Science & Mathematics

Random Linear Network Coding (RLNC), and in particular Sliding Window RLNC (SW-RLNC), is a promising approach for reliable, low-latency communication in dynamic environments such as 6G and non-terrestrial networks. While prior work has largely focused on encoder-side parameter tuning, such as the code rate and coding window size, decoder-side design remains comparatively underexplored despite its direct impact on end-to-end performance. This paper addresses decoding window management, namely the policies governing window-boundary updates, removal of obsolete packets, and delivery of recovered packets to upper layers, and shows that these policies critically shape the trade-off between computational complexity and decoding performance, particularly in-order delivery delay. We derive sufficient conditions under which packets can be safely removed from the decoding window without compromising decodability or violating in-order delivery guarantees, and, building on these conditions, propose two lightweight decoding window management schemes. Through extensive simulation-based evaluation across a range of channel conditions and code rates, we show that both schemes preserve decoding efficiency while substantially reducing decoding complexity, and that one additionally achieves a marked reduction in in-order delivery delay. These results establish decoding window management as a key design dimension for practical SW-RLNC systems.

This study compares ten digital 3D acquisition techniques applied to violin family instrument plates, assessing their accuracy and relevance for scientific and cultural heritage applications. Evaluated methods include two photogrammetric workflows, medical and industrial computed tomography, terrestrial laser scanning, two structured-light configurations (blue light and infrared), and three consumer-grade iPad Pro approaches: TrueDepth scanning and LiDAR using two different applications. The methods were applied to five historical instruments from the "Musical Instruments Museum" (Brussels) and the "Musée de la Musique" (Paris). Since no geometric ground truth is available, accuracy is assessed through systematic pairwise point-to-plane comparisons of the resulting point clouds. These comparisons are jointly analysed to estimate the intrinsic error of each technique (i.e. the expected discrepancy between the model and the true but unknown geometry) without designating any technique as reference. Results are discussed with respect to intended uses, offering recommendations for techniques depending on the dedicated objectives.

Reliable microstructural parameter estimation in diffusion MRI (dMRI) critically depends on the acquisition protocol. Yet, common protocol design criteria may not fully reflect estimator bias, parameter constraints, and nonlinear fitting behavior in complex inverse problems. Here, we introduce Protocol Optimization via Loss Assessing Parameter Recovery (POLAR), an end-to-end framework that searches the continuous acquisition space and directly minimizes multiparameter recovery error through noisy signal simulation and the chosen model fitting. We demonstrate POLAR spanning distinct microstructural environments: the Cylinder-Zeppelin model for white matter, Neurite Exchange Imaging (NEXI) for gray matter, and Cellular Exchange Imaging (CEXI) for permeable cells. Across independent held-out parameter sets, POLAR consistently outperformed model-specific reference designs, reducing RMSE by up to $50\%$ for Cylinder-Zeppelin and $74\%$ for CEXI, while reducing NEXI exchange time bias by $70\%$. These gains across distinct biophysical models, acquisition settings, and reference design strategies demonstrate the benefit of optimizing acquisition protocols through the downstream parameter-estimation problem. POLAR provides a flexible, model-specific and estimator-aware framework for designing dMRI protocols tailored to the microstructural parameters and experimental conditions of interest.

We present two Transformer-based methods for Task A of the 1st DAFx Parameter Estimation Challenge, which requires estimating six effective physical parameters of a synthetic plate-reverb model from its impulse response (IR). Method A1 combines an Audio Spectrogram Transformer encoder and Transformer regressor with differentiable IR refinement. Method A2 uses the same encoder to condition a continuous normalizing flow and refines sampled candidates using particle swarm optimisation (PSO) and gradient polishing. Both methods preserve the absolute IR scale to recover surface density. On a synthetic holdout set of 100 IRs, both refinement procedures reduce waveform and parameter errors by more than three orders of magnitude relative to the unrefined neural outputs. The PSO-based pipeline achieves the lowest errors, indicating near-perfect recovery in this matched synthetic setting.

