arXiv:2607.06653v1 Announce Type: new
Abstract: Federated learning (FL) enables collaborative model training across institutions without sharing sensitive patient data. However, the inherent heterogeneity of electrocardiogram (ECG) data across healthcare providers presents significant technical challenges for robust classification. We propose FedDualAtt, a personalized federated learning approach that splits transformer attention heads into global and local branches. Global heads are aggregated via FedAvg to capture shared cross-site patterns, while local heads remain client-specific to adapt to institution-level recording characteristics. Experiments on FedCVD, an FL benchmark for cardiovascular disease detection, demonstrate that FedDualAtt outperforms existing FL and personalized FL methods in ECG classification tasks. Analysis of global-local head ratios reveals that different clients benefit from varying levels of architectural personalization.
Science Journals
arXiv:2607.07356v1 Announce Type: new
Abstract: Magnetic fields generated by the nonlinear Rayleigh-Taylor growth of laser-seeded three-dimensional broadband perturbations were measured in laser-accelerated planar targets using ultrafast proton radiography. The experimental data show self-similar behavior in the growing cellular magnetic field structures. These observations are consistent with a bubble competition and merger model that predicts the time evolution of the number and size of the bubbles, linking the cellular magnetic field structures with the Rayleigh-Taylor bubble and spike growth.
arXiv:2607.07372v1 Announce Type: new
Abstract: The Quantum Control Processor (QCP) bridges the gap between compiler toolchains and control electronics, and is responsible for translating compiled quantum circuits into executable instructions that directly manipulate qubits and handle measurement feedback. However, existing designs rely primarily on customized instruction sets, limiting design reuse and requiring significant effort to build supporting toolchains. Furthermore, efficiently addressing qubits and scheduling operations in highly scalable scenarios remains a critical challenge. In this work, we present a vectorized quantum control approach built upon the RISC-V Vector (RVV) engine with a quantum-oriented extension. Leveraging the high parallelism of RVV, our approach can address up to 128 qubits in a single instruction. We also embed parameterized rotation information into the instruction set, enabling dynamic tuning of gate rotations in hybrid quantum-classical programs. To support mid-circuit measurements, we design a hardware-based halt-resume protocol that resumes pipeline execution within 80 $ns$ of receiving the measurement result. Comprehensive evaluation using both RISC-V toolchains and FPGA prototypes demonstrates that our design achieves up to 2.52$\times$ speedup over the baseline in program execution time, with excellent scalability.
arXiv:2607.07160v1 Announce Type: new
Abstract: We introduce the Stable Matching Problem with Minimum Utility Gap, which seeks a stable matching in which the utilities received by individual agents are as balanced as possible. Our framework can handle many-to-many matchings and general utility functions on partner sets that are consistent with the agents' preferences. We consider two measures for comparing agents' utilities: the difference between the maximum and minimum utilities, and their ratio.
We provide a polynomial-time algorithm for both versions. The algorithm exploits the rotation-poset representation of the set of stable matchings and, in particular, the fact that the rotations affecting each agent form a chain in this poset. To position our result, we also clarify its relation to existing frameworks: we show that our objectives are not captured by the recent minimum-cut representability framework, while identifying a special case that admits a submodular function minimization interpretation.
arXiv:2607.07165v1 Announce Type: new
Abstract: We propose an optimized stencil strategy for the Generalized Finite Difference Method (GFDM) applied to non-linear problems. We take advantage of the flexibility of GFDM to engineer specific stencils by agglomerating nodes and balancing size with numerical accuracy. Previous work, focusing on the stencil construction for the linear convection-diffusion problem, showed that optimizing stencils and scaling parameters improves the scheme. The present study aims to investigate similar benefits for non-linear problems such as the Burgers' equations and the weakly compressible Navier-Stokes system.
