Papers

Resolving the formation and dissociation mechanisms of hydrate requires identifying not only hydrate-like water, but also the specific cage topologies by which disordered liquid water reorganizes into, or escapes from, crystalline hydrate networks. Here we introduce HyTopo, a cage-topology-centered analysis framework that combines a hydrate order parameter, hierarchical ring and cage recognition, and cage-transition network analysis. Based on local orientational correlations, HyTopo classifies water molecules into liquid-like, precursor-like, hydrate-like, and ice-like environments, and uses this information to guide efficient topological searches for crystalline hydrate cages and defective intermediates. Applications to methane hydrate growth, homogeneous nucleation, and depressurization-induced dissociation reveal that hydrate evolution is governed by heterogeneous and pathway-dependent cage dynamics. During hydrate growth, EA-WPU suppresses the overall proliferation of cage-like structures while altering the balance between defective cages and crystalline sI cages. During nucleation, incomplete cage motifs dominate the early pathway and provide a structural reservoir from which a small fraction of stable 512 and 51262 cages. Hydrate dissociation proceeds through interfacial collapse, bulk decomposition, and dissolved-gas exsolution, with cage lifetimes and transition pathways strongly dependent on cage topology. These results show that HyTopo provides a physically interpretable and topology-resolved route to connect local water ordering, cage formation, structural defects, and hydrate phase transitions in complex molecular simulations.

Streaming video assistance requires models to answer asynchronous questions from an observed prefix under a fixed context budget. Existing approaches model response timing or compress history, but an online state formed before future questions are known can omit visual details before later questions reveal their relevance; the retained state alone cannot recover them. We introduce Watch-Think-Interact (WTI), a closed-loop framework for multi-question streaming video reasoning. WTI maintains compact natural-language memory entries tagged with source-video time ranges; these entries support direct reasoning when sufficient and otherwise anchor selective recall of finer visual evidence. For each question, WTI answers when current context and memory suffice, continues watching when required evidence has not appeared, or recalls a relevant past interval and decides again after incorporating the returned chunks, without replaying the full observed history. To train this behavior, we construct WTI-82K, comprising 82,335 timed questions across 4,812 causally aligned trajectories, and develop Stream-GDPO to optimize complete multi-question streaming rollouts using trajectory-level feedback for response timing, source-video recall, and memory updates. WTI achieves state-of-the-art aggregate performance among the compared open-source streaming baselines, reaching 83.3% on StreamingBench and 73.6% weighted overall accuracy on OVO-Bench.

We theoretically investigate quantum correlations and energy storage in a single-electron silicon double quantum dot (eDQD) coupled to a single-mode microwave cavity. The charge and spin degrees of freedom are hybridized by Rashba spin-orbit coupling (SOC), while the cavity interacts with the eDQD through spin-photon and charge-photon couplings. Using the reduced thermal density matrix, we characterize spin-charge entanglement by concurrence and quantum coherence by the $l_1$ norm. At low temperature, the cavity strongly modifies the spin-charge correlations, producing distinct regimes of enhanced and suppressed entanglement. We identify a nonlinear crossover between an eDQD-dominated regime and a photon-dressed regime. This crossover is revealed independently by two signatures: a rapid change in the concurrence and the onset of finite photon occupation in the cavity. Its boundary exhibits a dominant $G_{\rm c}\propto\sqrt{\Omega}$ dependence and is robust against cavity Hilbert-space truncation. Beyond the Rashba coupling studied in [Ferreira et al., Phys. Rev. A 107, 052408 (2023)], we show that the cavity provides an additional tunable means of controlling quantum correlations and energy storage. We further characterize the eDQD as a quantum battery under coherent charging and evaluate its stored energy and ergotropy. The ergotropy varies markedly across a cavity-dressing crossover resembling that identified from entanglement, connecting cavity-induced modification of the dressed states to extractable work. Optimization over the charge-photon coupling reveals parameter-dependent optimal charging regimes and enhanced maximum ergotropy with increasing Rashba coupling. Our results establish cavity coupling and SOC as complementary controls of quantum correlations and extractable energy in semiconductor eDQD-cavity systems.

