationships among phases measuring similar I/O behavior, but weaker and scale-sensitive relationships between bandwidth and metadata performance. Temporal and log-level analyses further reveal information that is not apparent from aggregate scores alone. These findings demonstrate the value of analyzing IO500 results beyond leader-board rankings and identify important considerations concerning system scale, benchmark execution, and repository provenance when using community benchmark data for comparative and longitudinal HPC performance analysis.

Nash–Williams's orientation theorem converts undirected edge-connectivity into directed connectivity. We extend it to prescribed outdegrees modulo $q$, whose residues must sum to the number of edges modulo $q$. For integers $q,k\ge1$, every such prescription has a realization with at least $k$ edges in each direction across each nontrivial cut if the graph is $(2q+2k-2)$-edge-connected or contains $q+2k-1$ edge-disjoint spanning trees. The method combines a cycle-space approximation bound with rounding that preserves the imbalance across every cut. Fractional spanning-tree packing gives an exact criterion: every compatible prescription has a strongly connected realization if and only if the fractional packing number exceeds $q$. This yields a polynomial-time recognition test. With no directed connectivity requirement, $q$ disjoint spanning trees suffice, proving Hasanvand's prescribed-residue conjecture. For every odd prime $p$, this proves the conjectured statement that the union of $p$ graphic bases over the field with $p$ elements is an additive basis. For every integer $p\ge5$, the theorem and the counterexamples of Han, Li, Wu and Zhang give the exact edge-connectivity threshold $4p+2$ for a modulo-$(2p+1)$ orientation. We also show that every six-edge-connected graph has circular flow number strictly below three, as conjectured by Li, Thomassen, Wu and Zhang. For three-flow-critical graphs on $n$ vertices and $m$ edges, we improve the previous general bound $m\le4n-10$ to $m\le3n-6$, and to $m\le3n-7$ when $n\ge5$. Further applications include Seymour's six-flow theorem, star decompositions, antisymmetric flows, and signed modular orientations.

Predicting transient urban winds is fundamental to understanding urban microclimates and designing climate-resilient cities. Building-resolving large-eddy simulation produces detailed incompressible urban wind fields at substantial computational cost for each layout. Neural surrogates offer a faster alternative by learning to predict the evolution of velocity fields. However, minimizing velocity prediction error does not guarantee local mass conservation and wall impermeability, which together define kinematic admissibility. This limitation stems from an unconstrained output representation: geometry conditioning guides predictions but does not restrict them to admissible velocity fields. Correcting boundary violations in these outputs changes the flux balance in adjacent fluid cells and may consequently compromise local mass conservation. To address the challenge, we propose Sculpt, a nested potential framework that builds the coupled, geometry-dependent constraints directly into its parameterization. This nested parameterization generates divergence-free velocity updates through the discrete curl of a volume vector potential on the native three-dimensional staggered grid. A shared scalar potential constrains the vector potential's boundary values so that the same operator also enforces impermeability, without a per-step pressure projection. Because backpropagation through this curl attenuates large-scale gradient signals, we parameterize the volume potential at multiple resolutions to better capture large-scale flow structures. We introduce UrbanWindFlow, an LES dataset spanning urban morphologies and inflow conditions, to evaluate accuracy and kinematic admissibility together.

Large-scale pretrained transformer models have achieved state-of-the-art performance across diverse machine translation tasks, including multilingual settings. Knowledge distillation has emerged as a sustainable approach for model compression, transferring knowledge from large teacher models to smaller, more efficient student models. Similarly, quantization, which reduces the numerical precision of model weights and activations (e.g., from 32-bit to 8-bit representations) is widely used to accelerate inference, enabling models to run several times faster during deployment. However, both techniques face limitations when applied to specialized domain data, particularly under low-resource conditions. In knowledge distillation, the effectiveness of transfer is often constrained by the scarcity of domain-specific parallel data, while quantization can lead to performance degradation as bit precision decreases. In this work, we investigate the combined application of knowledge distillation and quantization for French-to-English biomedical translation, a domain characterized by specialized terminology and limited parallel resources. We develop and compare multiple fine-tuning strategies to adapt compressed student models to this challenging setting. Our experiments demonstrate that a collaboratively distilled and quantized student model achieves a 69% reduction in size, a 98.21% increase in inference speed, and a 98.46% reduction in CO2 emissions compared to the original baseline all without sacrificing translation quality. These results indicate that jointly optimized compression techniques can yield efficient, high-performance models suitable for translation service providers operating under resource constraints.