Pointing error is critical in highly directional communication systems, where even slight misalignment leads to severe received power reduction. Existing pointing error models fail to cover diverse practical scenarios due to restrictive distributional and design limitations. This paper proposes a novel pointing error model incorporating jitters, AoD/AoA estimation errors, and comprehensive design impacts, including phase shifter quantization, antenna patterns, array design, beamforming architectures, steering capabilities, and bandwidth. To accomplish this, antenna patterns are modeled for both high- and low-gain configurations. Then, the array factor's main lobe is approximated as a Gaussian beam for subarrays and superarrays, accommodating analog and digital beamforming stages in wideband systems under beam squint presence or compensation. Jitters and estimation errors are incorporated via 3D rotations, while phase shift quantization is captured through the phase excitations of the arrays. It is shown that the resulting pointing error follows a novel distribution: the negative log-shifted Gamma (NLSG). Fully characterized by only shape, scale, and location parameters, the NLSG features closed-form statistical expressions. This formulation ensures tractable mathematical manipulation, making it a practical, scalable tool for system design and analysis. Validation via normalized moment error, Jensen-Shannon divergence, and Kolmogorov-Smirnov statistics confirms exceptional accuracy, yielding values primarily below 1%, 0.02, and 0.05, respectively. Furthermore, capacity analysis has demonstrated that specific design choices are paramount for achieving optimal capacity. Finally, important guidelines for designing highly directional wideband communication systems are provided to mitigate the detrimental effects of pointing errors, thereby paving the way for robust next-generation networks.

Scenario planning is a foresight practice in which analysts construct a small set of plausible futures and prepare an organisation for each. Each scenario is a narrative that must be plausible, internally consistent, relevant and challenging, and the set as a whole must be differentiated. The practice remains manual: a workshop (i.e., a facilitated meeting in which a small group of experts writes the narratives) considers only the factors it has time for, carries the biases of the experts who run it, and cannot be redone whenever conditions change. In this paper, we automate scenario planning by casting it as a diversity planning problem, in which two scenarios differ when they end in different narratives on the analyst's drivers. The result is a set of scenarios, one for each narrative that the topic can reach within a budget. Our implementation, Polyscene, is a generic engine that generates scenarios from a topic. We use geopolitical conflicts and shipping-route disruptions as case studies. For the first, we build the planning problem's initial state from the GDELT Project so it reflects current relationships among the countries involved. Polyscene then generates scenarios showing how relationships among those countries change. With a board of six countries as of 1 January 2021 and a budget of five events, the conflict set puts five of the 15 country pairs at war, and each of the three wars that break out among these countries by 1 January 2026 ends a scenario. For the second, we build the initial state from ship traffic, a supply-pressure index, and the news, and Polyscene generates scenarios showing what a closed or diverted route could do to a manufacturer's parts stock. With four routes per board and the same date and budget, each set includes three outcomes, and the Suez Canal set includes the March 2021 blockage and the December 2023 Red Sea container diversion.

Structural Causal Models (SCMs) provide a principled framework for observational, interventional, and counterfactual reasoning, all of which rely on probabilistic inference. Existing frameworks either transform SCMs into alternative tractable representations or rely on approximate inference with expressive generative models. We instead perform inference using a message passing scheme directly on the SCM by representing structural mechanisms and probability densities in a common finite-dimensional function space. Forward propagation combines a numerical realization of the Law of the Unconscious Statistician, while propagated densities are recovered by spline projection from LOTUS moments. The same representation naturally yields backward message passing for conditioning, providing a unified framework for causal inference. Using tensor-product B-spline bases, the resulting algorithm expresses inference through numerical quadrature and linear algebra without explicitly constructing conditional distributions. Experiments on randomly generated continuous SCMs demonstrate accurate approximation of Monte Carlo reference solutions and systematic improvement with increasing basis resolution.

Large Language Models (LLMs) are increasingly used for interpersonal advice, potentially shaping how users interpret ambiguous social encounters. Existing Theory of Mind evaluations typically rely on predefined answers, while uncertainty research overlooks LLMs' preferred explanations and confidence under psychological distress. To address these gaps, we introduce MentalLoom, a novel evaluation of LLM mentalisation comprising 604 ambiguous social episodes, matched depression and social anxiety disclosures, and four-turn dialogues with unchanged event evidence. Across eight LLMs, disclosures initially elicited more negative explanations overall but more positive narrator-directed interpretations. Scripted repeated questioning showed greater emotional support, stronger positive narrator-directed endorsement, and higher certainty without new event evidence. These findings show that user context can reshape interpretive preferences and commitment while the underlying ambiguity remains unresolved.