Federated parameter-efficient fine-tuning enables clients to adapt pre-trained models without sharing raw data or communicating the full model, but statistical heterogeneity makes a single global adapter insufficient for personalized prediction. Existing personalized methods typically use the same low-rank structure for both shared and private adaptation, overlooking their distinct requirements for aggregation and personalization. We propose FedLAFP, a role-aware framework that couples a compact, globally aggregated LoRA branch with a client-private, full-rank-capable RandLoRA branch. The shared branch provides an efficient interface for transferring common knowledge, whereas the private branch combines fixed random low-rank bases with learned scaling coefficients to provide expressive client-specific adaptation without additional communication. Client- and layer-specific mixing coefficients jointly fuse the two branches, and only the shared LoRA parameters are exchanged. A controlled linear study supports this role assignment: LoRA yields more aligned client updates and lower aggregation error, while RandLoRA more accurately recovers client-specific residuals. Experiments across four visual recognition benchmarks show that FedLAFP consistently outperforms local-only and federated LoRA baselines, achieving an average personalized accuracy of $86.93\%$ and exceeding the best baseline average by $1.30$ percentage points.

We study finite-momentum trimers and Cooper quartets in a one-dimensional two-component Fermi gas with coexistent $s$- and $p$-wave interactions, regularized in the relative momentum of each pair so that they remain Galilean invariant. Pauli blocking alone then makes the trimer dispersion deviate from free center-of-mass motion, so the trimer energy can be lowest at finite total momentum. Allowing this motion expands the $abb$ region among pairs and trimers and removes the region where an odd-wave pair is lowest. The $aabb$ quartet lies below all these branches at every sampled coupling, and in a strip starting on the crossing of the two dimer-pair channels its lowest breakup channel is a moving $abb$ trimer and an atom. At weak interspecies attraction, a variational bound shows that two odd-wave Cooper pairs bind into a quartet, whose binding exceeds that of the coexisting trimer by orders of magnitude as the attraction weakens. At stronger interspecies attraction, the quartet binding relative to two unlike-species dimers disappears when the odd-wave interaction is switched off, although the even-wave interaction dominates the interaction energy. Relative to the in-vacuum case, the Fermi sea enhances the weak-coupling quartet binding, and at fixed scattering lengths the quartet binding grows with density while the coexisting trimer becomes much shallower. These results show that neither the most strongly bound pair nor the dominant interaction energy alone determines whether the quartet binds.

Building extraction from optical remote sensing (RS) imagery is fundamental to urban mapping, yet existing methods are often dataset-specific and generalize poorly to unseen domains. Their practical use is also limited by insufficient detail recovery and weak geometric regularization, leading to blurred boundaries, irregular shapes, and merged adjacent buildings. To address these issues, we propose UniBuild, a unified building extraction framework for multi-source RGB optical RS imagery. First, a unified multi-dataset training scheme is constructed over heterogeneous RGB optical datasets to learn transferable building representations across sensors and resolutions. Second, a novel detail-preserving HR-DPT decoder is designed to integrate high-level semantic features with high-resolution spatial features, enhancing building detail recovery. Third, geometry-aware regularization is introduced through a structure-tensor-based direction-aware loss for boundary direction consistency and a saddle-aware loss for suppressing false activations in narrow inter-building gaps under low-resolution conditions. We train and evaluate UniBuild on multi-source RGB optical datasets, including 10 public high-resolution datasets and two self-collected low-resolution datasets. Experiments show that UniBuild consistently improves building-region accuracy, boundary sharpness, and adjacent-building separation across diverse datasets. It also generalizes well to unseen domains and supports practical building extraction from RGB optical RS imagery up to 10\,m resolution. The predicted masks can be further converted into GIS-compatible building footprints through simple polygonization. The trained model and inference code are released at https://github.com/zhu-xlab/UniBuild.