Recent video world models have witnessed the paradigm shift from single-agent to multi-agent involvements, which can reveal more complicated dynamics and cross-agent interaction in the real world. However, existing approaches commonly adopt implicit inter-agent communications via cross attention, which lack explicit geometry constraints and unified 3D state, thereby leading to poor multi-view consistency and struggling with recovering out-of-sight agents. In addition, most of them assume a static background, failing to represent uncontrolled background dynamics. To address these problems, we propose Artemis: a geometry-grounded multi-agent world model with explicit memory sharing. An explicit 3D world map is reconstructed from multi-agent observations to enforce a unified 3D state across agents, offering high cross-view consistency. Specifically, an action-guided geometric injection module is developed to simultaneously render decomposed foreground-background control maps, which are then injected into a diffusion transformer through a designed GeoAdapter block. Compared to previous methods assuming static-only background, our GeoAdapter can also distinguish uncontrolled non-agent dynamics, which are conditioned on their own multi-frame history positions to provide consistent motion cues. Keyframes selected from progressive video rollouts are used to progressively update the reconstructed 3D world maps. To effectively capture complex dynamic patterns, we curate a novel dataset sampled from the CARLA simulator called MA-CARLA. Extensive experiments demonstrate the superiority of our proposed method in terms of visual fidelity and cross-view consistency in the generated videos. In addition, our Artemis can support simultaneous multi-modal rollouts with both 2D video and 3D point map maintenance, scale to scenarios beyond two agents and multi-camera setting.

Container placement in data centers must simultaneously minimize power consumption and maximize affinity preferences, while satisfying multi-resource capacity and anti-affinity constraints. Traditional approaches typically rely on fixed rules, which lack adaptability to dynamic cluster states and are difficult to extend for adaptive decision-making. On the other hand, meta-heuristic methods, although more flexible, are often computationally expensive, slower and prone to getting trapped in local optima. In this work, we propose an adaptive and efficient approach based on an LLM-driven multi-agent collaboration framework, where four specialized agents operate in a closed-loop ReAct cycle at each placement step. A Power Consumption Agent and an Affinity Agent debate over competing objectives, while a Placement Agent resolves conflicts using Tolerance-Gated Arbitration (TGA). A Rearrangement Agent further refines decisions through the Monotone Improvement Rule (MIR) in a post-placement refinement phase. All the agents reason over deterministic, feasibility-filtered candidate tables provided by the environment, ensuring that every proposed action inherently satisfies hard constraints. Based on experimental evaluation using the Google Cluster Trace with a configuration of 100 applications and 25 machines, the proposed framework consumes 8.36% less power and achieves 39.30% higher affinity than the power-greedy baseline. Ablation studies confirm that each architectural component, the multi-round debate mechanism, the TGA in the Placement Agent, and the MIR in the Rearrangement Agent, contributes meaningfully to the overall performance, with removal of any single component degrading both power consumption and affinity satisfaction.

Resistive random-access memory (ReRAM)-based Processing-in-Memory (PIM) accelerator is a promising platform for processing massively memory intensive matrix-vector multiplications of neural networks in parallel domain, due to its capability of analog computation, ultra-high density, near-zero leakage current, and non-volatility. Despite many advantages, ReRAM-based accelerators are highly error-prone due to limitations of technology fabrication that lead to process variations and defects. These limitations degrade the accuracy of Deep Convolutional Neural Networks (Deep CNNs) running on PIM accelerators. While these CNNs accelerators are widely deployed in safety-critical systems, their vulnerability to fault is not well explored. In this paper, we have developed a fault injection framework to investigate the vulnerability of large-scale CNNs at both software- and hardware-level of inference phases. Faulty ReRAM devices are another reliability challenges due to significant degradation of classification accuracy when CNN parameters are mapped to the accelerators. To investigate this challenge, we map the CNN learning parameter to the ReRAM crossbar and inject faults into crossbar arrays. The proposed framework analyzes the impact of stuck-at high (SaH) and stuck-at low (SaL) fault models on different layers and locations of CNN learning parameters. By performing extensive fault injections, we illustrate that the vulnerability behavior of ReRAM-based PIM accelerator for CNNs is greatly impressible to the types and depth of layers, the location of the learning parameter in every layer, and the value and types of faults. Our observations show that different models have different vulnerabilities to faults. Specifically, we show that SaL further reduces classification accuracy than SaH.

