arXiv:2606.25116v1 Announce Type: cross Abstract: Respiratory acoustic foundation models (FMs) are benchmarked exclusively on smartphone recordings, yet clinical deployment increasingly targets body-coupled (BC) wearables whose sensors attenuate high-frequency content through tissue and bone, leaving FM reliability uncharacterised. We introduce BCoughBench, evaluating five FMs (OPERA-CT/CE/GT, HeAR, M2D+Resp) on nine classification tasks (AUROC, sensitivity at 95% specificity, Expected Calibration Error) and three age regression tasks (MAE vs. a mean-predictor baseline) across five EBEN-simulated BC sensor conditions on five labeled cough datasets. Mean AUROC declines from 0.785 (smartphone) to 0.689-0.723, degrading most under temple vibration pickup ($\Delta$ = -0.096) and least under the soft in-ear ($\Delta$ = -0.062). No FM meets the clinical sensitivity threshold (Se@Sp95 $\geq$ 0.20) on most disease tasks under any BC sensor. Sex classification on the CIDRZ cohort collapses (AUROC 0.954 to 0.596-0.628, $\Delta$ = -0.341) while COVID detection is nearly unaffected ($\Delta$ = -0.004). Age regression is robust, improving under the forehead accelerometer on CoughVID (MAE 9.61 to 8.97 yr); HeAR leads on regression and demographic tasks, M2D+Resp on disease and characteristic tasks. BCoughBench provides a reproducible framework for FM evaluation under wearable conditions.
Science Journals
arXiv:2606.25409v1 Announce Type: new Abstract: We present CV-rules, an alternative characterization of serializability in which a transaction order constructed by a protocol satisfies two per-read conditions, C-rule (Causality) and V-rule (View Consistency), that constrain the reads-from relation and competing writers. While classical Multi-Version Serialization Graph (MVSG) reasoning characterizes serializability via its acyclicity, our approach requires explicit order construction, enabling direct proofs that build on the protocol's own mechanisms. We prove that CV-rules, serializability, and MVSG acyclicity are all equivalent. Moreover, the C/V separation reveals that serializability is polynomial-time decidable for any fixed bound on the width of the order forced by C-rule. We verify five protocols: Two-Phase Locking, Multi-Version Timestamp Ordering, Serial Safety Net (SSN), Aria, and SnapChain. For SSN and Aria, whose original papers defined only certification conditions, we identify explicit transaction orders arising from their mechanisms; we also prove that Aria's unique-write constraint is unnecessary for serializability. SnapChain, in contrast, is designed directly from CV-rules, enforcing V-rule by construction. All results except the complexity bounds are mechanized in Lean with no additional axioms and no admitted goals.
arXiv:2606.25444v1 Announce Type: cross Abstract: Connecting a pre-trained speech encoder to a Large Language Model (LLM) is the standard architecture for building Speech LLMs. However, a structural misalignment exists between the encoder and the LLM. Unlike encoders based on automatic speech recognition, which often produce representations in separate language-specific spaces, LLMs operate within a unified language-agnostic space. A mechanism is required to align the encoder's language-specific representations with the LLM's shared space. We argue that speech translation provides a principled way to achieve this. Unlike monolingual transcription, translation requires the model to bridge different languages and learn language-agnostic representations. We experimentally evaluate the impact of incorporating translation objectives into speech encoder pre-training. Our results demonstrate that translation-enhanced pre-training improves cross-modal integration and leads to superior performance across downstream Speech LLM tasks.
arXiv:2606.25128v1 Announce Type: cross Abstract: Volume and quality of datasets are crucial for deep learning model training, yet they are often constrained by availability and data acquisition costs. Synthetic data augmentation can extend existing datasets with realistic images, and the quality of these images is generally assessed through fidelity metrics such as FID, KID, IS, LPIPS and SSIM that measure structural or distributional similarity. However, such metrics, including the widely used FID, focus on visual fidelity without reflecting downstream utility, and can diverge from human perception under perturbations that are imperceptible to human observers. In this work, we systematically evaluate Earth observation datasets alongside synthetic counterparts generated by deep generative models, comparing automatic metrics against human perception and downstream tasks. Our results reveal a stark misalignment: semantics-preserving perturbations such as rotation drastically alter metric scores while leaving human recognition unaffected, and synthetic samples that score poorly on automatic metrics achieve comparable or higher perceived realism, and can improve downstream performance when combined with real data. By benchmarking semantic segmentation models trained on mixed real-synthetic datasets, we demonstrate that quality metrics rooted in ImageNet-pretrained feature spaces are unreliable indicators for geospatial data. Our findings underscore that automatic quality evaluation of synthetic datasets should be grounded in downstream task performance and human evaluation.
