arXiv:2605.17471v1 Announce Type: new
Abstract: Quantization-aware training (QAT) is widely adopted to quantize language models by training full-precision weights using gradients from the quantized model. The main bottleneck is its slow convergence and early performance plateau, particularly below 4-bit-widths. While this problem has been observed in prior work, its precise cause remains unclear. In this paper, we analyze the convergence of QAT by estimating the spectrum of the loss-surface Hessians. We find that the weights converge to flat regions around saddle points, where a large fraction of the Hessian eigenvalues are both positive and negative. During training, an increasing fraction of Hessian eigenvalues concentrates around zero, whose magnitude decreases. At lower bit-widths, the magnitude of eigenvalues in the Hessian spectrum is significantly smaller. To mitigate these issues, we propose an algorithm called WinQ to accelerate QAT, which involves: (1) periodically resetting weights to the linear interpolation of full-precision and quantized weights, reducing the distance to the quantization grid and increasing eigenvalue magnitude, and (2) computing gradients of noise-injected weights to regularize the Hessian. Extensive experiments show that WinQ accelerates QAT by up to 4 times across various quantization methods and models. Under the same training cost, WinQ improves state-of-the-art sub-4-bit quantization by up to 8.8%. These results are consistent across 16 settings with different language models, quantization methods, and bit widths.
Science Journals
arXiv:2605.17339v1 Announce Type: new
Abstract: Sparse matrix reordering can significantly reduce the fill-in during matrix factorization, thereby decreasing the computational and storage requirements in sparse matrix computations. Finding a minimal fill-in ordering is known to be an NP-hard problem. Moreover, there is a paradox: matrix reordering is applied before matrix factorization, but fill-ins that matrix reordering methods aim at are generated from matrix factorization. To bridge the gap between reordering and factorization, we propose a deep learning framework to minimize a fill-in surrogate function based on spectral embedding. First, we employ a multi-grid-like GNN architecture to learn to approximate the smallest eigenvectors of its graph Laplacian matrix, i.e. spectral embedding, and capture the global structural information of the matrix. Then, another multi-grid-like GNN architecture is used to minimize the potential space where fill-in can occur based on the rank distribution. Experimental results indicate that our approach achieves competitive performance compared with traditional graph-theoretic algorithms and deep learning methods.
arXiv:2405.19189v3 Announce Type: replace
Abstract: With the great success of diffusion models (DMs) in generating realistic synthetic vision data, many researchers have investigated their potential in decision-making and control. Most of these works utilized DMs to sample directly from the trajectory space, where DMs can be viewed as a combination of dynamics models and policies. In this work, we explore how to decouple DMs' ability as dynamics models in fully offline settings, allowing the learning policy to roll out trajectories. As DMs learn the data distribution from the dataset, their intrinsic policy is actually the behavior policy induced from the dataset, which results in a mismatch between the behavior policy and the learning policy. We propose Dynamics Diffusion, short as DyDiff, which can inject information from the learning policy to DMs iteratively. DyDiff ensures long-horizon rollout accuracy while maintaining policy consistency and can be easily deployed on model-free algorithms. We provide theoretical analysis to show the advantage of DMs on long-horizon rollout over models and demonstrate the effectiveness of DyDiff in the context of offline reinforcement learning, where the rollout dataset is provided but no online environment for interaction.
arXiv:2605.17489v1 Announce Type: new
Abstract: Face Image Quality Assessment (FIQA) is a crucial control step in biometric pipelines. It ensures only reliable samples are processed to maintain system accuracy. State-of-the-art FIQA methods achieve high utility but typically operate as "black boxes." They produce scalar scores without human-interpretable justifications. This lack of transparency limits their effectiveness in human-in-the-loop scenarios, such as automated border control, where actionable feedback is essential. In this paper, we investigate the potential of off-the-shelf Vision-Language Models (VLMs) to bridge this gap by performing FIQA in a zero-shot setting. We present a comprehensive evaluation framework for assessing VLM performance. This involves benchmarking traditional FIQA methods through error-versus-reject curves. Additionally, using a diverse set of datasets, ranging from surveillance-oriented to synthetically generated, we analyzed their interpretability, consistency, and robustness to prompt changes. Our results show biometric utility performance depends significantly on architecture, not merely on parameter count. Most VLMs' outputs align with those of traditional methods. We also find that VLM ranking performance and the generated scores may vary across prompts. Our synthetic ablation study shows that while increasing the parameter count can improve internal consistency, it yields worse degradation-detection performance than smaller models. These findings suggest that zero-shot FIQA score estimation using VLMs is promising and could effectively complement conventional FIQA pipelines as an interpretability module. The codes are available at https://github.com/ThEnded32/VLM4FIQA.git.