Diffusion-based world models can generate and predict trajectories in high-dimensional dynamical systems, but predictive accuracy does not imply that their latent coordinates recover the underlying state variables or causal interactions. We ask whether a frozen pretrained diffusion model can be equipped with identifiable coordinates without retraining its generative backbone. We show that auxiliary-variable nonlinear ICA guarantees can be transferred to Contrastive Diffusion Alignment (ConDA), which learns only a lightweight alignment map on top of frozen diffusion latents. Under standard TCL/GCL assumptions, the aligned representation identifies latent dynamical states up to permutation and componentwise invertible transformations, preserves the latent dynamic structural causal model, and reduces lagged graph recovery to transition-Jacobian sparsity. We evaluate TCL-, GCL-, and CEBRA-based ConDA against TDRL, CaRiNG, IDOL, temporal SuaVE, and iVAE across physical and robotic video systems. TCL and GCL achieve near-perfect blockwise state recovery and competitive lagged graph recovery, including exact recovery in a simulated falling-body system. In a simulated bipedal robot, learned dynamics recover the sign and temporal structure of responses to held-out control perturbations. These results show that a frozen generative diffusion model can be equipped with coordinates that are identifiable, structurally interpretable, and useful for analyzing intervention-relevant dynamics.

As large language models (LLMs) are increasingly integrated into engineering workflows, students require hands-on experience to learn how to collaborate with them critically. This paper presents a scalable gamified workshop designed for engineering Master's students to practice human-AI collaboration in navigation planning. Using a mobile web interface across 10 workshop sessions, a total of 226 students wrote prompts for a non-reasoning and a reasoning LLM to solve grid-based navigation tasks of increasing complexity. The system returned robot-executable plans, trajectory visualizations, and automated scoring, culminating in a live demonstration on a Boston Dynamics Spot robot. In a post-workshop questionnaire, 81.5% reported substantial learning and 91.0% reported high engagement. Analysis of the submitted prompts revealed that students changed their strategies from step-by-step instructions for the non-reasoning LLM toward providing higher-level guidance for complex problem-solving tasks. We conclude that such interactive simulation-to-reality environments are viable for teaching the verification and collaboration skills necessary for responsible LLM use in engineering. Code is available at: https://github.com/renchizhhhh/LLM-robotics-workshop

The Offline AI Modules workstream enables practical, low-power, and community-accessible deployment of voice-first AI systems that operate fully offline. Designed for African language communities where speech is the dominant mode of interaction and internet connectivity is unreliable or absent, the workstream delivers three reinforcing components: a modular voice-first offline architecture, a low-cost hardware reference bill of materials, and a reproducible quantization and a reproducible quantization and benchmarking pipeline for instruction-tuned language models in the 2-5B parameter class. This paper presents the first end-to-end benchmark evaluation of the stack across two hardware tiers: an NVIDIA Jetson Orin NX (TierB) and a Raspberry Pi5 (TierA). Three instruction-tuned models are evaluated across four quantization formats, assessed for deployment metrics (decode throughput, chat latency, memory, power) and multilingual quality (topic classification accuracy on MasakhaNEWS across English, Hausa, Igbo, Nigerian Pidgin, and Yoruba; per-language perplexity drift). Speech recognition is evaluated using Ethio-ASR on Amharic and Oromo across both tiers. The principal finding is that Q4_K_M quantization represents the best size-to-quality trade-off for deployment on both tiers: gemma-4-E2B-it achieves 28.8t/s decode throughput and 89.2% topic classification accuracy at Q4_K_M on TierB, while all three models run within the 16GB memory budget on TierA.

Multi-person 3D pose estimation with WiFi channel state information (CSI) is challenging because reflections from different people overlap without directly identifying individual joints. Existing masked embedding objectives capture wireless relationships without explicit pose supervision, while structured decoders can retain coordinate errors. We propose WiSPER, a two-stage framework combining pose-aware predictive pretraining with conditional residual flow refinement. Pose-Aware Masked Embedding Learning (PAMEL) couples masked latent prediction with auxiliary pose-set supervision on the same CSI context, guiding the encoder toward joint localization from partial observations. Residual Flow refinement with Transformer (ReFT) generates a set of pose candidates to accommodate a variable number of people and refines each candidate through a conditional flow guided by its coarse coordinates and per-joint decoder features. Both stages use paired CSI and pose annotations during training, while inference requires only CSI. Experiments on the PiW3D dataset show that WiSPER achieves an overall mean per-joint position error of 63.72 mm, a 40.0% reduction relative to WiFi-JEPA. For experiments with two and three people, WiSPER reduces MPJPE by 42.1% and 38.1%, respectively. Pose-supervised pretraining configurations obtain lower errors than CSI-only JEPA, and enabling the trained residual refiner reduces overall MPJPE by 13.8-15.6% across the evaluated configurations.