Despite rapid progress in video generation models, they still exhibit obvious motion deficiencies, often manifested as incorrect object motion. However, most existing video quality evaluations focus on aesthetic quality or text-video alignment. To address this gap, we study object-centric motion fidelity assessment, evaluating target objects along object consistency, motion continuity, and physical plausibility. To achieve this, we first introduce VidMotion, a diagnostic dataset of 6,879 videos with designated moving objects and fine-grained annotations including dimension-wise scores and failure causes. We further propose MotionInsight, a diagnostic evaluator that shifts assessment from implicit RGB-frame observation to explicit motion-space diagnosis. By constructing motion-aware representations, MotionInsight makes subtle motion deficiencies more observable. We also introduce motion-specific rewards during GRPO to transform observed motion into a diagnostic assessment. Experiments demonstrate that MotionInsight provides an effective basis for diagnosing object motion deficiencies, producing human-aligned scores along three dimensions and grounded explanations.

Speculative decoding accelerates autoregressive inference by verifying multiple draft tokens in a single target forward pass. However, as the context grows, existing state-of-the-art drafters become increasingly expensive, eroding the very efficiency advantage they are designed to provide. We argue that this scaling is unnecessary. A standalone language model must grow with its prefix because it is solely responsible for every token it produces. A drafter, by contrast, only proposes candidates; the target catches and corrects every error before any token is committed. The drafter's decoding cost can therefore be made entirely independent of the prefix length. We introduce LongSpark, a block-diffusion drafter that achieves this by extracting fixed-size, multiscale views from the target's verification pass, thereby eliminating the need for a growing persistent state. Extensive evaluations demonstrate that LongSpark achieves state-of-the-art end-to-end efficiency across multiple model scales and realistic serving conditions. Notably, it delivers the lowest time-per-output-token on long-context tasks while reducing the drafter's context state by several orders of magnitude.

Zero-shot text-to-speech synthesizes new utterances in a speaker's voice from a short reference recording. Voice cloning requires accurate content and preserved speaker identity, but supervised acoustic-token prediction does not directly optimize these waveform-level properties. Reward-based post-training addresses this mismatch, but in discrete diffusion, token choices and reveal positions jointly define the sampling trajectory, complicating alignment. We introduce RAWD-TTS (Ratio-free Advantage-Weighted Denoising), which scores decoded samples with recognition and speaker rewards and uses group-relative advantages to weight masked-token reconstruction of those samples, without reverse-trajectory likelihoods or target audio. On 500 Russian CV3-Eval voice-cloning prompts, joint alignment reduces word error rate from 3.18% to 2.42% at the reward-selected checkpoint (24.0% relative) and to 2.58% at the final checkpoint (19.0%), while WavLM speaker cosine rises from 0.733 to 0.748 and 0.755. Controlled experiments characterize recognition-identity trade-offs and the effects of corruption count, group composition, and weighting.

Low-Rank Adaptation (LoRA) is a widely used approach to parameter-efficient fine-tuning (PEFT), yet a performance gap can remain relative to full fine-tuning (FFT). Many LoRA variants improve the initialization or optimization of low-rank factors. At each training step, however, their first-order weight-space directions are constrained by the current parameterization. We characterize the corresponding LoRA-accessible gradient space and show that it coincides with the tangent space induced by the current LoRA parameterization. This characterization yields an orthogonal decomposition of the full weight gradient at the current model parameters. We term the component orthogonal to this space the normal gradient. Based on this decomposition, we propose GDLoRA (Gradient-Decomposed Low-Rank Adaptation). GDLoRA reconstructs the full weight gradient from forward activations and backward signals, extracts its normal component, and directly updates the base weights with this component, while retaining standard AdamW optimization for the LoRA factors. GDLoRA incorporates complementary normal gradients without increasing standard LoRA's optimizer-state memory budget under matched adapter and optimizer configurations. Experiments on natural language understanding, mathematical reasoning, commonsense reasoning, and image classification show that GDLoRA consistently improves over LoRA and narrows the performance gap to FFT. The code is available at https://anonymous.4open.science/r/GDLoRA.