arXiv:2606.25170v1 Announce Type: cross Abstract: We study PAC learning in tabular discounted Markov decision processes with exogenous i.i.d. contexts, with discount factor $\gamma$, finite state space $\mathcal X$, action space $\mathcal A$, and context space $\mathcal Z$. At each time step, a context is drawn independently from an unknown distribution $\mu$ and revealed before the agent acts. This context may affect both rewards and transitions, while remaining uncontrolled by the agent. Depending on the regime, the learner has access either to a sampling oracle for $\mu$, to a sampling oracle for the transition kernel conditioned on state-context-action tuples, or to both. Oracles can be accessed before and during policy execution. The sample complexity is measured by a couple $(n,m)$, where $n$ is the number of calls to the sampling oracles before execution and $m$ is the number of calls to the sampling oracles during execution. When rewards and transitions are known and only the context distribution $\mu$ is sampled, we give a variance-reduced algorithm that solves policy evaluation (PE), best-value estimation (BVE), and best-policy extraction (BPE) with $\left(\widetilde O\left(1/((1-\gamma)^3\varepsilon^2)\right), 0 \right) $ sample complexity. The rate is independent of $|\mathcal Z|$ and minimax optimal up to logarithmic factors. As a corollary, we also obtain tight rates in the case of one-step perfect look-ahead, improving upon the existing guarantees. In the fully unknown regime, where both $\mu$ and P must be learned, we show that PE remains $|\mathcal Z|$-free, with matching upper and lower bounds $\bigl(\widetilde O(|\mathcal X|/((1-\gamma)^3\varepsilon^2)),\, \widetilde O(1/((1-\gamma)^2\varepsilon^2))\bigr)$.
arXiv:2606.26059v1 Announce Type: new Abstract: The laser welding full-penetration is of critical importance, as it constitutes one of the fundamental factors in achieving defect-free welded joints. Accurate prediction of the penetration state is therefore essential for ensuring weld quality. To this end, this paper introduces SimPhysNet, a novel algorithm that achieves high classification accuracy in laser welding penetration prediction using only a limited number of labelled images. This approach effectively overcomes the limitations of supervised learning classification algorithms, which are hindered in industrial applications by their dependence on extensive, high-quality labelled data. The core of SimPhysNet is a unique self-supervised learning paradigm that embeds physical priors into a contrastive learning framework. By incorporating a physics-informed neural network (PINN), the model is guided to extract physically meaningful features of the molten pool and keyhole from a large set of unlabelled data, while three image augmentation tasks further enhance its generalization capabilities. Subsequently, a few-shot learning strategy, based on prototypical networks, enables robust classification by constructing class representations from a minimal set of labelled images. Experimental results demonstrate that SimPhysNet achieves a classification accuracy of 96.06% using only 200 labelled images (approximately 5% of the total labelled dataset), which is comparable to the performance of conventional supervised learning algorithms that utilize the entire labelled dataset. This work presents a new, efficient, and highly accurate method, providing the way for the intelligent automation of laser welding.
arXiv:2606.26068v1 Announce Type: new Abstract: We identify the scale ratio between the Helmholtz transverse-wavenumber scale and the paraxial Gaussian-beam-width scale as the physical parameter connecting separable Gaussian beams with Helmholtz waves. Our optical scale ratio factors the Helmholtz-to-paraxial angular-spectrum restriction with the paraxial-to-Helmholtz large-Rayleigh-range limit. It turns the algebraic In\"on\"u--Wigner contraction and the Cayley--Klein deformation into optical limits governed by scale. Our spectral contraction connects Hermite, Laguerre, Ince, and Boyer--Wolf Gaussian beam families with plane, Bessel, Mathieu, and Weber wave families, respectively, at the coordinate, differential wave equation, algebraic deformation, separation operator, separated differential equation, and separation-spectrum levels.