Accurate skull-to-array registration is essential for transcranial focused ultrasound (tFUS) because CT-derived skull models must be aligned with the physical-array coordinate system to support patient-specific aberration correction. Current workflows typically use MRI and/or external tracking to establish this spatial relationship, adding imaging infrastructure and coordinate transformations to the treatment workflow. Here, we describe an MRI-free CT-to-ultrasound registration framework using a hybrid hemispherical CMUT tFUS array that combines a sparse therapeutic array with an integrated CMUT US-Cross imaging array. The therapeutic array provides pulse-echo measurements of skull location, while the US-Cross array provides local high-resolution images of the skull surface. The approach was evaluated using a CT-derived outer cortical skull surface across a 25-pose matrix with 10-mm lateral offsets between the skull and array centers and +/-5 degree angular perturbations. Five randomized radial-noise seeds were applied to therapeutic pulse-echo measurements to model measurement or element-placement errors, yielding 125 pose-seed realizations per noise and active-element setting. US-Cross imaging features were generated using Field II and delay-multiply-and-sum plane-wave compounding. Four registration modes were evaluated: Therapeutic only, US-imaging only, Sequential, and Fused registration. Fused registration yielded the best overall performance, with a median total displacement error of 0.271 mm [0.169-0.420] and a median rotation error of 0.904 degrees [0.432-1.639]. These results support an all-ultrasound-based approach using a hybrid CMUT tFUS array for MRI-free CT-US skull registration.

Large Language Models (LLMs) have achieved remarkable capabilities but remain vulnerable to jailbreak attacks that elicit harmful or unsafe outputs. Existing safety alignment approaches, including Supervised Fine-Tuning (SFT) and Reinforcement Learning from Human Feedback (RLHF), often require substantial attack-specific supervision and computational resources, while remaining susceptible to shallow safety alignment and over-refusal. To address these challenges, we introduce SSRFT(Supervised Safe-Role Fine-Tuning), the first framework that reformulates safety alignment as the internalization of a predefined safe role. SSRFT constructs a Safe-Role Question-Answer (SRQA) dataset from psychometric questions, limited jailbreak prompts, and a safe-role description. Role-consistent responses are synthesized, validated, and expanded into diverse scenarios, enabling models to internalize safety-oriented values and principles rather than explicit refusal patterns. Experiments across multiple Base and Instruct models show that SSRFT achieves more robust and generalizable safety alignment than standard SFT. SSRFT shows substantially greater robustness to prefilling attacks and better generalization to unseen jailbreak domains, while reducing over-refusal on benign queries and preserving the model's general capabilities. These results establish safe-role internalization as an effective alternative to refusal-centric safety alignment. Warning: This paper contains examples of harmful and toxic language.

For a fixed integer $k\ge 2$, let $kK_3$ denote the vertex-disjoint union of $k$ triangles. A recent fixed-size spectral theorem of Das and Yamini asserts that, for all sufficiently large $m$, every $kK_3$-free graph $G$ of size $m$ satisfies $\lambda(G)\le (k-1)+\sqrt{m-k(k-1)},$ and equality holds if and only if $(2k-1)\mid m$ and $ G\cong \bigl(K_{2k-1}\vee qK_1\bigr)\cup tK_1,\, q=\frac{m}{2k-1}-(k-1) $ for some $t\ge 0$. They explicitly posed the open problem: Let $k\ge 2$ be fixed and $\ell$ be a residue in $\{1,\dots,2k-2\}$. For all sufficiently large integers $m\equiv \ell\pmod{2k-1}$, determine the exact value of $\max\bigl\{\lambda(G): e(G)=m,\ G\text{ is }kK_3\text{-free}\bigr\}, $ and characterize all graphs attaining this maximum. In this paper, using the positive-defect version of the bounded-core method for divisible sizes together with several new ideas developed in this paper, we give a complete solution to the aforementioned open problem.