We investigate the Morse index and spectral asymptotics of least-energy solutions to the planar Choquard equation \begin{equation*} \begin{cases} -\Delta u = \displaystyle\left(\int_{\Omega} \frac{u^{p+1}(y)}{|x - y|^\alpha}dy\right) u^p & \text{in } \Omega, \\ u > 0 & \text{in } \Omega, \\ u = 0 & \text{on } \partial \Omega, \end{cases} \end{equation*} where $\Omega\subset\mathbb R^2$ is a smooth bounded domain and $0<\alpha<1$. Under some geometric assumptions, we obtain sharp asymptotics for the first four eigenpairs of the linearized problem as $p\to+\infty$. The first eigenvalue is exactly $(2p+1)^{-1}$. The second and third eigenfunctions give the translation modes of the limiting bubble, and their eigenvalue expansions are determined by the Hessian of the Robin function at the concentration point. The fourth eigenfunction gives the dilation mode with eigenvalue $1+3(4-\alpha)/(2p)+o(p^{-1})$. As a consequence, we derive a comparison between the spectral indices of the solution and the critical-point indices of the Robin function.The proofs combine variational estimates, blow-up analysis and nonlocal Pohožaev identities.

As large language models (LLMs) advance, AI agents are increasingly deployed in open-world environments to tackle complex sequential tasks (e.g., document processing, cross-application collaboration), relying heavily on actions ranging from GUI operations to semantic APIs. However, three core challenges persist: the "scale dilemma" of massive tool ecosystems exceeding LLM context windows, the "non-stationarity" of tool quality due to updates or outages, and the "heterogeneity" of feedback formats (pixels, text, structured data) creating information silos. To address these, we propose AnyAct, a universal action layer that unifies available capabilities into a self-evolving action space, enabling agents to operate efficiently and reliably in large-scale, dynamic tool ecosystems. AnyAct's core design focuses on two objectives: constructing this action space via hierarchical progressive retrieval (filtering task-relevant actions) and test-time reliability evolution (pruning unreliable actions), and enabling reliability-aware action orchestration through a heterogeneous observation grounding module that unifies multi-modal feedback. Additionally, it defines a hybrid action space (primitive + semantic actions) and optimizes for a balance between task success rate and execution cost. Evaluations on LiveMCPBench and OSMCP (a new benchmark we developed for multi-granularity action collaboration) demonstrate state-of-the-art performance. AnyAct delivers substantial performance gains over baseline methods across various LLM base models on LiveMCPBench and improvements are particularly notable for models with constrained native capabilities. On OSMCP, it achieves 77.27% overall success with only 50 steps, which is half the steps required by most competitors.

Integrating transcriptomic and electrophysiological data is essential for building multimodal foundation models for neuroscience. Patch-seq provides paired measurements of gene expression and intrinsic electrophysiology from the same neuron, establishing a basis for training cross-modal models. Here we introduce LangPatch, a foundation-model-based contrastive learning framework that uses paired Patch-seq data to align pretrained GenePT representations with electrophysiological phenotypes through a language-based interface. Gene descriptions and verbalized electrophysiological profiles are embedded by the same frozen text encoder. A context adapter and projection modules connect the modalities through paired contrastive learning. Across mouse visual, mouse motor, and human cortical cohorts, LangPatch achieves the highest mean transcriptome-to-electrophysiology prediction correlation among the evaluated foundation-model and representation-learning methods. It also improves held-out cross-modal alignment in the two mouse cohorts (FOSCTTM 0.107/0.135 vs. 0.208/0.222 for JAMIE, an existing cross-modal Patch-seq imputation method). It predicts transcriptomic family, type, cortical layer, and marker-gene expression from electrophysiology, exceeding other baselines on most endpoints. More importantly, the method transfers across brain areas and species: a model trained on mouse visual cortex predicts electrophysiology in motor cortex with approximately 70% correlation retention and in human cortex with 47% (58% on acute-slice recordings). Together, these results demonstrate alignment between molecular and functional representations of neurons, providing a building block for multimodal foundation models in neuroscience.

We establish an Arakelov inequality for variations of Hodge structures underlying families of principally polarized complex Abelian varieties. In the equality case, this leads to a numerical characterization of certain totally geodesic ball quotients inside the moduli space of Abelian varieties. Our result extends work of Möller-Viehweg-Zuo by removing the strong positivity conditions imposed in their statement. For families over compact base spaces, our proof involves showing that the period map associated with a family of Abelian varieties factors through certain MMP operations and then generalizing the results of Möller, Viehweg, and Zuo to an appropriate singular setting. In the quasiprojective surface case, we implement a new approach that does not pass through Miyaoka-Yau-type uniformisation theorems but uses an argument going back to an idea of Mok, symmetric space theory and semistability considerations instead.