arXiv:2607.07400v1 Announce Type: new
Abstract: In programming courses, instructors may need to interpret whether a submission is consistent with a student's prior programming profile, especially when code similarity alone is inconclusive. Existing source-code authorship methods are often evaluated on programming-contest or open-source datasets, where reusable templates and local code patterns can produce strong author-related signal. Educational repositories present a different setting. Students solve shared assignments while their programming practices are still developing. This study uses task-aware evaluation to contrast these production contexts and tests whether repository-visible process features add information beyond final code in six matched educational comparisons. Contest data provide a high-signal contrast, with a Kick Start mean top-1 of 0.938. Educational datasets produce substantially lower attribution performance. Adding process features raises the educational mean from 0.094 to 0.233 and mean pairwise verification ROC-AUC from 0.556 to 0.752. The comparisons show that measured signal depends on production context and that process patterns can complement weak final-code signal in educational repositories. Such models are therefore appropriate only as instructor-mediated decision support, not as independent proof of authorship.
arXiv:2607.07438v1 Announce Type: new
Abstract: Action Quality Assessment (AQA) aims to evaluate how well a person performs a movement, which is essential in applications such as sports scoring, skill assessment, and healthcare. However, unimodal approaches often struggle to capture subtle cues of movement quality in real-world settings. Although multi-modal inputs provide complementary information, existing methods still face two major challenges: heterogeneous modalities often lead to cross-modal misalignment and unstable fusion, and reliable multi-modal annotation is costly, resulting in limited dataset diversity. To address these challenges, we propose DualAlign, a two-stage multi-modal fusion framework with adaptive alignment. The framework first constructs a coherent visual representation by maximizing shared structural information across RGB video, optical flow, and skeleton modalities. Textual semantics are then incorporated after visual stabilization, allowing high-level descriptions to complement rather than distort the underlying visual manifold. To evaluate the framework under realistic multi-modal conditions, we introduce MM--JDM, a movement-quality assessment dataset integrating RGB videos, optical flow, skeleton sequences, and structured text. MM--JDM naturally exhibits modality noise, class imbalance, and label scarcity, making it a challenging benchmark for studying multi-modal fusion and alignment. Extensive experiments show that DualAlign improves average correlation on MM--JDM by 21.16% over the state-of-the-art methods and achieves gains of 3.53% and 5.95% on the RG and Fis-V benchmarks, respectively. DualAlign also remains robust under missing-modality and label-scarce conditions.
arXiv:2607.06818v1 Announce Type: new
Abstract: For any E-commerce website it is a nontrivial problem to build enduring advertisements that attract shoppers. It is hard to pass the creative quality bar of the website, especially at a large scale. We thus propose a programmatic solution to generate product advertising headlines using retail content. We propose a state of the art application of Reinforcement Learning (RL) Policy gradient methods on Transformer based Masked Language Models. Our method creates the advertising headline by jointly conditioning on multiple products that a seller wishes to advertise. We demonstrate that our method outperforms existing Transformer and LSTM + RL methods in overlap metrics and quality audits. We also show that our model-generated headlines outperform human submitted headlines in terms of both grammar and creative quality as determined by audits.
arXiv:2604.07308v2 Announce Type: replace-cross
Abstract: Conventional delay-Doppler (DD) communication and sensing systems require transmitting pilot frames at every channel coherence time interval in order to keep track of channel variations at the cost of spectral efficiency. In this paper, we propose an approach to utilize data transmissions that modulate arbitrary waveforms with zero-mean, unit average energy symbols for DD channel estimation without requiring pilot transmissions in every coherence time interval. Numerical evaluation over practical doubly-selective channel models demonstrate $\sim 1.8 \times$ improvement in uncoded spectral efficiency with our proposed data-based approach over conventional pilot-based approaches across various $6$G modulation schemes.