arXiv:2605.18227v1 Announce Type: new
Abstract: Pseudo-random number generators (PRNGs) are widely used in modern computing and are expected to exhibit excellent statistical performance and repeatability. This study evaluates and compares modern PRNGs used in high performance computing and artificial intelligence. Our selections comes from different families, including Xoshiro, Philox, PCG, and MRG32k3a. We systematically assess the quality of these generators; instead of testing a single stream for each generator, we test more than 10 3 streams with the BigCrush battery form the TestU01 library. The results, involving more than 4.5 years of cumulative computing time, are analyzed against the claims made by the generators' creators. The highest success rate is 72%, and all tests have been failed by almost every generator, the failed tests are documented. To ensure fairness, all tests are conducted under consistent conditions and are designed to closely simulate real-world usage. The results of each test are available, usable and reproducible with a git repository.
arXiv:2510.24680v2 Announce Type: replace
Abstract: While imitation learning (IL) has enabled successful visual navigation in many common environments, IL policies are prone to unpredictable failures under out-of-distribution (OOD) scenarios. This necessitates failure-resilient policies, which not only detect failures, but also recognise their sources and recover from them autonomously. We propose InFeR, a general framework for building IL policies with informed failure resilience without failure or recovery demonstrations. InFeR retrains an IL policy with a Variational Information Bottleneck (VIB) loss to structure its latent space for OOD failure detection. It applies a visual explainability technique, Grad-CAM, to localise an image region as the source of failure and inform a heuristic policy for recovery. All these are achieved without requiring additional training data. Real-world experiments show that InFeR enables informed failure recovery across two different policy architectures, yielding robust long-range navigation in complex environments.
arXiv:2605.17969v1 Announce Type: new
Abstract: Despite rapid advances in text-to-image generation, faithfully realizing user intent remains challenging, often requiring manual multi-turn trial and error. To automate this process, existing systems rely on either simple prompt rewriting or closed-loop agents driven by hand-crafted rules, rather than learning to adapt actions to the evolving generation process. In this paper, we reformulate image generation as a state-conditioned action-making problem and propose Generation Navigator, a multi-turn T2I agent that learns to dynamically steer the generation trajectory and output the next action. However, training this agent via reinforcement learning introduces a critical credit assignment challenge: naively rewarding a trajectory based solely on a single state assigns equal credit to all actions in the rollout, ignores the quality dynamics across turns, and fails to distinguish actions that improve the trajectory from those that degrade it or waste turns without progress. We resolve this with PRE-GRPO (Peak-Retention-Efficiency Group Relative Policy Optimization), a trajectory-level reinforcement learning objective that explicitly rewards discovering a high-quality image (Peak), avoiding subsequent quality degradation across turns (Retention), and minimizing unnecessary turns (Efficiency). Experiments show substantial improvements across benchmarks, reaching a WISE score of 0.90 and 79.06% reasoning accuracy on T2I-ReasonBench.