arXiv:2606.00463v4 Announce Type: replace-cross Abstract: This paper examines how future technosignature searches may constrain competing resolutions of the Fermi Paradox, with particular attention to the possibility that technologically capable entities (TCEs) are either intrinsically rare or deliberately concealed. I propose a multi-pronged observational strategy comprising expanded radio and optical SETI, spectroscopic searches for biosignatures and technosignatures in exoplanet atmospheres, astronomical searches for large-scale extraterrestrial engineering, and Solar System searches for extraterrestrial artefacts (SETA). The latter is identified as having a distinctive temporal advantage because it can probe evidence accumulated over Solar System (and perhaps even Galactic) history, rather than requiring temporal overlap with TCEs. In this context, I argue that searches for micron-scale interstellar technological debris (Arkhipov particles) in lunar and planetary regoliths may provide an additional constraint on past Galactic technological activity. Additionally, I briefly argue that this programme of scientific exploration should be considered alongside its geopolitical and governance implications.
arXiv:2606.17397v3 Announce Type: replace-cross Abstract: How should recommender systems be designed when recommendations shape access to scarce, short-lived opportunities? We study this question in a production setting: Timee, Japan's largest platform for spot work, where workers favorite job templates and receive notifications when firms post shifts from those templates. Maximizing predicted favoriting can generate misdirected concentration: recommendations accumulate on popular templates that create few viable job openings, while templates with unmet labor demand receive too little exposure. We design exposure-control mechanisms for favorite-list management, reallocating template exposure based on posting activity and unfilled capacity. The proposed recommender, thresholded eligibility control (TEC), is fully parallelizable and suitable for large-scale digital platforms. In simulations calibrated to Timee data, TEC raises the per-round job-finding rate from 57.6% to 70.0%. A prefecture-level randomized field experiment increases realized matches and exposure per active template, reduces the share of low-exposure templates, and improves impression-level favoriting and downstream matching.
arXiv:2606.25460v1 Announce Type: cross Abstract: Recent advances in sequence modeling have significantly improved ASR systems, bringing them close to human-level recognition accuracy and enhancing robustness across diverse acoustic conditions and languages. In contrast, Forced Alignment has not experienced comparable progress, and traditional HMM-GMM frameworks remain widely adopted and highly competitive. To address this gap, we propose an end-to-end, fully differentiable neural architecture specifically designed for phoneme alignment. The model consists of an encoder that processes the input signal and a decoder that produces alignment decisions. The encoder is structured into two complementary branches: one dedicated to phoneme identity verification and the other to phoneme boundary detection. The decoder is implemented as a trainable module based on differentiable soft dynamic programming. The entire system is optimized end-to-end using a novel contrastive loss that encourages clear separation between steady-state phoneme regions and transition boundaries. The proposed approach outperforms the current state of the art in phoneme alignment on hand-annotated English benchmarks, achieves strong word-level generalization results, and demonstrates generalization on unseen languages.
arXiv:2606.25281v1 Announce Type: new Abstract: The growth of artificial-intelligence workloads is increasing the electrical and thermal demands on data-center power-delivery systems, making conversion efficiency, power density, and reliability critical design priorities. This review examines how gallium-nitride (GaN) power devices can be matched to specific stages of the grid-to-load conversion chain, including power-factor correction, isolated DC/DC conversion, 48-V intermediate-bus conversion, and point-of-load regulation. Si, SiC, and GaN are compared using converter-relevant metrics, and lateral, vertical, and specialized GaN architectures are evaluated in terms of voltage scalability, switching behavior, reverse conduction, thermal pathways, gate control, and technology maturity. The analysis shows that GaN provides a stage-dependent rather than universal advantage. Commercial lateral GaN HEMTs are particularly effective in high-frequency, low-to-mid-voltage stages, while specialized and hybrid devices support bidirectional operation, normally-off control, extreme conversion ratios, and integration. Vertical GaN remains an emerging option for higher-voltage and higher-power conversion. A quantitative framework links cascaded converter efficiency to electrical-loss reduction, cooling demand, annual facility energy use, and operational carbon emissions. Broad deployment further requires low-parasitic packaging, disciplined gate-drive and EMI co-design, mission-profile reliability qualification, scalable manufacturing, and supply-chain resilience. GaN is therefore best treated as a stage-specific system lever whose value depends on coordinated device, topology, package, and thermal co-design.