Reconstructive spectrometers (RSs) shift spectral analysis from hardware-heavy dispersion to encoding-decoding schemes. In the speckle-based RSs, spectra are encoded into disordered intensity patterns and then recovered algorithmically. However, characterizing such spectral-encoding transmission matrices (TMs) typically demands empirically complex calibrations, which rely on expensive tunable light sources to generate known wavelengths. In this work, we demonstrate a fully calibration-free RS that leverages conical phase modulation from an axicon. The target spectra are encoded into a Bessel-form intensity profile, enabling computational construction of a deterministic TM and thus completely removing the need for optical calibration. With this computer-generated TM, broadband spectral reconstruction is experimentally achieved across a 650 nm range (1050-1700 nm)–nearly three times the bandwidth of existing calibration-simplified RSs–while a spectral resolution of 2.9 nm remains comparable to that of these counterparts. Compared with an RS requiring optical calibration for equivalent performance, our solution achieves an approximate 36-fold cost reduction.Unlike most speckle RSs, the axicon RS requires no fixed input polarizer, further enhancing energy efficiency and simplicity through polarization-insensitive operation. To the best of our knowledge, this is the first reported experimentally calibration-free RS, featuring superior simplicity, practicality, and cost-effectiveness for practical applications.

We investigate equatorial timelike motion of a massive test particle and the associated gravitational radiation around a Kerr black hole surrounded by perfect fluid dark matter. In the presence of a dark matter fluid, the radius of the innermost stable circular orbit decreases, while the specific angular momentum is modified and the specific energy increases as the dark matter (DM) parameter, $\alpha$, increases. We compute periodic timelike orbits using the rational frequency ratio $q$, with the $(z,w,v)$ classification, which is modified by the presence of DM while preserving the topological class. We analyze the gravitational waveforms corresponding to these periodic orbits and find that dark matter modifies the timing and distribution of the oscillatory features of the orbits. For a representative extreme-mass-ratio system, the corresponding characteristic strain is concentrated mainly in the millihertz (mHz) frequency band relevant to LISA. These results establish a direct connection between the dark-matter-induced modifications of strong-field orbital dynamics and the corresponding gravitational-wave signatures.

Systematic reviewers who use a model to read randomized trials need each answer and a measure of how far to trust it. If the model is calibrated, so that its stated probabilities match how often it is right, a reviewer can accept its confident answers and check the rest by hand. This paper fine-tunes Qwen3-4B-Base with low-rank adapters and a classification head for evidence inference (Lehman et al., 2019; DeYoung et al., 2020): given a trial's article and a prompt that names an intervention, a comparator and an outcome, classify the result as a significant increase, a significant decrease or no significant difference. The articles of Evidence Inference 2.0 come from the open-access subset of PubMed Central, which includes articles under non-commercial and no-derivatives licenses, so the released model was fine-tuned only on the 1,431 of 2,657 training articles under CC BY, CC0 or the public domain; it is released under the Apache License 2.0 and runs on a laptop. On the test split, its expected calibration error is 0.017 and its log loss is lower than that of Gemma 4 31B-it given the same input zero-shot, meeting criteria registered on the Open Science Framework before testing. Its macro-F1, 0.925, is above every published result on that split that I found, though those systems used other inputs, scoring rules and prompt filters, and the test labels are public. Accepting its 80 percent most confident answers leaves an error rate of 2.6 percent among them. Context ablations show that the answers depend on the article, but the title alone carries some signal, which an informative title or recall of the trial could explain.

Vision-language models (VLMs) have achieved strong performance in multimodal reasoning, yet they remain prone to generating plausible but incorrect answers. Self-verification offers a practical way to improve answer reliability without relying on external judges, but existing methods typically depend on a single verification criterion or fixed prompt, resulting in incomplete and unstable reliability estimates. We first systematically analyze how verifier capability and prompt design affect verification performance. Our findings show that stronger verifiers provide more reliable judgments, while verification performance is highly sensitive to prompt choice, with no single prompt consistently dominating across tasks. Guided by these findings, we propose MOTIVE, a Multi-View Self-VerificatiOn wiTh RelIability-Guided SelectiVE Rethinking framework for reliable multimodal reasoning. MOTIVE evaluates each candidate answer from complementary verification perspectives and learns a correctness-aligned reliability score through correctness-grounded multi-view verification learning. During inference, this score governs an accept-or-rethink decision, allowing reliable answers to be returned directly while uncertain ones trigger history-guided rethinking. Extensive experiments across diverse multimodal benchmarks and VLM backbones demonstrate that MOTIVE consistently outperforms strong self-verification and self-correction baselines. Further results show that reliable verification improves accept-or-rethink decisions and reduces unnecessary reasoning turns, enabling more reliable and efficient self-verification without an external judge.