arXiv:2512.11446v3 Announce Type: replace
Abstract: Driver fatigue remains a leading cause of road accidents, responsible for 24% of crashes. While yawning serves as an early behavioral indicator of fatigue, existing approaches face significant challenges due to the presence of systematic noise in video-annotated datasets arising from coarse temporal annotations. Training robust machine learning (ML) models requires rich supervisory labels that help learn salient features from the training data. Moreover, efficient on-device training and inference of models on edge devices is crucial in driver fatigue detection tasks to enable accurate real-time decisions on vehicles without reliance on cloud infrastructure. To address this issue, we develop a semi-automated labeling pipeline with human-in-the-loop verification to annotate YawDD videos to YawDD+ frame-level annotations, enabling more accurate model training on edge platforms such as NVIDIA Jetson NANO. Training the established MNasNet classifier and YOLOv11 detector architectures on YawDD+ improves frame accuracy by up to 6% and mAP by 5% over video-level supervision, achieving 99.34% classification accuracy and 95.69% detection mAP on Jetson NANO and AGX. Moreover, MNasNet completed the epoch time in just 8.69 min/epoch while delivering up to 115 frames-per-second (FPS) inference time on AGX, confirming that enhanced data quality alone supports on-device driver fatigue monitoring systems without server-side computation. The YawDD+ dataset and trained models are available online.
arXiv:2605.18430v1 Announce Type: new
Abstract: Text-to-CAD generation aims to create parametric CAD models from natural language, enabling rapid prototyping and intuitive design workflows. However, existing benchmarks focus on basic primitives and simple sketch-extrude sequences, lacking advanced features essential for real-world applications and covering only traditional mechanical parts. We introduce Text2CAD-Bench, the first benchmark systematically evaluating text-to-CAD across geometric complexity and application diversity. Our benchmark comprises 600 human-curated examples spanning four levels: L1-L2 cover fundamental geometry with standard features, L3 introduces complex topology and freeform surfaces, and L4 extends to real-world domains beyond mechanical parts. Each example pairs dual-style prompts -- geometric descriptions mimicking non-expert users, and procedural sequences aligned with expert-level conventions. Evaluating mainstream general LLMs and domain-specific models, we find that current models perform reasonably on basic geometry but degrade substantially on complex topology and advanced features. We release our benchmark to drive progress in text-to-CAD research.
arXiv:2605.17526v1 Announce Type: new
Abstract: As autonomous coding agents become capable of handling increasingly long-horizon tasks, they have gradually demonstrated the potential to complete end-to-end software development. Although existing benchmarks have recently evolved from localized code editing to from-scratch project generation, they remain confined to structurally simplified, single-stack applications. Consequently, they fail to capture the heterogeneous environments, full-stack orchestration, and system-level complexity of real enterprise Software as a Service (SaaS) systems, leaving a critical gap in assessing agents under realistic engineering constraints. To fill this gap, we introduce SaaSBench, the first benchmark designed to explore the boundaries of AI agents in enterprise SaaS engineering. Spanning 30 complex tasks across 6 SaaS domains with 5,370 validation nodes, it incorporates 8 programming languages, 6 databases, and 13 frameworks to meticulously mirror real-world software heterogeneity. Furthermore, we design a dependency-aware hybrid evaluation paradigm tailored for complex systems with long horizons and multi-component coupling, enabling fine-grained, reproducible assessment. Crucially, our extensive experiments reveal a striking insight: the primary bottleneck for state-of-the-art agents is not generating isolated code logic, but successfully configuring and integrating a multi-component system. Over 95\% of task failures occur before agents even reach deep business logic, with models often falling victim to overconfidence and prematurely halting during foundational system setup, or getting trapped in ineffective debugging loops. We hope SaaSBench serves as a practical and challenging testbed to drive the evolution of reliable, system-level coding agents. The code is available at \url{https://github.com/ShadeCloak/SaaSbench}.
arXiv:2603.23672v2 Announce Type: replace
Abstract: Biological sensory systems are inherently adaptive, filtering out constant stimuli and prioritizing relative changes, likely enhancing computational and metabolic efficiency. Inspired by active sensing behaviors across a wide range of animals, this paper introduces a principled 1D event-based visual servoing framework for ground robots operating in structured environments. Utilizing a Dynamic Vision Sensor (DVS), we demonstrate that by applying a fixed spatial kernel to the asynchronous event stream generated from structured logarithmic intensity-change patterns, the resulting net event flux analytically isolates specific combinations of kinematic states. We establish a generalized theoretical bound for this event rate estimator and show that linear and quadratic spatial profiles isolate the robot's velocity and position-velocity product, respectively. Leveraging these properties, we employ a multi-pattern stimulus to directly synthesize a nonlinear state feedback term entirely without traditional state estimation. To overcome the inescapable loss of linear observability at equilibrium inherent in event sensing, we propose a bio-inspired active sensing limit-cycle controller. Experimental validation on a 1/10-scale autonomous ground vehicle confirms the efficacy, extreme low-latency, and computational efficiency of the proposed direct-sensing approach.