arXiv:2606.24607v1 Announce Type: cross Abstract: Stellar toroidal magnetic fields are known to be unstable to the Tayler instability. Here we demonstrate the existence of a complementary current-driven instability of essentially arbitrary toroidal-field configurations in stably stratified nonrotating stars with the following properties: (i) in ideal magneto-hydrodynamics, it grows on the Alfv\'{e}n timescale $\tau_{\rm A}$; (ii) under certain conditions, it may reveal itself by driving shellular differential rotation about an arbitrary axis perpendicular to the magnetic-field symmetry axis; (iii) it is large-scale in the angular directions $\theta$ and $\varphi$, and develops at radial wave-numbers $k \lesssim \mathcal{N}\tau_{\rm A}/R$, where $\mathcal{N}$ is the Brunt-V\"ais\"al\"a frequency and $R$ is the stellar radius. Thus, unlike the Tayler instability, the proposed instability is intrinsically global. Consequently, it may be less susceptible to dissipative suppression than the Tayler instability and can prevail over it in some regimes. This instability may have broad implications for magnetic field generation in stars and could modify scenarios of magnetic field amplification within the Tayler-Spruit dynamo, contributing to models of efficient angular-momentum transport and chemical mixing in stellar interiors.
arXiv:2606.25784v1 Announce Type: new Abstract: Data augmentation is known to improve generalization of deep visual models. Recent methods favor mixup strategies that generate interpolated samples to improve model performance. However, these techniques not only incur significant computational overhead, they also lead to semantic disruption of augmentation data due to cross-sample mixing. We first propose Self-Saliency ($S^2$) Mixup, which constructs challenging yet label-consistent samples by extracting multi-scale salient patches and reinserting them into non-salient regions of the same image. This promotes scale-invariant feature learning while avoiding cross-sample interference. To further enhance model robustness, we introduce FracMix, a mixing scheme that injects self-similarity patterns into salient regions using adaptive ratios. Collectively, our unified framework, $S^{2}$-FracMix, enables simultaneous learning from fractal and non-fractal structures within a single image, yielding a targeted and structurally coherent augmentation strategy. We theoretically analyze the advantage of our technique, and empirically establish its superiority over the existing methods by achieving state-of-the-art performance in extensive evaluation with seven benchmarks across classification (coarse and fine-grained), robustness, calibration, object detection, and transfer learning tasks. Project page is available at \href{https://fracmix-data-augmentation.github.io/}{fracmix-data-augmentation.github.io}
arXiv:2606.25480v1 Announce Type: new Abstract: Unmanned aerial vehicle (UAV) can provide on-demand, high-capacity connectivity in disaster and normal situation. However, it faces a challenge of curse of dimensionality in trajectory optimization, where interference-limited environments and vast search spaces make real-time coordination computationally expensive. To overcome this challenge, we propose the Rate-Aware Quantum-Annealed Graph Condensation (RA-QAGC) scheme, which combines rate-aware graph abstraction with decentralized reinforcement learning to enable scalable, interference-aware UAV coordination. By identifying high throughput locations and guiding UAV trajectory adaptation toward throughput-optimal regions, RA-QAGC effectively balances network capacity by maintaining quality-of-service (QoS) requirements. Simulation results demonstrate the proposal outperformed over existing schemes by achieving 59.4 Mbps total throughput and 23.9 Mbps priority-user throughput, representing gains of approximately 15% and 34%, respectively, over the baseline schemes.
arXiv:2606.25796v1 Announce Type: new Abstract: Localization in challenging environments, such as GNSS-denied, geometrically repetitive, or textureless scenes commonly found in offices, hotels, and underground parking facilities, remains an open problem for reliable autonomous mobile robot (AMR) deployment. Single-modality localization methods are inherently limited by the constraints of individual sensors. Although multimodal fusion frameworks have shown improved robustness, most existing approaches still rely heavily on geometric or texture features, or on infrastructure-based beacons, which increase installation and maintenance costs while reducing deployment flexibility. Recently, ambient magnetic field (AMF)-based localization has attracted growing attention because it does not depend on geometric or texture features, nor does it require additional infrastructure, making it a promising complementary modality for AMR localization. However, existing studies have only explored such fusion in pedestrian scenarios using smartphone-mounted sensor suites, and practical solutions for AMR systems remain largely unexplored. To address this gap, this article proposes a magnetometer-inertial-LiDAR fused multimodal localization framework with a custom-designed sensor suite, termed MIL-LC, which provides reliable localization even when LiDAR suffers from geometric degeneration or when the magnetic map changes during long-term deployment. Extensive experiments in both simulation and real-world environments demonstrate that the proposed MIL-LC framework achieves robust and accurate localization performance.