Efficient patient flow coordination across autonomous hospital departments is critical for mitigating overcrowding and balancing resource utilization. While classical queueing theory, specifically open Baskett–Chandy–Muntz–Palacios (BCMP) networks, provides an interpretable mathematical topology for healthcare operations, analytical models rely on stationary assumptions and fixed routing matrices that degrade under state-dependent real-world dynamics. Conversely, centralized reinforcement learning approaches struggle to accommodate the decentralized structure of hospital governance, where individual clinical departments function with localized observations, heterogeneous resources, and divergent operational objectives. In this paper, we present a Multi-Agent Systems (MAS) framework titled Physics-Informed Multi-Agent Coordination, which embeds empirically calibrated BCMP queueing topologies as physical priors within a decentralized multi-agent reinforcement learning architecture. Formulated as a Decentralized Partially Observable Markov Decision Process (Dec-POMDP) under coupled resource constraints, our method enables autonomous departmental agents to cooperatively negotiate patient routing and dynamic service scaling. To mitigate environmental non-stationarity without inducing excessive communication overhead, agents exchange localized action fingerprints along network edges and optimize a spatially decomposed reward structure. Empirical evaluations driven by real-world MIMIC-IV patient trajectories indicate that this cooperative multi-agent approach substantially reduces cumulative system delay compared to static Markovian approximations, heuristic dispatching, and independent multi-agent baselines, while maintaining clinical safety constraints.

This paper explores the applicability of functional programming to the design of Application-specific Integrated Circuits (ASICs). We investigate the impact of designing ASICs using high-level, abstract Hardware Description Language (HDL) features versus employing low-level optimizations on the area of the synthesized circuits. The aim is to determine whether using low-level optimizations is beneficial and, if so, whether it is worth the added implementation effort. To carry out the investigation, we implement an unsigned bit-serial multiply-accumulate (MAC) unit in 16 different configurations using the functional HDL Clash. We make use of both low-level bit-manipulation techniques as well as Clash's high-level constructs for circuit design. The experimental evaluation shows that some high-level constructs of Clash have negligible influence on the resulting circuit size, suggesting that using the full power of functional programming is a viable approach to hardware design. To evaluate the impact of the used HDL itself, we also implemented versions of the MAC in Verilog. The experiments clearly show that there seems to be an inherent overhead in using Clash compared to Verilog code written by a seasoned engineer.

Motivated by the pioneer work of Viswanath \cite{VIS2000} on random Fibonacci sequences and the development of random Euclidean addition chains in cryptography, we introduce a new family of random Binomial-Fibonacci sequences. We study the asymptotic behavior of their finite-time Lyapunov exponents. In the special case of random Bernoulli-Fibonacci sequences, we show that they convergence almost surely to an explicit Lyapunov exponent and prove a central limit theorem with an explicit formula for the variance. We also show that they satisfy a large deviation principle.

We study the backward-Euler scheme for a class of stochastic functional differential equations (SFDEs) with memory driven by a fractional Brownian motion with Hurst parameter H > 1/2. We first establish the local consistency of the method, with a local truncation error of order H -$\rho$- $\beta$ + 1, and then, as the main result, prove the uniform global pathwise convergence of the numerical approximation, on a set of probability one, with order H -$\rho$ -$\beta$, for $\beta$ $\in$ (1-H, 1/2) and 0 < $\rho$ < H -$\beta$. A uniform high-probability formulation with deterministic constants follows as a corollary. In particular, the convergence order can be taken arbitrarily close to 2H -1, matching the order obtained for the explicit Euler scheme. The convergence proof relies on an exact integral representation of the scheme at mesh points, an a priori bound for the numerical solution, fractional calculus estimates, and a Gronwall inequality for weakly singular kernels.