arXiv:2605.18554v1 Announce Type: new
Abstract: Federated Bayesian neural networks require fixing a prior on the model parameters together with a likelihood. Eliciting meaningful priors on the weight space of modern overparameterized models is notoriously difficult, and misspecification of either component can severely degrade accuracy and calibration. Motivated by the rapid progress of predictive models such as large language models, the martingale posterior, also known as predictive Bayes, replaces the prior--likelihood pair with a predictive distribution and recovers parameter uncertainty by repeatedly drawing predictive samples and refitting the model. A direct federated implementation, however, would require clients to share the local data sets. This letter proposes {federated martingale posterior} (FMP) sampling, a one-shot embarrassingly parallel protocol in which each client uploads a small set of trainable data embeddings and the server runs the predictive sampler centrally. Experiments on MNIST, CIFAR-10, and CIFAR-100 show that FMP closely matches the centralized counterpart and significantly improves calibration over consensus-style baselines.
arXiv:2601.05653v2 Announce Type: replace
Abstract: Existing traffic simulation frameworks for autonomous vehicles typically rely on imitation learning or game-theoretic approaches that solve for Nash or coarse correlated equilibria, implicitly assuming perfectly rational agents. However, human drivers exhibit bounded rationality, making approximately optimal decisions under cognitive and perceptual constraints. We propose EvoQRE, a principled framework for modeling safety-critical traffic interactions as general-sum Markov games solved via Quantal Response Equilibrium (QRE) and evolutionary game dynamics. EvoQRE integrates a pre-trained generative world model with entropy-regularized replicator dynamics, capturing stochastic human behavior while maintaining equilibrium structure. We provide rigorous theoretical results, proving that the proposed dynamics converge to Logit-QRE under a two-timescale stochastic approximation with an explicit convergence rate of O(log k / k^{1/3}) under weak monotonicity assumptions. We further extend QRE to continuous action spaces using mixture-based and energy-based policy representations. Experiments on the Waymo Open Motion Dataset and nuPlan benchmark demonstrate that EvoQRE achieves state-of-the-art realism, improved safety metrics, and controllable generation of diverse safety-critical scenarios through interpretable rationality parameters.
arXiv:2605.18556v1 Announce Type: new
Abstract: Embodied control increasingly requires models to follow compositional language instructions while reasoning over dynamic visual states. However, current vision-language-action policies and world-action models often couple linguistic knowledge with visual computation in a shared backbone or conditioning pathway, leading to modality competition and making knowledge extension dependent on backbone updates. In this paper, we introduce Key-Gram, a conditional-memory framework that separates language-derived world knowledge from visual-state reasoning for embodied control. At its core is a memory module that decomposes an instruction into task-specific key-grams, retrieves static linguistic priors through deterministic hashed lookup, and injects the retrieved entries into selected hidden layers through context-aware gating and lightweight convolutional fusion. This design allows the backbone to devote its main capacity to visual reasoning and action inference, while reusable instruction knowledge is stored in an extensible external memory. The logical memory table can be conveniently partitioned during training and, due to its $O(1)$ lookup pattern, efficiently placed on host memory during inference. Across RoboTwin2.0, LIBERO/LIBERO-Plus, and real-world dual-arm manipulation, Key-Gram consistently improves both $\pi_{0}$ and $\pi_{0.5}$ backbones, with average relative gains of $29.5\%/9.9\%$ on RoboTwin2.0, $35.8\%/4.5\%$ on LIBERO-Plus transfer without target-domain fine-tuning, and $15.4\%/8.1\%$ on real-world long-horizon tasks. These results demonstrate that externalized linguistic memory provides an effective and extensible mechanism for improving compositional grounding, transfer, and real-world manipulation.