arXiv:2606.25363v1 Announce Type: new Abstract: Mathematical knowledge is organized around statements and their dependencies, but this structure is exposed unevenly: informal papers cite mostly at the document level, while formal libraries record fine-grained dependencies over a much smaller body of mathematics. We introduce TheoremGraph, a unified statement-level dependency graph spanning both informal and formal mathematics. On the informal side, we parse 11.7M theorem-like environments from mathematics arXiv and recover 18.3M candidate directed dependencies, each labeled by the extractor that proposed it so downstream users can trade coverage for precision. On the formal side, we release LeanGraph, a Lean 4 elaborator-level extractor producing 388,105 declaration nodes and 11.3M typed edges across 25 Lean projects. We bridge the two graphs by embedding generated natural-language slogans into a shared semantic space, linking related statements across papers and across the informal/formal divide; an LLM judge affirms 47,952 such matches above a 0.8 cosine floor, with the judge-acceptance rate rising from 48% across the floor to 87% in the >=0.9 tier. On formal concept retrieval, our name-and-signature representation with graph expansion comes within 0.5pp of LeanSearch v2's reranked Recall@10 (0.775 vs. 0.780) without an LM reranker. We release the dataset, extractors, HTTP API, and MCP interface as infrastructure for mathematical search, attribution, and retrieval-augmented reasoning, available at theoremsearch.com and huggingface.co/datasets/uw-math-ai/theorem-matching.
arXiv:2606.24938v1 Announce Type: cross Abstract: Tensor decomposition of donor $\times$ cell-type $\times$ gene single-cell data recovers \emph{multicellular programs}: coordinated axes of inter-individual transcriptional variation that span cell types and stratify disease. Yet immune single-cell atlases are increasingly multi-institution, multi-ancestry, and governed, so patient cells often cannot be pooled. We present a federated estimator: each site computes a local program subspace, and a coordinator merges these by stacked SVD under federated global-mean centering, provably equivalent (up to truncation) to the centralised decomposition. This centering makes the merge robust to site-label confounding (program AUC $0.957$ vs.\ $0.861$ for naive per-site centering). Only program subspaces leave a site, and aggregation is compatible with secure aggregation. On a 261-donor systemic lupus erythematosus atlas it recovers the canonical interferon program (ISG enrichment AUC $0.998$; case--control separation $0.958$; bootstrap $\Delta\text{AUC}=-0.000$, 95\% CI $[-0.004,+0.012]$ vs.\ centralised), across institution-scale and multi-ancestry partitions, and across three \emph{real} COVID-19 sites (subspace correlation $0.989$). It recovers the program when \emph{no site observes all cell types} (correlation $1.000$, exact by construction), which fixed-feature federated PCA cannot. On an interstitial-lung-disease atlas the recovered program predicts disease better than the best single cell type (AUC $0.96$ vs.\ $0.91$; gap 95\% CI excludes zero) and the advantage survives federation; a liver cohort is consistent ($p=0.005$). Membership-inference shows secure aggregation cuts attack AUC from $0.91$ to $0.61$. The method enables cross-institution, cross-ancestry recovery of multicellular immune programs without sharing cells.
arXiv:2606.25194v1 Announce Type: cross Abstract: Topological information is widely used in data analysis, network design, and machine learning, and topological constraints naturally arise when optimizing or generating objects with prescribed homological structure. However, directly controlling Betti numbers and persistent homology is difficult because they are discrete and combinatorial. We propose a differentiable framework for topology-constrained optimization based on Hodge-spectral relaxations of homological constraints and low-pass spectral filters. From soft graphs and soft clique complexes, we construct Hodge-Laplacian-type spectral relaxations that unify graph clique complexes and Vietoris--Rips filtrations of point clouds. In the hard limit, the penalty-regularized ambient operator recovers the ordinary Hodge Laplacian on the active subcomplex, while in the soft regime it serves as a differentiable low-frequency spectral surrogate. Homological information is represented by zero and near-zero modes, and differentiable topological objectives are defined using heat filters, resolvent filters, and polynomial Laplacian moments. For point clouds, we show that the proposed Hodge spectral-filter losses yield more spatially distributed gradients, smoother scale-normalized behavior under persistence-pairing changes, and geometry-aware update directions than persistent-homology-based losses. For graph clique complexes, Laplacian moments control normalized first-Betti-type quantities and can be combined with ordinary graph-feature objectives. We also discuss connections to trace-based normalized Betti-number estimation, polynomial spectral methods, and possible quantum trace estimation.