Let $\operatorname{UT}_n$ denote the group of $n\times n$ real upper triangular matrices with positive diagonal entries. We derive a closed-form analytic expression for the Karcher mean on $\operatorname{UT}_2$ and obtain a sharp convergence threshold for the projection mean iteration with fixed initial diagonal, together with a sufficient convergence condition for arbitrary initial points. In addition, we obtain explicit formulas for the log-Euclidean mean and establish a Lie–Trotter type formula for the Karcher mean on $\operatorname{UT}_2$. For $\operatorname{UT}_n$ we derive the diagonal and first superdiagonal entries of every solution of the Karcher equation and transfer nonconvergence examples from $\operatorname{UT}_2$.

LLM-based multi-agent systems (MAS) increasingly use latent collaboration to avoid the information loss and repeated encoding-decoding overhead of natural-language communication. However, directly forwarding all sender latents makes the receiver-side context scale with both the number of agents and the reasoning length, increasing computation, memory usage, and collaboration latency. A natural solution is latent compression. But we find that cross-agent redundancy remains unresolved in existing latent compression approaches, which typically compress each sender independently and then concatenate the results. We propose LatCom, a cross-agent latent compression framework for efficient multi-agent latent collaboration. LatCom maps multiple sender latents into a fixed number of receiver-readable and task-relevant slots. Rather than reconstructing all sender hidden states, it optimizes the compressed latents for receiver-side task utility. LatCom trains the compressor in two stages: single-sender readability learning first establishes a latent interface interpretable by the frozen receiver, and multi-sender fusion learning then trains the compressor to fuse complementary evidence and remove redundancy across agents. Experiments on multiple benchmarks with Qwen3-4B show that LatCom achieves an average 2.46x inference speed-up over LatentMAS and reduces output token usage by 70.3% while maintaining comparable average accuracy.

Structured images, such as diagrams, charts, and flowcharts, are inherently symbolic and can be compactly represented in an editable format, yet in practice, they are often rendered as images, and therefore not graphically editable. This mismatch presents a significant challenge for researchers, engineers, and designers who wish to incorporate modified versions of existing graphic content into new materials without manually reconstructing it. In this study, we presentBack2Struct, which "makes structured images editable again" by directly recovering vector graphics code (SVG / XML) from image representations. Given an image of a structured graphic, Back2Struct predicts semantically object-level SVG / XML code that explicitly encodes text, shapes, topology, and layout, rather than performing low-level pixel vectorization. The generated code can be seamlessly imported into tools such as PowerPoint, allowing users to edit, refine, restyle, and reuse graphic content while preserving structural fidelity. Beyond supervised fine-tuning on ground-truth SVG token sequences, we further optimize Back2Struct with reward-based learning to better match deployment-time requirements: the output should be syntactically valid, properly concise, and visually faithful to the input diagram. Specifically, we design a composite reward that jointly encourages SVG / XML compilability, length consistency with the reference code, and structural or semantic similarity between the generated and ground-truth graphics. These complementary signals guide the model to produce SVGs that are not only closer to the training distribution, but also more complete, editable, and renderable in practice. Experiments show that Back2Struct improves accuracy, editability, validity, and user alignment over baselines. Dataset and code are available at: pengyu965.github.io/Back2Struct.github.io

We propose RBF-GNN, a new pseudo-coordinate based graph neural network architecture that takes into account Euclidean, spherical or angular coordinates and uses them to induce a powerful spatial inductive bias. Similar in architecture to SplineCNN, we improve upon the latter by replacing the less efficient sparse-activation based B-splines whose number grows exponentially with dimension by rational Padé basis functions. For effective training we propose a spline-subspace initialization and a variance-preserving weight rescaling. Experimentally, we evaluate on a number of popular neural network architectures that use SplineCNNs. We replace only the SplineCNNs with RBF-GNN. We achieve improved results, including on semantic keypoint matching, shape matching, event based camera computer vision tasks. Code is available at https://github.com/pawelswoboda/RationalBasisCNN.

Automatic speaker verification must remain reliable across devices, rooms, and compression pipelines. We present ReDimNet2+, which scales training of the compact ReDimNet2 backbone across seven public corpora (63,934 speakers, about 8,675 hours). Analysis of a VoxBlink2 subset reveals a shift in predicted spectral coloration, motivating codec and waveform augmentation alongside this multi-corpus training, large-margin fine-tuning (LMFT), and graph-based retrieval reranking. With random 4-second evaluation windows for all models, ReDimNet2+ LMFT reduces pooled VoxCeleb1 EER from 2.42% to 0.82% and a 26-condition robustness stress-test EER from 7.21% to 1.99%. Under this shared local protocol, it reaches 0.35% EER on VoxCeleb1-O versus 0.787% for the best evaluated WeSpeaker checkpoint. On a VoxBlink2 retrieval subset, reranking improves the final model's Pr@k from 0.7413 to 0.7687.