arXiv:2607.06838v1 Announce Type: new
Abstract: Humans can navigate an unfamiliar city and gradually form a coherent spatial mental map spanning tens of square kilometers. Can AI build spatial representations at a comparable scale? Although recent foundation models have advanced scene reconstruction and embodied intelligence, scaling to entire cities remains an open challenge, primarily due to the lack of city-scale data. To bridge the gap, we introduce WildCity, a real-world multimodal dataset collected by autonomous fleets traversing complex urban environments. Our dataset includes 18 trajectories, each averaging 83.7 kilometers in length, and preserves the core challenges of in-the-wild perception, e.g., dynamic objects, lighting variations, and imperfect camera poses. We further establish an urban-tailored reconstruction baseline and convert the reconstructed environments into a closed-loop simulator. Beyond the dataset and baseline, we systematically analyze the key challenges on the path to simulation-ready urban digital twins: scalability, extrapolation, and uncertainty. Ultimately, WildCity aims to catalyze progress not only in city-scale rendering, but more broadly in the pursuit of AI that can perceive, remember, and reason across space at a scale comparable to human cognition. Project page: https://han-xiangyu.github.io/Wild-City/
arXiv:2607.07486v1 Announce Type: new
Abstract: 3D Morphable Models (3DMMs) remain the standard parametric shape priors for many state-of-the-art 3D face reconstruction algorithms. However, as these models are derived from a finite number of 3D face samples, they inherit the morphological biases of their training data, potentially limiting their generalizability across diverse global populations. In this paper, we propose a novel framework to analyze 3DMM reconstructions through the lens of surface curvature, with the objective to discover, quantify and visualize biases. While standard evaluation metrics often rely on Euclidean distances, our reconstruction error captures subtle surface nuances such as local topology or undulations. To do so, we leverage the Laplace-Beltrami Operator (LBO) to generate high-resolution curvature error maps, providing a localized and geometrically meaningful visualization of discrepancies between ground truth faces and reconstructed meshes. We derive from it an error metric that we validated through a user study, observing a significantly higher correlation to human perception compared to traditional methods. Furthermore, we conduct extensive experiments across several 3DMM bases and fitting algorithms, uncovering systematic age-related biases and providing preliminary evidence of biases associated with gender and ethnicity. Our findings highlight the necessity of adopting curvature-aware evaluation protocols to ensure demographic fairness and geometric precision in future 3D face reconstruction research.
arXiv:2607.07575v1 Announce Type: new
Abstract: While the thermodynamically formulated generalized Prandtl-Ishlinskii stop-type operator effectively captures hysteresis nonlinearities, it requires a local iterative procedure to update each hysteron, resulting in considerable computational effort. In this work, we propose a simplified thermodynamic formulation of the generalized Prandtl-Ishlinskii stop operator. The nonlinear mapping on the stop operator is replaced by an identity, such that the hysteresis operators are directly weighted through their outputs, while the nonlinear anhysteretic response, represented by ramp dead-zone basis functions, is fully preserved. For isotropic cases, this simplification enables a closed-form solution for the local plastic correction, eliminating per-hysteron iterative Newton updates. The resulting constitutive mapping is integrated into a finite element solver, and numerical results show a significant reduction in computation time with accuracy comparable to the generalized model.
arXiv:2607.06997v1 Announce Type: new
Abstract: Biochemical traveling waves transmit signals across cells and tissues, but the thermodynamic cost of reliable propagation remains unclear. We develop a stochastic thermodynamic framework for reaction--diffusion systems with stable traveling waves and show that diffusion of the wave position is bounded by the dissipation specifically associated with propagation. The bound follows by projecting noisy field dynamics onto the adjoint translational mode, which maps the wave position to an effective biased random walk. Its tightness is controlled by the non-self-adjoint part of the linearized dynamics, with finite wave speed and antisymmetric reaction dynamics generically producing deviations from equality. For excitable trigger waves in a FitzHugh--Nagumo model, we show that the slow inhibitor dominates the propagation cost, yielding a trade-off among wave speed, inhibitor amplitude, and dissipation. We test these predictions in stochastic simulations of a microscopic Belousov--Zhabotinsky reaction--diffusion system and find consistent signatures in mitotic trigger-wave experiments in \textit{Xenopus} egg extracts. The same relation further imposes an annihilation-limited bound on the reliable signaling rate of wave trains.
arXiv:2607.07697v1 Announce Type: cross
Abstract: The induced Erd\H{o}s--P\'osa property in graphs relates the maximum number of pairwise anti-adjacent copies of an object with the minimum number of neighborhoods required to hit all copies. In this paper, the objects we consider are long cycles and long thetas, both as induced minors. Let $C_t$ denote the cycle with $t$ vertices and let $\Theta_t$ be the graph consisting of three internally disjoint and anti-adjacent paths, each with $t$ internal vertices, connecting the same pair of distinct vertices.