arXiv:2605.17967v1 Announce Type: new
Abstract: This paper explores a scientific question in supervised fine-tuning (SFT): why SFT is broadly effective for small-scale deep neural networks, yet can produce inconsistent or even detrimental effects when applied to large language models (LLMs). Recent advances in interaction-based explanations suggest that interactions between words/tokens provide a faithful metric for quantifying the inference patterns encoded by LLMs. We find that the evolution of interactions during SFT can effectively explain the inconsistent effectiveness of SFT for LLMs. Specifically, we find that (1) SFT primarily removes noise-like interactions, while rarely acquiring reliable new interactions. (2) This denoising stage is extremely brief, after which continued fine-tuning tends to introduce overfitted interactions. We validate these findings across multiple LLMs and datasets. Our findings provide new insights into early stopping and offer practical guidance for LLM training.
arXiv:2605.17530v1 Announce Type: new
Abstract: Network intrusion detection systems play a vital role in protecting networks by detecting malicious network traffic which can then be investigated by a cybersecurity operations centre. State-of-the-art approaches utilise supervised machine learning methods to train a classification model to recognise known cyberattacks; however, these models require a large labelled dataset to train and show poor performance when trained on smaller datasets. In an attempt to address this shortcoming, anomaly detection models learn the distribution of benign traffic and flag non-conforming traffic as malicious. While these methods do not require malicious examples to train, they suffer from high false-positive rates rendering them impractical. As a result, networks may be particularly vulnerable when there are insufficient labelled instances of a specific attack class to train an effective classifier. This often occurs in newly established networks or when previously unseen types of attacks emerge. To address this challenge, this work proposes the use of a triplet network, utilising online triplet mining and a KNN classifier, which is able to perform few-shot classification, enabling effective intrusion detection after being trained on a limited number of malicious examples. Various online triplet mining algorithms were explored and model design choices, such as the inference algorithm and optimised distance metrics, were compared and evaluated through a series of ablation studies. The final model was compared against other state-of-the-art approaches in few-shot binary and multiclass classification, where the proposed approach was found to be competitive with existing methods when trained on as little as 10 malicious samples of each class.
arXiv:2605.17533v1 Announce Type: new
Abstract: This letter proposes a distributed 3D leader-follower formation (3D-LFF) control framework for multi-UAV systems that achieves formation tracking while enforcing perception safety constraints. Maintaining safe, vision-based 3D-LFF is challenging because onboard cameras impose strict Field-of-View (FOV) limitations, and demanding formation commands can drive the leader outside the follower's camera frustum, resulting in loss of visibility. To address this issue, we develop a perception-aware safe control architecture that guarantees visibility by construction. First, we derive a relative kinematic model in a line-of-sight coordinate representation and design a distributed 3D-LFF tracking controller using only locally available relative states. Next, we embed the nominal formation controller within a Control Barrier Function-based Quadratic Program (CBF-QP) safety filter that minimally modifies the commanded velocities to maintain the leader inside the follower's camera frustum while preserving formation tracking whenever feasible. Gazebo simulations and Crazyflie hardware experiments validate the proposed approach, demonstrating accurate formation tracking and effective FOV enforcement, including scenarios in which the nominal desired formation conflicts with visibility constraints.
arXiv:2605.17535v1 Announce Type: new
Abstract: Legacy modernization breaks business logic. Most tools and LLM-based approaches treat modernization as syntax translation, losing implicit rules, edge-case handling, and cross-module constraints. We present AgentModernize, a multi-agent framework that treats modernization as a behavioral preservation problem. Four specialized agents handle extraction, specification, code generation, and validation. The key intermediate artifact -- a Behavioral Specification Graph (BSG) -- forces extracted business logic to be explicit and inspectable before any code is generated. We evaluated on LegacyModernize-8, eight scenarios spanning telecom and banking, using three models (GPT-4o-mini, GPT-4o, GPT-5.3-codex) under a fair protocol: same gold-standard tests, 3 trials, temperature 0.0. Full AgentModernize with feedback was the only configuration with non-zero mean BER under every backbone. SP-LLM and CoT-LLM scored 0.0% on every scenario, on every backbone. AgentModernize without feedback scored 0.0% mean BER with GPT-4o-mini and GPT-5.3-codex; under GPT-4o it achieved non-zero BER only on S1 (44.4%; 5.6% mean over scenarios). Mean BER for full AgentModernize was 9.4% (mini), 8.1% (GPT-4o), and 19.4% (codex). The BSG captures 91.2% of gold-standard rules, confirming that the bottleneck is code generation, not extraction.