Reach-avoid sets, defining the states from which a task can be completed while avoiding unsafe regions, are fundamental tools for deploying robots in environments subject to disturbances and uncertainty, as well as for runtime task-feasibility monitoring, allowing prediction of whether a state is safe and will lead to successful task completion. Classical Hamilton-Jacobi Reachability (HJR) and its recent neural extensions (e.g., Lipschitz Continuous Reachability Learning, DeepReach) treat disturbances adversarially, computing the reach-avoid set under worst-case disturbances from a bounded set. This guarantees safety but can yield overly conservative sets when disturbances are stochastic rather than adversarial, and additionally requires the disturbance set to be known or specified in advance. We propose Probabilistic reach-avoid Value Function (PVF), a self-supervised method, applicable to black-box systems, for computing reach-avoid sets as the states with a certain reach-avoid probability under stochastic, potentially unbounded disturbances. We evaluate our approach against HJR-based baselines on systems of increasing complexity, up to a Hopper locomotion task, and show that our probabilistic sets are less conservative and more accurate than adversarial HJR sets when disturbances are non-adversarial.

Gradient clipping is a key tool for obtaining high-probability convergence guarantees for stochastic first-order methods under heavy-tailed noise. Existing analyses typically assume a bounded central moment of order $α\in(1,2]$, which excludes distributions with tails of order $u^{-α}$ at the critical exponent. In this paper, we consider smooth convex optimization under affine heavy-tailed noise, whose scale may grow with the gradient norm. We derive high-probability convergence guarantees for clipped SGD (Clip-SGD) and its accelerated variant Clip-SSTM under a weak-$L^α$ tail condition, which is strictly weaker than the corresponding moment bound. For $1

The single-stage isolated eXtended-functionality rectifier (X-Rectifier) electric vehicle (EV) on-board charger (OBC) employs 650V GaN monolithic bidirectional power transistors (M-BDS) in the input stage, provides sinusoidal input currents and four-quadrant operation in three-phase and split/single phase grids without power derating, operates from asymmetrical three-phase mains, provides reactive power if needed, and supports extremely asymmetric phase loading in standalone vehicle-to-load (V2L) mode. This paper presents a new improved phase-modular control method for the X-Rectifier that enables tracking of arbitrary and independent phase current references regardless of the instantaneous grid voltage (i.e., enables four-quadrant operation). The method supports grid-tied (i.e., grid-to-vehicle, G2V, or vehicle-to grid, V2G) and grid-forming (i.e., V2L) operation. Further, aiming towards a practical realization of the X-Rectifier, the paper addresses implementation aspects of the M-BDS half-bridges, like commutation logic around the grid-voltage zero crossing, fault handling, and soft-switching behaviour. Finally, a realized 5.5 kVA hardware demonstrator of the X-Rectifier is used to validate the proposed control method experimentally for all operating modes, including grid-tied three- and single-phase four-quadrant operation (G2V, V2G, and full reactive power modes). Further, robust V2L operation is verified with real-world loads, i.e., in three-phase mode with asymmetrical loading of the phases including a woodworking power tool, and in single-phase mode with a high-power vacuum cleaner with high inrush current and hence momentary overload of the X-Rectifier demonstrator.

Vision-language-action (VLA) models have achieved strong performance on standard robotic manipulation benchmarks. However, their reliability can degrade substantially when task conditions deviate from the nominal setting. Perturbations in object positions and task instructions may lead to complete task failure. Existing VLA policies generally lack an explicit runtime mechanism for distinguishing these failure modes and selecting an appropriate recovery strategy. To systematically mitigate this issue, we propose FIT-MAC, a failure-aware, closed-loop multi-agent control framework for improving the robustness of VLA policies at inference time. FIT-MAC augments a VLA policy with an inference-time control layer. First, the Planner Agent parses the language instruction into a structured task specification that captures the manipulation goal and spatial constraints, and the Grounder Agent resolves each instruction-level reference to a visual target. Next, the Verifier Agent tracks phase-specific execution progress, distinguishes target-binding errors from local execution failures, and determines whether intervention is warranted. The Correction Agent then routes permitted interventions to target rebinding or bounded grasping or placement recovery, depending on the failure type. When no intervention is required, the pretrained VLA remains the default action generator. Extensive experiments on LIBERO-PRO demonstrate that FIT-MAC improves robustness across multiple VLA backbones, consistently increasing success rates and achieving state-of-the-art performance under diverse perturbations.