Rapid assessment of building damage after natural disasters is essential to support emergency response. Earth Observation satellites can acquire relevant imagery shortly after an event, but exploitation is limited by uplink and downlink capacity and by ground-processing latency. We address this with a bi-temporal building damage assessment pipeline built on a siamese detector derived from YOLOX, designed to compress information at both ends of the ground/space link. On the ground, pre-disaster reference images are encoded into a compact latent space – compressed by up to a factor of 64 – and uplinked to the satellite. On board, this reference is compared with a fresh post-disaster acquisition so that the downlink carries only actionable object-level products, bounding boxes and damage classes, instead of full scenes. This cuts the data exchanged in both directions, while on xBD the strongly compressed reference still preserves most of the detection performance. Because on-board acquisitions suffer from residual pre/post co-registration errors, we introduce a latent-space shift estimation and correction module that regresses the global offset from the coarse feature level and realigns the post-disaster features before fusion. It substantially improves robustness to de-registration – especially under large shifts, where fusion-only variants collapse – while also raising nominal accuracy and remaining compatible with the strongest compression. We finally port the pipeline to two embedded targets, a Xilinx Versal VCK190 and an NVIDIA Jetson AGX Orin, and report hardware performance (latency, throughput, power efficiency). The core detector and its compression port cleanly to both, but the operators needed for long-range robustness survive only on the Jetson GPU, whereas the Versal DPU does not.

For a long-horizon agent, context is the bottleneck: the history is resent with every request, the window caps task length, and reasoning degrades as the history grows. Replacing structured objects with compact retrieval Cards shortens the prompt and keeps the exact originals retrievable, but editing the history can break prefix-cache reuse, and prior recoverable methods time their edits by forecasts of future reuse or by preset intervals. We propose CADOC (Cache-Aware Dynamic Object Context), an online algorithm that replaces structured objects with compact Cards while preserving exact, on-demand retrieval of their original contents. CADOC schedules replacements in batches by balancing accumulated waiting cost against shared cache-reconstruction cost. Its scheduling rule follows from an economic order quantity trade-off, recovers the optimal integer batch under stationary assumptions. Across evaluation, CADOC consistently achieves the lowest aggregate input cost among the compared configurations, which reduces input cost by approximately 40% on average while maintaining task performance close to full context. CADOC thus provides a cost-derived approach to compressible context management, demonstrating that efficient compression depends not only on shortening prompts but also on scheduling edits to preserve cache reuse.

Federated learning enables multiple clients to collaboratively train models without sharing their private data. However, the lack of visibility into local training makes it difficult to verify whether clients follow the prescribed training procedure or submit malicious updates, such as model poisoning. A natural approach is to replay client training for verification. However, privacy-preserving replay produces numerical results that cannot be directly matched with local client execution because the two run in different environments. We present OPFL, an optimistic verification framework for privacy-preserving federated learning. To protect data privacy, OPFL performs replay inside secure multi-party computation (MPC). Although gradients computed on MPC and local GPUs are not bitwise identical, we observe that their absolute differences are stable and bounded. OPFL therefore calibrates an empirical boundary offline and uses it to distinguish benign numerical deviations from malicious manipulation. To reduce the cost of expensive MPC replay, OPFL adopts optimistic verification by post auditing only sampled training steps. Experiments on LeNet, BERT, and Qwen show that the boundary generalizes across datasets, input lengths, and GPUs, while achieving $0$% ASR against model poisoning and PGD-based attacks. On a LeNet workload, at $p=0.01$, OPFL is approximately $98.6\times$ faster than full MPC-based FL and $625.5\times$ faster than ZK-based approach.