We show that for every fixed $t$, both $C_t$ and $\Theta_t$ have the induced Erd\H{o}s--P\'osa property with respect to the induced minor relation. More precisely, for every integer $k$ and every graph $G$, one of the following two outcomes occurs: (i) $G$ contains $k$ pairwise vertex-disjoint and anti-adjacent copies of $C_t$ (resp., $\Theta_t$) as induced minors, or (ii) there is a set $X \subseteq V(G)$ of size $\mathcal{O}(tk \log k)$ such that the set $N[X]$, consisting of $X$ and its neighbors, hits all $C_t$ (resp., all $\Theta_t$) induced minors in $G$.
This resolves in a strong form a special case of a conjecture of Ahn, Gollin, Huynh, and Kwon [SODA 2025]. From these results we derive that graphs that exclude $k$ disjoint copies of $\Theta_t$ as an induced minor admit balanced separators consisting of the neighborhood of $\mathcal{O}(tk \log k)$ vertices. This in turn resolves a special case of a conjecture of Gartland and Lokshtanov and, combined with known techniques, yields a QPTAS for Maximum Weight Independent Set and a number of its generalizations.
arXiv:2607.07450v1 Announce Type: new
Abstract: Computational on-chip spectrometers are emerging as a powerful platform for portable spectral analysis, combining photonic integration with advanced signal processing to enable a wide range of in-situ sensing applications. We propose a broadband reconstructive spectrometer based on wave chaos in a stadium microresonator with a nanostructured scattering layer for full-area speckle readout. Wavelength dependent interference within the chaotic microresonator encodes the spectral information into a spatial intensity pattern that can be computationally inverted to reconstruct the input spectra. The optimal fabrication parameters of the SU-8 polymer nanostructured layer yield a surface roughness of 176nm and a root mean square thickness of ~2um. We experimentally validate our spectrometer at visible and infrared wavelengths, with resolutions of 43pm at 630nm and 8.2pm at 1550nm. The spectral reconstruction is demonstrated for single and multiple narrowline sources as well as for a broadband (~1nm) pulsed laser source. The broad experimental validation and compact footprint (0.05mm2) establishes our chaotic microresonator-based speckle spectrometer as a robust and versatile platform for high-resolution, on-chip spectral sensing.
arXiv:2603.12775v2 Announce Type: replace
Abstract: We introduce the concept of Randomly Modulated Gaussian Processes as a unifying framework for elaborating, analyzing and classifying anomalous diffusion models in annealed heterogeneous media. This formulation incorporates correlations in the displacements together with correlated fluctuations of their amplitudes. Most known models of anomalous diffusion (including continuous-time random walk, fractional Brownian motion, and L\'evy flights) and random diffusivity can be described and further generalized within this framework. Moreover, the unified view identifies the main statistical properties to be probed experimentally for a reliable classification of diffusive dynamics. The proposed matrix formulation facilitates the computation of the first four moments and allows for a systematic statistical characterization of the considered processes. The necessary and sufficient conditions are provided for the emergence of anomalous diffusion. General expressions for the non-Gaussian parameter, the ergodicity breaking parameter and the covariance of squared increments are derived. An expression for the characteristic function and the codifference (i.e., a generalized measure of correlations) are obtained and used to study the special cases of L\'evy flights and Laplace motion with correlated displacements. Potential applications of this framework for systematic analysis and biophysical interpretations of experimental single-particle trajectories are discussed.
arXiv:2607.07029v1 Announce Type: new
Abstract: Reinforcement learning (RL) policies can be unsafe and vulnerable to attacks. Ensuring their reliability is often a pain point as existing automated testing methods target only selected environments, testing scenarios, and RL algorithms. To address this, we propose a comprehensive framework for testing single- and multi-agent RL policies under varying conditions. Our implementation of this framework, Gimitest, is an open-source tool that supports various gym frameworks and allows for modifications of their integrated components. This article describes the framework and details Gimitest's functionality and architecture. It showcases its effectiveness in testing multiple RL policies in environments such as the official Farama Gymnasium and PettingZoo.