arXiv:2605.17538v1 Announce Type: new
Abstract: This letter investigates the problem of output synchronisation in heterogeneous dynamical networks with nonlinear diffusive couplings in the presence of disturbances on the coupling links. By exploiting relative dissipativity properties between adjacent agents, distributed conditions are established to guarantee output synchronisation. Specifically, these conditions can be verified using only local information associated with neighbouring agents and coupling links. As an illustration, a heterogeneous network of Goodwin oscillators is considered, where the relative dissipativity properties between neighbouring oscillators are characterised and used to analyse synchronisation.
arXiv:2605.17541v1 Announce Type: new
Abstract: We present a semi-analytical model to analyze multi-tile metasurface antennas consisting of a set of metasurface tiles and a practical power-dividing network that excites the tiles. The metasurface tiles consist of arrays of rectangular waveguides with subwavelength metamaterial radiators etched into their top walls, each of which can be accurately modeled as polarizable dipoles. The feed structure for the arrays comprises a slotted waveguide attached to their bottom wall, with coupling slots inserted into the common wall that are likewise modeled as polarizable dipoles. The proposed semi-analytical model employs a coupled-dipole framework that accurately captures dipolar interactions among constituent elements within the metasurface tiles, along with a multi-port network analysis technique that accounts for electromagnetic interactions between the tiles and the power divider, thereby forming a self-consistent formulation. The proposed model enables the prediction of key performance metrics, including overall S-parameters, radiation patterns, and gain, and is validated through full-wave numerical simulations. By significantly reducing the computational complexity associated with electrically large apertures, the proposed framework enables rapid and efficient modeling of the overall structure, thereby facilitating iterative optimization. The proposed model has potential applications as an efficient forward model for the design of wireless systems requiring large-aperture metasurface antennas, including remote sensing and next-generation wireless communication networks.
arXiv:2605.17542v1 Announce Type: new
Abstract: This paper proposes a systematic and explicit quantum circuit framework for solving advection-diffusion equations with boundary conditions, based on the Linear Combination of Hamiltonian Simulations (LCHS) method. By employing the Finite Volume Method (FVM) combined with various flux construction schemes, we elaborate the design of quantum circuits tailored explicitly for Robin boundary conditions (including Dirichlet and Neumann boundary conditions as special cases) and periodic boundary conditions. In contrast to prior works on quantum simulation of advection-diffusion equations, we present a detailed error analysis for the linear combination of unitaries (LCU) induced by the constructed quantum circuits. A comprehensive gate complexity analysis demonstrates the quantum advantages over classical computing in high-dimensional scenarios. We simulate the proposed circuits on a fault-tolerant emulator, and numerical results validate the effectiveness of the proposed framework across homogeneous, inhomogeneous, and high-dimensional cases. The proposed framework is compatible with numerous spatial discretization methods and numerical schemes, extends naturally to other linear PDEs, and establishes a practical foundation for solving large-scale PDE problems on future fault-tolerant quantum computers.
arXiv:2605.17336v1 Announce Type: new
Abstract: Tactile sensing is a fundamental modality for embodied intelligence, offering unique and direct feedback on contact geometry, material properties, and interaction dynamics that remote sensors cannot replace. However, unimodal tactile perception is inherently limited by its sparse spatial coverage and lack of global semantic context. With the recent explosion in deep learning and large language models, integrating tactile with vision and language has become essential to bridge physical interaction with semantic reasoning, leading to the emergence of Multimodal Tactile Fusion. Despite rapid progress, the existing researches remain fragmented across disparate datasets, sensing modalities, and tasks, lacking a unified theoretical framework. To address this gap, this paper provides a comprehensive survey of multimodal tactile fusion research up to the first quarter of 2026. We propose a hierarchical taxonomy that organizes the field into two primary dimensions: multimodal datasets and multimodal methods. On the data side, we categorize resources ranging from Tactile-Vision datasets, Tactile-Language datasets, Tactile-Vision-Language datasets, and Tactile-Vision-Other datasets. On the method side, we structure prior work into three core pillars: (1) Multimodal Perception and Recognition, which focuses on object understanding and grasp prediction; (2) Cross-Modal Generation, focusing on bidirectional translation between tactile, vision, and text; and (3) Multimodal Interaction, emphasizing feedback control and language-guided manipulation. Furthermore, we summarize representative tactile sensing hardware, review commonly used evaluation metrics and benchmark settings, and discuss current challenges and promising future directions.