We demonstrate that twisted and vector optical beams can spatially separate two-dimensional electrons with opposite spin projections. Spin accumulation arises from diverging spin currents induced by structured light together with spin relaxation and diffusion. We derive analytical expressions for the resulting spin density in the quasi-local and diffusion-controlled regimes and numerically calculate the spin textures for the parameters relevant to transition metal dichalcogenide monolayers. Our results suggest a mechanism by which the twist and vector pattern of optical beams can be imprinted onto the electron spin degree of freedom.

Modeling heterogeneous evacuation behavior under a moving threat is difficult because human perception, memory, and evidence evaluation are not well captured by fixed rules. We propose a novel LLM-powered agent-based framework to represent these internal decision processes. Each pedestrian agent perceives a private symbolic ASCII view, maintains a Memory-based Knowledge Graph derived solely from individual observations, and makes decisions through persona-conditioned prompts under a common sampling configuration. A compressed decision context with stateless memory preserves trial-and-error experience across turns while excluding reasoning traces, and a validation engine separates behavioral choice from physical feasibility by executing routes only over observed terrain. We evaluated eight personality compositions in eight paired randomized blocks within a simulated public plaza. Usable-exit knowledge was strongly associated with evacuation success: 89.5% of agents possessing such knowledge evacuated, compared with 1.05% of those without it. Personality compositions also differed in their evaluation of remembered threat evidence: the proportion of danger assessments varied by 0.265, while high-urgency, low-directness decisions ranged from 11.41% to 34.33%. After direct threat sightings, responses converged, with 99.6% of assessments classifying the situation as dangerous. Movement was selected in 99.8% of decisions. Overall, evacuation outcomes were strongly associated with information access, while evacuation time was jointly associated with spatial geometry, information, and affect. The framework provides an auditable approach to generating endogenous behavioral heterogeneity through persona-conditioned LLM agents in crowd-evacuation simulations.

The orbit selected by a magnetic particle in a viscous liquid under a rotating field is commonly described in the instantaneous Stoner–Wohlfarth limit, in which the magnetic moment always occupies a minimum of the anisotropy energy. In this limit, precession out of the field plane exists only for reduced fields $B_0/B_a<1/\sqrt2$, and stronger fields admit only planar synchronous rotation or libration. We study a macroscopic egg-yolk particle, a printed shell containing two freely rotating permanent magnets in fluid-filled cavities, driven by a compensated rotating field and tracked optically. In addition to synchronous rotation, precession, and libration, the particle exhibits a polar phase: above the instantaneous bound and at high frequency, the shell axis leaves the field plane and locks onto a narrow cone about the rotation axis. We extend the model by a finite viscous friction between the magnets and the shell. The extension introduces a single dimensionless parameter, the ratio $\kappa$ of inner to outer rotational friction, and reduces to the instantaneous theory as $\kappa\to0$. Numerical phase diagrams show that increasing $\kappa$ tilts the bistable band to higher fields at higher frequencies and extends stable precession above $B_0/B_a=1/\sqrt2$, whereas for large $\kappa$ planar libration returns at high frequency. With a calibrated anisotropy field, relaxation rate, and $\kappa$ of order unity, the model reproduces the terminal state of all measured records and their tilt angles, including the polar cone. In the model the polar state is maintained by a steady rotation of the magnets relative to the shell, driven by the field and anisotropy torques and dissipated by the inner friction; it therefore exists only for intermediate $\kappa$.

Autoregressive decoding (AR) with Transformer models is memory bandwidth bound at single stream inference, the typical deployment regime for on device text to speech (TTS). Real time streaming with Qwen3-TTS requires more than 200 sequential model calls per second, dominated by the inner loop MultiCodeDecoder that emits the 15 residual vector quantization (RVQ) codes per 80 ms audio frame. We propose RVQ position aware speculative decoding for the MultiCodeDecoder, attaining 2.47 accepted tokens per model call at $5\times10^{-4}$ percent added parameters and 10 to 20 percent per round speculation/verification overhead, reducing real time synthesis from 200 to 88 sequential model calls per second. The scheme is distributionally lossless under the deployed top-k sampling, and WER parity with the original system is consistent with this guarantee. We deliver 2 to 2.2x speedup for RVQ token generation with Qwen3-TTS 0.6B on recent iPhone and Apple Silicon Mac devices.