arXiv:2606.25463v1 Announce Type: cross Abstract: This study introduces Blasto-Net, a multi-task deep learning model for comprehensive blastocyst analysis. The proposed model performs three tasks simultaneously in a single forward pass: segmentation of the ZP, TE, and ICM compartments, morphological grading, and implantation outcome prediction. Accurate blastocyst analysis in in vitro fertilization (IVF) is challenging. The compartments often have similar textures but very different structures. To address these challenges, Blasto-Net employs an EfficientNet-B3 encoder with a UNet-style decoder enhanced by the Convolutional Block Attention Module (CBAM) and a novel Edge-Aware Attention Module (EAAM) to effectively capture both semantic and boundary information. To handle distinct compartment topologies, the network employs specialized segmentation heads and a composite region- and boundary-based loss. Additionally, Grad-CAM++ visualizations are used to verify the anatomical consistency of the model's predictions. Evaluated on a public HMC blastocyst dataset, Blasto-Net achieves Dice scores of 94.93%, 91.60%, and 88.82% for ICM, ZP, and TE, respectively, alongside an implantation F1-score of 80.0%. These results demonstrate that Blasto-Net offers an accurate, interpretable, and efficient solution for automated blastocyst assessment, with strong potential to support clinical decision-making in IVF.
arXiv:2606.25527v1 Announce Type: new Abstract: Online reinforcement learning (RL) agents increasingly depend on knowledge acquired offline to achieve practical efficiency. Originally studied in offline-to-online RL, this paradigm now spans foundation model post-training and embodied intelligence, with prior types expanding from offline datasets and pre-trained policies to increasingly diverse knowledge sources such as multimodal foundation models and generative world models. Offline priors have become central to how deep RL is developed and deployed. However, this reliance introduces a challenge that the prevailing benchmark-driven paradigm cannot resolve: because prior validity varies across deployments and shifts during training, no single approach to managing it is universally optimal, and benchmark rankings offer limited guidance for real-world deployments. Rather than pursuing universal solutions, we argue that the field should shift to diagnosis-driven tension management, in which deployment-specific evidence guides how the learner relates to its priors throughout training, enabling both flexible and adaptive deployment. We support this position with a framework characterizing how priors reshape online optimization through three functional roles, controlled experiments demonstrating help-or-hurt reversals, cross-domain evidence from foundation model post-training to embodied intelligence, and engagement with five substantive counterarguments.
arXiv:2606.25528v1 Announce Type: new Abstract: The particle-in-cell (PIC) simulation method is often understood to solve the collisionless Vlasov equation due to the finite shape of its macroparticles. In reality, it can suffer from artificially high collisionality due to the underresolution of particle number; i.e., the use of a large macroparticle weight. The degree to which particle shape effects compensate for a large macroparticle weight in 1D, 2D, and 3D is presented. The collision time is calculated from PIC simulations based on the decay rate of the velocity autocorrelation function and compared directly with the kinetic theory of Okuda, Birdsall, and Langdon. The theory is found to accurately predict the simulated collision time with varied grid spacings, plasma conditions, and simulation dimensionalities. The result is a means to predict the timescale of self-consistent Coulomb interactions in the PIC simulation and thus characterize the relevance and implications of numerical thermalization as a function of grid spacing and macroparticle weight. It is determined that reaching the physical thermalization time, let alone approximating the collisionless Vlasov limit, may often be intractable in 3D for macroparticle sizes that resolve the Debye length.
arXiv:2606.25401v1 Announce Type: new Abstract: Accurate channel state information (CSI) prediction is essential for proactive beamforming and resource management in 5G massive MIMO systems, yet the deployment of high-accuracy transformer-based predictors on base-station hardware remains challenging because the most capable models carry upwards of 30\,M parameters. This paper introduces Lightweight PCGAE-Net, which addresses the efficiency problem not by post-hoc compression but by correcting two architectural flaws in the current state of the art. The first is a sequential attention ordering bias: in CS3T-UNet, group-wise temporal attention (GTA) always operates on features that have already been transformed by cross-shaped spatial attention (CSA), distorting what temporal information GTA can capture. We remove this dependency by routing both attention modules to the same layer-normalized input and combining their independent outputs through a learned per-channel sigmoid CrossGate. The second flaw is an uncompressed bottleneck: applying full self-attention at the deepest encoder stage, where channel depth reaches $4C$, is quadratically expensive and carries redundant features. A Bottleneck AutoEncoder (BAE) with $1\times1$ convolutions halves this depth and uses an auxiliary reconstruction loss to prevent information collapse. Wrapping these components inside a shallower encoder-decoder with frequency-domain dimensionality reduction ($N_f\!=\!32$, $C\!=\!48$) produces a model with just 8.54\,M parameters -- 58\% fewer than the CS3T-UNet baseline -- that outperforms it by up to 3.26\,dB at 5\,km/h and 6.0\,dB at 9\,km/h in single-step prediction on QuaDriGa dataset.