We prove that the categories of strictly unital (resp.\ non unital) A$_{\infty}$-algebras and A$_{\infty}$-categories, with strictly unital (resp.\ non unital) A$_{\infty}$-morphisms and A$_{\infty}$-functors have no coequalizers.

In few-shot industrial anomaly detection, the few normal target images provide no direct defect supervision, making anomaly prompts difficult to learn from these samples alone. Some vision-language methods therefore use manually specified descriptions to supply explicit anomaly semantics. However, constructing these descriptions requires product-specific effort, and their effectiveness depends on prompt selection. We propose Anchor and Adapt, a two-stage prompt learning framework that separates the acquisition of anomaly semantics from adaptation to target normal appearance. Stage I learns transferable normal and abnormal anchors from annotated auxiliary data. Stage II keeps these anchors fixed and adapts an additional normal branch using the few target normal samples. The inherited and adapted normal branches jointly characterize target normality, with text-anchor regularization encouraging consistency with the generic normal prior and separation from the abnormal anchors. This design retains learned anomaly knowledge while reducing dependence on category-specific anomaly templates, without requiring synthetic anomaly generation. Cross-dataset experiments between MVTec-AD and VisA under 1-, 2-, and 4-shot settings demonstrate competitive detection and localization performance. Controlled ablations assess the roles of transferred anchors, asymmetric adaptation, dual-normal representations, and anchor regularization.

Privileged 3D supervision uses additional geometric information during training to guide RGB-based visuomotor policy learning, without requiring geometric inputs at deployment. However, low reconstruction error does not ensure that visual representations capture geometry useful for control. We identify three limitations that weaken this supervision: proprioceptive shortcuts, dominant-view reliance, and reconstruction objectives dominated by task-irrelevant geometry. To address these limitations, we propose Latent Encoding with Aligned Privileged Geometry (LEAP). Our framework uses an auxiliary decoder to reconstruct point clouds within the manipulation workspace from visual features alone, while retaining proprioception for action prediction. Alongside full reconstruction, we introduce wrist-view dropout and partial reconstruction targets matched to the retained wrist, encouraging the encoder to capture complementary local geometry. The auxiliary decoder is removed at inference, leaving only RGB observations and robot state as policy inputs. Extensive experiments on RoboTwin, ManiSkill, and real-world tasks demonstrate consistent and substantial improvements over Diffusion Policy and ACT, with only a small increase in parameter count and no reduction in inference speed.

RGB-D semantic segmentation has made notable progress by fusing RGB and Depth, yet mainstream models still learn features almost exclusively from pixel-level supervision, lacking direct high-level semantic constraints. This raises a central question-can external knowledge such as language priors inject stronger semantic discriminability into mainstream RGB-D segmentation models. We present DTFormer, a novel tri-modal (RGB-D-Text) semantic segmentation framework. At its core is Text-guided Semantic Alignment Module (TSAM) that first encodes textual cues into a set of semantic prototypes and then explicitly aligns multi-modal RGB-D features with these prototypes at multiple encoder and decoder layers. This design imposes strong semantic regularization on representation learning, guiding the network toward more discriminative features. Extensive experiments on multiple benchmarks show that DTFormer delivers consistent gains while remaining simple and efficient. Our results demonstrate that explicit semantic alignment offers an effective and practical route to improving RGB-D semantic segmentation. The code will be released upon acceptance.

As electron sources are reduced in size, fabricated in new materials, or pushed for performance, the analytical formulations for thermo-field electron emission, namely Fowler-Nordheim and Murphy-Good equations and their JWKB-based extensions, are increasingly applied out of the range of conditions for which they were derived. We present GETELEC-3, an open-source code that replaces them with a direct numerical solution of the one-dimensional time-independent Schrödinger equation (1D-TISE). Several methods to solve the 1D-TISE have been implemented to test for speed, robustness, and solution's uncertainty. The Noumerov algorithm is found to be quickest, with the solution uncertainty below 1%, and together with an optimised kernel that advances all electron energies in parallel, the cost of calculating the current density is reduced to a few milliseconds - comparable to the JWKB evaluation. To further boost the calculation speed, especially in complicated barriers, a neural network has been trained on the residual of the logarithm of the semiclassical reference and it reproduces the current density better than 0.4% at a fraction of the computational cost. The code is developed for metals and semiconductors, accepts arbitrary potentials and density of states, and comes with a user-friendly interface and compiled executable to democratise rigorous electron emission calculations and data analysis.