Spiking Neural Networks (SNNs) have shown great potential for low-power neuromorphic computing due to their event-driven processing and spatiotemporal sparsity, but they remain vulnerable to input perturbations, hardware-intrinsic noise, and adversarial attacks in real-world deployment. We attribute this vulnerability partly to input-specific hidden representations learned by standard SNNs trained with conventional cross-entropy loss, as our information-plane analysis empirically shows that substantial input-related information can still be retained after label-predictive information becomes stable. To address this issue, we propose SCIB, a simple yet effective robust learning objective for SNNs based on the Conditional Information Bottleneck. Different from conventional variational IB methods, SCIB replaces the mutual information compression term $I(X;S)$ with the conditional mutual information $I(X;S|Y)$ and derives a tractable variational upper bound. Moreover, it introduces a learnable class-conditional prior instead of a label-agnostic isotropic Gaussian prior, improving the expressiveness of variational compression and encouraging class-structured representations when combined with the supervised classification objective. Systematic experiments on Fashion-MNIST, CIFAR-10, and CIFAR-100 under external environmental noise, neuromorphic hardware noise, and adversarial attacks show that SCIB improves robustness over standard SNNs and IB-based baselines under most evaluated noise conditions, improves adversarial accuracy under the evaluated attacks without using adversarial training, and reduces the SynOps-based single-sample inference energy consumption by 24.9% relative to an unconstrained SNN baseline.

Large language model (LLM) agents increasingly execute business operations through tools and interact directly with application data and services. Beyond completing user requests, these agents must obey domain-specific business rules that constrain when actions are permitted and what data or resources may be affected. Violating such rules can lead to business logic bugs, even when a task appears successful. Detecting these bugs is difficult because the rules are rarely available as explicit, testable specifications. They are distributed across system prompts and tool definitions, while others remain implicit in domain knowledge. Existing agent testing approaches typically evaluate task success, predefined policies, or known attack scenarios. This limits their ability to systematically uncover previously unspecified business-rule violations. We present AgenTest, which mines business rules from system prompts, tool definitions, and domain knowledge, refines them into Given-When-Then specifications, and systematically tests each specification. For each extracted specification, AgenTest instantiates its Given conditions by constructing a satisfying environment state and simulated user, exercises the agent through multi-turn interactions, and checks its Then conditions over messages, tool calls, and state changes using a hybrid oracle. On the airline, retail, and telecom agents of tau2-bench with two underlying models, AgenTest produces 551 refined specifications, and detects violations of 89 and 121 distinct specifications, respectively -- 3.4X and 2.5X as many as all baselines combined. Only AgenTest detects cases where an agent improperly accesses or modifies another user's data. Moreover, about 60% of bug-triggering runs cannot be detected from the final state alone, highlighting the need to check the full interaction against explicit business rules.

We introduce OvisOCR3, a 0.8B model for a broad range of OCR-related tasks. OvisOCR3 unifies document parsing, general OCR, key information extraction (KIE), document figure captioning, and cross-page table reconstruction in a single model. We train the model on real and synthetic data through continued pretraining and use domain-balanced mid-training to balance supervision across tasks with different data volumes and response lengths. Multi-teacher on-policy distillation transfers expertise from five domain-specific 4B teachers into the shared model, followed by mixed-task reinforcement learning with domain-specific rewards. We compare OvisOCR3 with general-purpose vision-language models and specialized OCR models on public and in-house benchmarks covering all five task families. In these comparisons, it achieves an overall score of 96.74 on OmniDocBench v1.6 and the highest overall scores on PureDocBench, Real5-OmniDocBench, and MDPBench. On OCRBench v2, it ranks first, scoring 65.9 and 61.3 on the English and Chinese splits, respectively. It also outperforms the baselines on OCRBench and all four KIE benchmarks. On DocFigCapBench, it achieves the highest localization F1 and a caption score close to that of Qwen3.8-Max. For cross-page tables, it leads in Relation F1 on CrossTabBench and truncated-table TEDS on MPDocBench-Parse. These results demonstrate that a compact 0.8B model can achieve strong performance across diverse OCR-related tasks. OvisOCR3 is available at https://huggingface.co/ATH-MaaS/OvisOCR3.