arXiv:2607.07034v1 Announce Type: new
Abstract: We introduce Intrinsic Green's Learning (IGL), a framework that models a target function on a manifold as the solution to a linear PDE whose source term is learned from data. Rather than approximating the target directly, IGL learns a source and integrates it against a Green's kernel. An encoder discovers a low-dimensional coordinate chart on the manifold where both the source and the kernel decompose as low-rank tensors, collapsing a high-dimensional integral into independent one-dimensional integrals with cost linear in the intrinsic dimension. A two-stage algorithm separates coordinate discovery from source fitting, a near-convex linear solve, preventing the dimensional collapse of joint training. Learnable gates on each coordinate automatically discover the intrinsic dimension of the manifold. We validate IGL on synthetic manifolds and on MNIST, where it simultaneously achieves near-optimal classification and automatic recovery of the intrinsic dimension.
arXiv:2607.07035v1 Announce Type: new
Abstract: The architecture of deep feedforward neural networks is ubiquitous in deep learning, either as a whole system or as a subnetwork of other architectures, and thus its mechanism is a key ingredient of the black box of neural networks. On the basis of the simplest two-layer ReLU network, this paper systematically studies the mechanism of deep feedforward ReLU networks with multiple hidden layers and successfully explains the training solution obtained by the back-propagation algorithm. The concept of a path, especially in terms of the relationships between paths, plays a central role in uncovering the mystery of the black box. It is shown that a unit of a deep ReLU network can form a piecewise linear manifold to divide the input space, instead of a hyperplane of the two-layer case. How to efficiently use the hidden-layer units to produce both linear functions and partitions of the input space is also a central problem. The principles of a two-layer ReLU network can be generalized to the deeper case to a large extent, such as multiple strict partial orders and continuity restriction. The combination of the basic and simple principles proposed can yield complicated instantiations including the training solutions, and in this sense the black box of deep feedforward ReLU networks is revealed.
arXiv:2607.07678v1 Announce Type: new
Abstract: Rotary Position Embeddings (RoPE) provide transformers with a fixed grid of positional frequencies, yet trained models use these frequencies highly non-uniformly. We study what determines this frequency usage and propose a data-centered explanation: RoPE frequencies are selected to match the relative-distance structure of the training data. Viewing each frequency as a positional lens, we formalize a field-resolution tradeoff and show that, for a data-induced dependency profile of width $W$, the optimal frequency scales as $1/W$. This frequency-matching principle explains controlled observations on synthetic and text-based data, and suggests that the mid-low frequency bands observed in language models arise from the multi-scale dependency structure of natural language. We further connect frequency selection to position-interpolation-based length generalization: scaling frequencies down expands the effective field while reducing resolution. This helps when longer-context dependencies are approximate dilations of those seen during training, but can fail when relevant dependencies do not scale with context length. Empirically, we show that natural language exhibits approximate self-similarity across positional scales, explaining why test-time frequency scaling can support long-context generalization. Overall, our results identify a data-driven mechanism behind emergent RoPE frequency usage and show that long-context generalization depends on two forms of scale matching: between learned frequencies and training-time dependencies, and between frequency scaling and how those dependencies extend to longer contexts.
arXiv:2607.07066v1 Announce Type: new
Abstract: Transformers have demonstrated a remarkable ability to learn algorithmic reasoning, yet mechanistic analyses have mostly focused on globally invertible operations such as cyclic addition and group composition. In this work, we investigate how small transformers learn modular integer multiplication over composite moduli, a fundamentally non-invertible operation due to the presence of zero-divisors. We propose the monoid extension: a localized generalization of Group Composition via Representation (GCR) that suggests the learned computation does not rely on a single global representation space. Instead, the model partitions the input space into local hierarchical algebraic regions, where group-like structure survives and Fourier mechanisms can be applied. In transformers trained on square-free modular multiplication, we find that embeddings organize around these regions, attention exhibits class-sensitive routing and low-rank write directions, and local character features explain a large fraction of the model's output logits. Our results suggest that representation-theoretic mechanisms previously identified for group operations can extend beyond groups to more general structures.