This paper investigates the dielectric breakdown behavior of Litz wire insulation subjected to repetitive fast-edge bipolar square-wave voltages representative of wide-bandgap (WBG) power converter stress. Two conductor sizes, 14 AWG and 16 AWG, were tested in a twisted-pair configuration over frequencies ranging from 1 kHz to 30 kHz using a controlled high-voltage pulse generator with 200 ns rise time. The breakdown voltages obtained at each frequency were statistically analyzed using the weighted two-parameter Weibull method in accordance with IEEE Std 930-2004. Between 1 and 30 kHz, the mean breakdown voltage decreases by approximately 10.7% for the 14 AWG construction and 20.0% for the 16 AWG construction. The 14 AWG specimens exhibit a modest increase between 1 and 5 kHz followed by a decline, whereas the 16 AWG mean decreases across the tested frequencies. The Weibull shape parameter varies nonmonotonically with frequency, characterizing differences in the dispersion of the measured breakdown voltages. These findings provide construction-specific breakdown data relevant to Litz-wire insulation assessment in high-frequency power electronic applications. Possible contributions from charge accumulation, localized discharge activity, and dielectric heating are discussed, but were not directly verified in this study.

Modeling the channel in a single-antenna fluid antenna system (FAS) using extreme value distributions (EVDs) provides an accurate and tractable framework for FAS performance evaluation. When the objective of FAS channel fitting is outage probability (OP) evaluation, accurate characterization of the low-probability left-tail region becomes crucial, while existing fitting strategies that emphasize global fitting accuracy may fail to capture the critical tail behavior required for precise OP evaluation. In this paper, we propose an OP-oriented channel fitting strategy with a left-tail-sensitive target distribution and fitting criterion. Specifically, a combined EVD (CEVD) is introduced as the target distribution, where a generalized Pareto distribution (GPD) is employed to characterize the left tail and a generalized extreme value (GEV) distribution is used to model the global behavior. Furthermore, a modified mean-square-error (MMSE) criterion is developed, which employs logarithmic-domain errors to enhance sensitivity to left-tail discrepancies. Meanwhile, the evaluation points are constructed via uniform discretization on the logarithm of the cumulative distribution function, ensuring uniform sampling across all probability scales. This mitigates the under-representation of tail errors in the overall MMSE, which cannot be effectively addressed by error amplification alone due to the sparsity of tail samples. Simulation results demonstrate that the proposed fitting strategy significantly improves the OP evaluation accuracy in the ultra-low-OP regime.

Unitary synthesis underlies universal quantum computation by translating unitary transformations into circuits of elementary gates. We show that for any budget $m\geq 0$ of ancilla, every $n$-qubit unitary can be exactly implemented using a quantum circuit consisting of $O(4^n)$ CNOT and single-qubit gates in $O(n+4^n/(n+m))$ layers attaining the asymptotically optimal depth–ancilla tradeoff while simultaneously achieving optimal circuit size. The key idea is to encode computational basis states as wavefunctions sharing a Gaussian envelope, converting unitary action on amplitudes into coordinate transformations that admit shallow circuit implementations. The continuous encoding is then discretized, with approximation errors removed by exact residual correction. The depth-ancilla tradeoff follows by recursively applying the quantum Shannon decomposition, synthesizing its smaller unitary factors, and reusing ancillary qubits.

Image-to-text jailbreaks place harmful intent in text, image content, or the relationship between them. We examine image-side factors across four published attack families on a 313-prompt StrongREJECT slice, using five multimodal models and an additional appendix evaluation of InternVL3.5-8B. The harmful instruction is held constant across conditions; the baseline matrix uses one draw per prompt, and paired ablations use three draws with an automated rubric judge. A bare harmful query, with or without a benign unrelated image, produces little attack success on most victims, while attack images substantially increase it. First, per-tile entropy and JPEG size do not reliably distinguish attack tiles from size-matched benign distractors, limiting density-only screening. Second, earlier tile-count ladders were confounded by payload visibility. A corrected region-count test found no detectable effect, so the role of tile-count structure remains unresolved. Third, on Qwen3-VL-8B, the E4 manipulation that removes query-specific relatedness lowers ASR by about 0.12. This supports a bounded attribution to the relatedness manipulation, although image-text congruence remains unmeasured. A within-category control reproduces the direction on overlapping stimuli. Five tests survive the global statistical correction, but only E4 supports attribution to one measured descriptor; the within-category result is a robustness check, not a separate attribution. These conclusions remain conditional on the rubric judge.