arXiv:2606.25821v1 Announce Type: new Abstract: Sparse Mixture-of-Experts (MoE) architectures have emerged as an increasingly influential paradigm as they offer a strategic balance between parameter scalability and computational efficiency. However, low-resource languages, which suffer from a scarcity of high-quality training data, often have their tokens routed to different experts than those predominantly activated by high-resource inputs, which limits cross-lingual expert sharing. This cross-lingual routing divergence consequently hinders their efficacy in multilingual contexts. To address this issue, we propose SARA (Semantically Anchored Routing Alignment), a framework designed to transfer specialized capabilities from high-resource languages as anchors to low-resource languages. SARA explicitly aligns the routing distribution of multilingual inputs with high-resource semantic anchors using a symmetric Jensen-Shannon (JS) divergence constraint. Unlike traditional distillation methods that operate on output logits, SARA directly aligns the internal routing distributions of MoE layers, encouraging mechanistic consistency in expert selection across languages. We conduct experiments on 2 LLMs across 5 low-resource languages and 3 benchmarks. Experiment results demonstrate that SARA outperforms standard instruction tuning, e.g., +0.8% on Qwen3-30B-A3B and +1.2% on Phi-3.5-MoE-instruct on Global-MMLU. Further analyses show that SARA effectively addresses performance bottlenecks in low-resource languages, providing a scalable pathway to enhance multilingual capabilities in sparse architectures.
arXiv:2606.25823v1 Announce Type: new Abstract: Manufacturing companies look increasingly at Private 5G networks to manage Automated Guided Vehicles (AGVs). While 5G promises Ultra-Reliable Low Latency Communication (URLLC), its service quality is challenged by industrial environments characterized by dense metallic structures, which frequently cause line-of-sight (LOS) blockage events, causing deep fades in received signal levels that can degrade channel capacity to near-zero. Standard transport protocols and rate adaptation mechanisms fail to react sufficiently fast to these deep fades, resulting in bufferbloat and latency spikes that violate safety margins. In this paper, we propose a cross-layer rate control algorithm based on Lyapunov Drift-plus-Penalty theory. The proposed controller dynamically optimizes the trade-off between service utility and queue stability based on instantaneous buffer states, without requiring predictive channel models. We validate the approach using a trace-driven simulation framework that replicates the stochastic dynamics of 5G blockage using 3GPP-compliant capacity data. Numerical results demonstrate that while baseline scheduling schemes suffer from catastrophic queue accumulation, leading to excessive delays upon reconnection, the proposed Lyapunov controller effectively eliminates bufferbloat. By preventing congestion-induced backlog, the system ensures immediate low-latency operation as soon as the channel recovers, maintaining near-deterministic behavior.
arXiv:2606.25410v1 Announce Type: new Abstract: Machine unlearning is an emerging domain that ensures the safe removal of elements (includes concepts, attributes, entity and class) from the trained model along with least drop in model performance. The domain of machine unlearning brings its own indigenous challenges since the removal of pre-trained elements from model will always degrade the model performance on remaining elements. The existing methods basically rely on retraining for removal of elements from the pre-trained model, which is compute extensive. In this work, we propose a machine unlearning method which helps to reduce the computational requirement for faster retain-dataset accuracy convergence which also does not require extensive retraining of the pre-trained model. The proposed method, Data-Frugal Machine Unlearning (DFMU) requires only a single forward and backward pass for computing the importance score of various computational blocks of a model. The importance score computation is based on knowledge preserving pruning which helps to converge faster and requires far less data as compared to the existing methods. Experimentally, it achieves 40% more retain-accuracy with just 13% of data samples in comparison with SOTA method on various public datasets and also averages 88% faster processing time for forgetting a given class.