arXiv:2607.07494v1 Announce Type: new
Abstract: Gradient communication is a primary scaling bottleneck in large language model (LLM) pretraining. Communicating gradients in low-precision formats, such as FP8 and NVFP4, can significantly reduce the communication volume. Existing methods quantize gradients via linear or nonlinear mappings in Euclidean space, often degrading model performance because highly anisotropic gradients incur direction-dependent distortion. We present GIFT, a geometry-informed gradient scaling method that performs low-precision communication in geometry-aware coordinates. By transforming gradients into a near-isotropic space before quantization, GIFT makes low-precision representations substantially more faithful to their high-precision counterparts. GIFT only changes the coordinate system used for low-precision gradient communication and does not change the optimizer, training recipe, communication collective, or low-precision format. We also develop a simplified geometry-aware transformation algorithm with low-rank approximation and selective application to balance the computation overhead and communication reduction. We examine the empirical convergence of GIFT using Llama-300M and Llama-600M models. Our results show that GIFT reduces the end-to-end pretraining time of Llama-600M by 7.6% on 64 NVIDIA GH200 Superchips, while improving the downstream task preservation profile over direct Euclidean FP8 communication under the same optimizer and communication path.
arXiv:2607.07101v1 Announce Type: new
Abstract: Proprioception is fundamental to robotic manipulation, yet standard fusion methods often treat it as an isolated vector lacking explicit alignment with visual tokens. Without a direct correspondence between 3D kinematics and 2D feature maps, manipulation policies struggle to ground the robot's state within the scene, frequently underperforming even vision-only baselines. To address this, we introduce GeoProp, a lightweight, plug-and-play adapter that aligns proprioception with vision through explicit geometric grounding and spatial feature sampling. GeoProp projects the robot state onto the image plane to sample localized visual features, constructing a grounded state token. It then injects state-derived spatial priors into the corresponding visual features via FiLM modulation. To capture motion intent, GeoProp further samples features at a short-horizon predicted coordinate derived from recent kinematics, providing look-ahead visual context. Across 67 tasks, GeoProp improves Diffusion Policy by 8.7% on 63 simulation tasks and pi_0 by 4.0% on the RoboTwin subset, and yields a 10.6% average gain across both policy families in the real world, while adding only 2-3% to the parameter count. These results demonstrate that GeoProp is a simple yet high-impact inductive bias for generalist embodied policies. Project page: https://alibaba-damo-academy.github.io/GeoProp/.
arXiv:2502.05906v2 Announce Type: replace
Abstract: We consider a strategic M/M/1 queueing model under a first-come-first-served regime, where customers are split into two classes and class $A$ has priority over class $B$. Customers can decide whether to join the queue or balk, and, in case they have joined the queue, whether and when to renege. We study the equilibrium strategies and compare the equilibrium outcome and the social optimum in the two cases where the social optimum is or is not constrained by priority.
arXiv:2607.07142v1 Announce Type: new
Abstract: Low Earth Orbit (LEO) satellite Internet introduces rapid path variability, intermittent capacity shifts, and non-terrestrial delay dynamics that challenge transport-layer congestion control. Although Bottleneck Bandwidth and Round-trip propagation time (BBR) achieves high throughput in such environments, its aggressive bandwidth probing can cause excessive retransmissions and unstable pacing over LEO links. This paper presents a global experimental evaluation of BBR over a SpaceX Starlink testbed spanning six geographically distributed AWS endpoints and compares its behaviour against Cubic, Vegas, and Hybla under isolated and competing traffic scenarios. The measurements show that BBR consistently delivers superior throughput but incurs significantly higher retransmission overhead, revealing a critical throughput-stability trade-off in LEO satellite Internet. To address this limitation, we propose a Small Language Model (SLM)-guided BBR adaptation framework that learns phase-safe pacing-gain decisions from real Starlink traces. The framework combines a structured BBR state encoder, LoRA-based parameter-efficient fine-tuning, and a constrained networking head to generate feasible pacing actions with low inference latency. Evaluation using GPT-2, T5, GPT-Neo, and SmolLM2 shows that lightweight SLMs can retain BBR's throughput advantage while substantially reducing retransmissions, with performance comparable to larger language models but at much lower computational cost.