Few-shot class-incremental learning (FSCIL) aims to learn novel classes from limited annotations while preserving prior knowledge. Existing methods typically assume a sufficiently large base session, but this assumption fails when both base and incremental data are scarce, leading to weak initial representations, semantic drift, and unstable boundaries. We study this underexplored yet realistic setting, termed Generalizable FSCIL (G-FSCIL), where the base session itself contains only a few classes. Although synthetic data can alleviate supervision scarcity, naively mixing generated samples often introduces semantic noise and exacerbates old-new boundary conflicts. To address this, we propose a framework that curates trustworthy synthetic knowledge for stable G-FSCIL. Specifically, we first construct class-specific synthetic candidate pools using a frozen latent diffusion model, where class inversion is performed at the first observation and the resulting condition embeddings are reused for on-demand generation. Building on these candidates, we learn a knowledge curation strategy that selects samples with both semantic consistency and visual diversity, and distill this process into a transferable selection policy during the base session, which is then reused without further optimization. Leveraging the curated synthetic data, we further design a boundary-stable incremental adaptation scheme, including synthetic-informed prototype initialization and bidirectional boundary calibration to mitigate old-new conflicts. Extensive experiments demonstrate that our method consistently outperforms existing FSCIL baselines, with reduced forgetting and improved balance between old and new classes. Code is available at https://github.com/NiHaoWoJiaoYYC/G-FSCIL.

The identification of plastic inorganic semiconductors becomes challenging when their mechanical responses depend on crystallographic orientation, sample size, and loading conditions. Using layered GeSe as a model system, we examine its deformation behavior through macroscopic compression, bending, conventional micropillar compression, and eccentric micropillar compression. Under macroscopic compression perpendicular to the layers, GeSe sustains approximately 23% strain without fracture, whereas bending along the c-axis armchair direction produces brittle cleavage fracture. Conventional micropillar compression results in brittle fragmentation, while eccentric loading introduces a shear component that activates pronounced interlayer sliding and accommodates deformation. These contrasting responses reflect the coupled effects of bonding topology, interlayer van der Waals interactions, defect density, stress constraints, and strain path on the competition between sliding and fracture. The results highlight the limitations of identifying plasticity from a single direction, scale, or test and support a systematic evaluation framework combining multiple directions, length scales, loading modes, and characterization techniques. More rigorous and unified criteria are needed to guide reliable materials selection for flexible electronics and devices integrated on curved surfaces.

Learning effective representations helps characterize the structure and dynamics of equity markets from financial data with a low signal-to-noise ratio. Black-box deep models can capture complex patterns but may overfit sample noise and lack explicit economic structure. Meanwhile, classic linear financial models provide interpretable references, but their oversimplified assumptions leave non-linear signals uncaptured. To combine the strengths of these two directions, we propose Stock-JEPA, a joint-embedding predictive framework that learns predictable incremental revisions relative to a point-in-time financial prior. First, we leverage a low-complexity financial model to produce fixed statistics summarizing multi-horizon return and risk. A prior projector then maps these statistics into the target encoder's latent space as an anchor. Second, we design a context-conditioned revision predictor to estimate the future representation's predictable displacement from the anchor. Separate losses update the two branches: the anchor learns from prior statistics, while the revision captures additional predictable information from historical context. Third, we freeze all representation modules and train a downstream readout, evaluating its forecasts through cross-sectional ranking and portfolio performance. Theoretically, we prove that optimal revision reduces the prior anchor's expected squared error for the same future representation by exactly $\mathbb{E}[\|\boldsymbol{\Delta}\|_2^2]$. This non-negative gain is the expected squared magnitude of the additional signal predictable from historical context. Experimentally, Stock-JEPA outperforms 13 strong baselines across large-scale China and U.S. equity universes on 5 key evaluation metrics. Ablation studies and representation analysis further demonstrate the value of the learned revisions for representation learning in equity markets.