arXiv:2605.17858v1 Announce Type: new
Abstract: This paper proposes a Transformer-based hybrid beamforming framework for reconfigurable pixel antenna (RPA)-equipped massive multiple-input multiple-output (MIMO) in high-altitude platform station (HAPS) communications. The proposed pattern reconfigurable hybrid beamforming network (PR-HBFNet) comprises two key components: 1) a pattern reconfigurable network that leverages a Transformer encoder to determine the radiation pattern for each antenna element, and 2) a hybrid beamforming network that employs model-driven residual learning to compute analog and digital precoders over SVD-based initializations. Simulation results demonstrate that the proposed PR-HBFNet closely approaches the spectral efficiency of a greedy benchmark while significantly reducing computational complexity.
arXiv:2605.17562v1 Announce Type: new
Abstract: EEG foundation models (EEG-FMs) have been evaluated predominantly on clean, in-distribution accuracy, leaving their robustness, interpretability and representational quality largely unexamined. This study addresses these gaps by benchmarking six EEG-FMs against a baseline deep learning model across eight datasets. Beyond clean accuracy, we conduct three layers of analysis: (i) Robustness: we apply test-time perturbations including additive noise, random and region-based channel dropout and region-specific noise injection. Our analyses show that no single model dominates all failure modes. The most noise-robust model is among the most fragile under channel dropout and much of the dropout fragility disappears when channels are removed rather than zero-padded. (ii) Interpretability: we present the first application of Attention-Aware Layer-Wise Relevance Propagation (AttnLRP) to EEG-FMs and show that models broadly concentrate relevance on task-appropriate brain regions consistent with known neurophysiology. However, attribution maps remain spatially stable under perturbation while predictions degrade, suggesting that the models attend to the correct brain regions but decode corrupted content. (iii) Expressiveness: With block-wise probing we show that late blocks are repurposed during fine-tuning, while early blocks already hold task-related information. Furthermore, we demonstrate that the poor head-only performance previously attributed to low-quality pre-trained representations is largely explained by pooling and that EEG-FMs possess sufficient representational capacity when their token-level embeddings are preserved. Together, these findings provide the first systematic assessment of robustness, interpretability and expressiveness for EEG-FMs and highlight critical considerations for their development.
arXiv:2603.27341v3 Announce Type: replace
Abstract: Recent Artificial Intelligence (AI) models have matched or exceeded human experts in several benchmarks of biomedical task performance, but surgical benchmarks in particular are often missing from prominent medical benchmark suites. Since surgery requires integrating disparate tasks, generally-capable AI models could be particularly attractive as a collaborative tool if performance could be improved. On the one hand, the canonical approach of scaling architecture size and training data is attractive, especially since there are millions of hours of surgical video data generated per year. On the other hand, preparing surgical data for AI training requires significantly higher levels of professional expertise, and training on that data requires expensive computational resources. These trade-offs paint an uncertain picture of whether and to-what-extent modern AI could aid surgical practice. In this paper, we explore this question through a case study of surgical tool detection using state-of-the-art AI methods available in 2026. We demonstrate that even with multi-billion parameter models and extensive training, current Vision Language Models fall short in the seemingly simple task of tool detection in neurosurgery. Additionally, we show scaling experiments indicating that increasing model size and training time only leads to diminishing improvements in relevant performance metrics. Thus, our experiments suggest that current models could still face significant obstacles in surgical use cases. Moreover, some obstacles cannot be simply ``scaled away'' with additional compute and persist across diverse model architectures, raising the question of whether data and label availability are the only limiting factors. We discuss the main contributors to these constraints and advance potential solutions.